Assumptions register
The quantities that no source provides and that Projet Celsius sets itself, such as how many contracts a lawyer reviews or what share they delegate to AI, each with what it rests on.
Machine operations
1 assumption · 1 entry
Synthesise one minute of voice1 assumption
The 1,200 audio tokens of one minute of voice, at the cost of a written token: no published measurement Source: OpenAI
Uses
47 assumptions · 22 entries
Have a conversation of 10 exchanges1 assumption
The context re-read from cache at each turn: the previous exchanges, and the document already in the conversation if there is one Source: Epoch AI
Write an email1 assumption
The instruction and the message being answered
Carry out deep research4 assumptions
The search engine queries, 30 by default, counted as conventional searches; divided by 1.10, as the conventional search already counts its data centre
The web pages read, about a hundred by default: Google announces up to hundreds of sites Source: Google
The context re-read from cache at each of the 40 steps, 40,000 tokens on average Source: Anthropic
The draft of each step, 500 tokens, then the report of 5 pages
Analyse a spreadsheet2 assumptions
The spreadsheet, 200 rows and 10 columns, and the question Source: Projet Celsius
The analysis code, about 40 lines, then the answer
Carry out an AI-assisted web search2 assumptions
The 2 searches launched by the assistant, counted as conventional searches; divided by 1.10, as the conventional search already counts its data centre
The 5 pages found, read by the model, and the question: web agents visit 6 pages per task on average Source: Zhang and Nam
Read an AI-generated overview at the top of a search1 assumption
A small model reads the excerpts of the top results, about 3,000 tokens, and writes an overview of 150 tokens
Create a large image1 assumption
An image of 4 megapixels Source: Bertazzini et al.
Retouch a photo1 assumption
The image redone, counted as a generated image
Create a visual in 4 attempts1 assumption
The 4 images generated, to keep one
Create a clip of 10 seconds1 assumption
A single continuous video costs as much as the square of its duration: 10 seconds is worth 3.4 times the 5.4-second video that was measured Source: Delavande et al.
Speak in voice mode, 1 minute3 assumptions
What the user says, half of the minute, in audio tokens Source: OpenAI
The conversation re-read from cache, 4 turns per minute
What the AI answers, the other half, in audio tokens Source: OpenAI
Have a coding agent work, 1 hour3 assumptions
The context re-read from cache at each of the 120 steps of the hour: 96,000 tokens, 96% of a call in the median published session Source: Hausfather
The new tokens of each step (files read, command results): 3,600, 3.6% of a call Source: Hausfather
The tokens written at each step, draft included: 900 on average in the published logs Source: Hausfather
Get a code suggestion2 assumptions
The code around the cursor, about 6,000 characters or 125 lines, read by a small model Source: GitHub
The suggestion, about 8 lines, written by a small model
Give a task to a browser agent4 assumptions
A screenshot read at each step, 15 steps: published web agents take 13 on average Source: Zhang and Nam
The instruction and the new text of each page, 500 tokens per step
The agent's instructions and its history, 10,000 tokens re-read from cache at each step
The action chosen at each step and its justification, 250 tokens
Run an automated workflow2 assumptions
The workflow instructions and the email received
A structured output of about 100 tokens: the category and the extracted fields
Review a contract of 20 pages3 assumptions
Read the contract of 20 pages and summarise its risks on one page
The 3 follow-up questions, the contract re-read from cache
Reword the clauses to be reviewed, one page
Prepare a presentation of 10 slides3 assumptions
The text of the 10 slides, about 2 pages
The 3 illustrations generated
The 3 retouching requests
Prepare a one-hour lesson3 assumptions
A conversation of 10 exchanges to build the outline
The lesson material, 2,000 words
The exercises, one page
Carry out a literature search, 10 sources3 assumptions
The 3 assisted searches to find the sources
Each source read and summarised, 10 pages each, 10 sources by default
The synthesis, 1,000 words
Handle a customer request3 assumptions
The automatic sorting of the request
The answer looked up in 3 pages of documentation
The reply email to the customer
Write meeting minutes2 assumptions
Tidy up 2 pages of notes
The email that sends the minutes
Question to Claude, Projet Celsius estimate1 assumption
The database's quick question, 100 tokens read and 400 written, put to the chosen model; Anthropic does not publish the size of its models, so the class of each is an assumption Source: Anthropic
Occupations and activities
370 assumptions · 38 entries
Office worker, 1 year14 assumptions
Write an email : 15 per day; with AI: 3% light, 10% typical, 34% heavy
About 15 emails written per day, correspondence being a daily task for office workers in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 8 per day; with AI: 10% light, 30% typical, 100% heavy
About 8 information needs per day; share done with AI based on the share of users who turn to an AI first to look something up (Crédoc, Baromètre du numérique, 2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Proofread a page : 3 per day; with AI: 8% light, 25% typical, 86% heavy
About 3 texts to rework per day; share done with AI based on the share of work messages that ask for a supplied text to be edited (OpenAI, 2025), adjusted to the occupation. Source: OpenAI
Write a page : 1 per day; with AI: 7% light, 20% typical, 69% heavy
About 1 page written per day (note, letter, procedure), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 5 per day; with AI: 7% light, 20% typical, 69% heavy
About 5 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 3 per week; with AI: 7% light, 20% typical, 69% heavy
About 3 reflection sessions per week (preparing a decision, unblocking a file), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 2 per week; with AI: 10% light, 30% typical, 100% heavy
About 2 sets of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Analyse a spreadsheet : 3 per week; with AI: 5% light, 15% typical, 52% heavy
About 3 spreadsheets to analyse per week, assumption; share done with AI set low, data analysis carrying little weight in messages (OpenAI, 2025), adjusted to the occupation. Source: OpenAI
Summarise 30 pages : 1 per week; with AI: 10% light, 30% typical, 100% heavy
About 1 long document to read per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 1 per week; with AI: 20% light, 59% typical, 100% heavy
About 1 page to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Presentation of 10 slides : 22 per year; with AI: 8% light, 25% typical, 86% heavy
About 2 presentations per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 1 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 5 meetings per week, assumption; with AI in the office suite, most are transcribed and summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it. Source: Microsoft Research
Automated workflow : 40 per day; with AI: 10% light, 30% typical, 80% heavy (if tooling: agents)
About 40 emails received per day, assumption; with agents, a share is sorted and prepared by an automated workflow, assumption with no published survey.
Browser agent : 5 per week; with AI: 5% light, 20% typical, 60% heavy (if tooling: agents)
About 5 online procedures per week (form, booking, statement), assumption; share delegated to an agent set without a published survey.
Accountant, 1 year12 assumptions
Write an email : 12 per day; with AI: 3% light, 9% typical, 33% heavy
About 12 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 5 per day; with AI: 9% light, 28% typical, 98% heavy
About 5 questions per day (accounting rules, tax), assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Analyse a spreadsheet : 2 per day; with AI: 5% light, 14% typical, 49% heavy
About 2 spreadsheets to analyse per day (reconciliations, checks), the core of the occupation in O*NET; share done with AI set low, data analysis carrying little weight in messages (OpenAI, 2025), adjusted to the occupation. Source: O*NET
Proofread a page : 2 per day; with AI: 8% light, 23% typical, 82% heavy
About 2 texts to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 3 per week; with AI: 6% light, 19% typical, 66% heavy
About 3 pages of notes or reports per week, report writing being mostly monthly or annual in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Query a PDF : 2 per week; with AI: 8% light, 23% typical, 82% heavy
About 2 long documents to query per week (tax file, contract, report), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 1 per week; with AI: 9% light, 28% typical, 98% heavy
About 1 regulatory text or report to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 1 per week; with AI: 9% light, 28% typical, 98% heavy
About 1 set of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 2 per day; with AI: 6% light, 19% typical, 66% heavy
About 2 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 12 per year; with AI: 8% light, 23% typical, 82% heavy
About 1 presentation per month (closing, budget), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 0.6 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 3 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Automated workflow : 30 per day; with AI: 10% light, 40% typical, 90% heavy (if tooling: agents)
About 30 items to enter per day (invoices, expense claims), assumption; with agents, a share goes through an automated workflow, assumption with no published survey.
Human resources officer, 1 year11 assumptions
Write an email : 20 per day; with AI: 3% light, 9% typical, 31% heavy
About 20 emails per day, explaining the rules to employees being a daily task in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 5 per day; with AI: 9% light, 27% typical, 94% heavy
About 5 questions per day (employment law, procedures), assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Proofread a page : 2 per day; with AI: 7% light, 22% typical, 78% heavy
About 2 texts to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 1 per day; with AI: 6% light, 18% typical, 63% heavy
About 1 page per day (job description, letter, note), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Query a PDF : 2 per week; with AI: 7% light, 22% typical, 78% heavy
About 2 batches of applications to sort per week, application review being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Write meeting minutes : 4 per week; with AI: 9% light, 27% typical, 94% heavy
About 4 interview reports per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 2 per week; with AI: 6% light, 18% typical, 63% heavy
About 2 delicate situations to prepare per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 2 per day; with AI: 6% light, 18% typical, 63% heavy
About 2 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 12 per year; with AI: 7% light, 22% typical, 78% heavy
About 1 presentation per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 0.8 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 4 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Automated workflow : 10 per day; with AI: 10% light, 30% typical, 80% heavy (if tooling: agents)
About 10 repetitive procedures per day (reminders, certificates), assumption; share delegated to an automated workflow set without a published survey.
Executive assistant, 1 year12 assumptions
Write an email : 30 per day; with AI: 3% light, 10% typical, 36% heavy
About 30 emails per day, mail being handled every day in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 6 per day; with AI: 10% light, 31% typical, 100% heavy
About 6 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Proofread a page : 4 per day; with AI: 9% light, 26% typical, 89% heavy
About 4 texts to rework per day, preparing letters being a daily task in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Write a page : 3 per week; with AI: 7% light, 20% typical, 72% heavy
About 3 pages per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 4 per week; with AI: 10% light, 31% typical, 100% heavy
About 4 sets of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 2 per week; with AI: 9% light, 26% typical, 89% heavy
About 2 presentations to format per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 20% light, 61% typical, 100% heavy
About 2 pages to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Analyse a spreadsheet : 2 per week; with AI: 5% light, 15% typical, 54% heavy
About 2 spreadsheets per week (budget, diary), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 1 per week; with AI: 10% light, 31% typical, 100% heavy
About 1 long document to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 4 per day; with AI: 7% light, 20% typical, 72% heavy
About 4 searches per day (travel, suppliers), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 1.2 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 6 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Browser agent : 3 per week; with AI: 10% light, 30% typical, 70% heavy (if tooling: agents)
About 3 bookings or procedures per week, organising travel being weekly or more often in O*NET; share delegated to an agent set without a published survey.
Manager or executive, 1 year12 assumptions
Write an email : 40 per day; with AI: 3% light, 8% typical, 29% heavy
About 40 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 5 per day; with AI: 8% light, 25% typical, 86% heavy
About 5 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Proofread a page : 2 per day; with AI: 7% light, 21% typical, 72% heavy
About 2 texts to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 3 per week; with AI: 8% light, 25% typical, 86% heavy
About 3 reports to read per week, reviewing results being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Conversation of 10 exchanges : 3 per week; with AI: 5% light, 16% typical, 58% heavy
About 3 decisions or interviews to prepare per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 3 per week; with AI: 8% light, 25% typical, 86% heavy
About 3 sets of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 2 per week; with AI: 5% light, 16% typical, 58% heavy
About 2 pages per week (note, strategy), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Analyse a spreadsheet : 2 per week; with AI: 4% light, 12% typical, 43% heavy
About 2 dashboards to analyse per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 44 per year; with AI: 7% light, 21% typical, 72% heavy
About 1 presentation per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Deep research : 11 per year; with AI: 3% light, 8% typical, 29% heavy
About 1 in-depth question per month (market, competitor), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 2 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 10 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Browser agent : 2 per week; with AI: 5% light, 20% typical, 60% heavy (if tooling: agents)
About 2 online procedures per week, assumption; share delegated to an agent set without a published survey.
Communications officer, 1 year13 assumptions
Write a page : 2 per day; with AI: 8% light, 24% typical, 84% heavy
About 2 pages per day (press release, publication, letter), updating content being a daily task in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Proofread a page : 4 per day; with AI: 10% light, 30% typical, 100% heavy
About 4 texts to rework per day; share done with AI based on the share of work messages that edit a supplied text (OpenAI, 2025), adjusted to the occupation. Source: OpenAI
Write an email : 20 per day; with AI: 4% light, 12% typical, 42% heavy
About 20 emails per day, media requests being a daily task in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 5 per day; with AI: 12% light, 36% typical, 100% heavy
About 5 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Article of 1,000 words : 1 per week; with AI: 12% light, 36% typical, 100% heavy
About 1 long article per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 3 per week; with AI: 8% light, 24% typical, 84% heavy
About 3 brainstorming sessions per week (angle, slogan, campaign), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Create an image : 5 per week; with AI: 6% light, 18% typical, 63% heavy
About 5 supporting visuals per week for social media, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Visual in 4 attempts : 2 per week; with AI: 8% light, 24% typical, 84% heavy
About 2 campaign visuals searched for per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 24% light, 72% typical, 100% heavy
About 2 pages to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Clip of 5 seconds : 1 per week; with AI: 0% light, 5% typical, 30% heavy
About 1 short animation per week for social media, assumption; share done with AI set low, generated video remaining rare at work, assumption.
Web search with an AI : 3 per day; with AI: 8% light, 24% typical, 84% heavy
About 3 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 22 per year; with AI: 10% light, 30% typical, 100% heavy
About 2 presentations per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 0.6 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 3 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Journalist, 1 year11 assumptions
Article of 1,000 words : 4 per week; with AI: 1% light, 5% typical, 25% heavy
About 4 articles per week, writing happening several times a day in O*NET; share done with AI set low, newsrooms' rules limiting the use of AI to write articles, assumption. Source: O*NET
Summarise a meeting : 3 per week; with AI: 25% light, 60% typical, 100% heavy
About 3 hours of interviews to transcribe per week, assumption; share done with AI set high, transcription being the most accepted use in newsrooms, assumption with no French survey reviewed.
Proofread a page : 5 per day; with AI: 10% light, 31% typical, 100% heavy
About 5 texts to rework per day (headline, standfirst, proofreading); share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 8 per day; with AI: 12% light, 37% typical, 100% heavy
About 8 checks or questions per day, information search happening several times a day in O*NET; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: O*NET
Web search with an AI : 5 per day; with AI: 8% light, 25% typical, 87% heavy
About 5 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 3 per week; with AI: 12% light, 37% typical, 100% heavy
About 3 reports to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Query a PDF : 3 per week; with AI: 10% light, 31% typical, 100% heavy
About 3 long documents to query per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 3 per week; with AI: 25% light, 74% typical, 100% heavy
About 3 foreign-language sources to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Write an email : 15 per day; with AI: 4% light, 12% typical, 43% heavy
About 15 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Literature search : 1 per week; with AI: 10% light, 31% typical, 100% heavy
About 1 documentary investigation per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 2 per week; with AI: 8% light, 25% typical, 87% heavy
About 2 angle searches per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Consultant, 1 year14 assumptions
Presentation of 10 slides : 2 per week; with AI: 10% light, 30% typical, 100% heavy
About 2 presentations per week, the typical deliverable of the occupation, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 2 per day; with AI: 8% light, 24% typical, 83% heavy
About 2 pages per day (note, proposal, deliverable), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Proofread a page : 4 per day; with AI: 10% light, 30% typical, 100% heavy
About 4 texts to rework per day; share done with AI based on the share of work messages that edit a supplied text (OpenAI, 2025), adjusted to the occupation. Source: OpenAI
Write an email : 20 per day; with AI: 4% light, 12% typical, 42% heavy
About 20 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 5 per day; with AI: 12% light, 36% typical, 100% heavy
About 5 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Analyse a spreadsheet : 1 per day; with AI: 6% light, 18% typical, 62% heavy
About 1 spreadsheet to analyse per day, analysing collected data being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Summarise 30 pages : 3 per week; with AI: 12% light, 36% typical, 100% heavy
About 3 long documents to summarise per week, information gathering being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Query a PDF : 3 per week; with AI: 10% light, 30% typical, 100% heavy
About 3 long documents to query per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 4 per week; with AI: 12% light, 36% typical, 100% heavy
About 4 sets of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Literature search : 1 per week; with AI: 10% light, 30% typical, 100% heavy
About 1 literature search per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Deep research : 22 per year; with AI: 4% light, 12% typical, 42% heavy
About 2 in-depth questions per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 3 per week; with AI: 8% light, 24% typical, 83% heavy
About 3 reflection sessions per week (structuring a problem), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise a meeting : 1.2 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 6 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Browser agent : 2 per week; with AI: 5% light, 20% typical, 60% heavy (if tooling: agents)
About 2 online collections per week, assumption; share delegated to an agent set without a published survey.
Lawyer, 1 year12 assumptions
Review a contract : 4 per week; with AI: 7% light, 21% typical, 74% heavy
About 4 contracts of 20 pages reviewed per week, text analysis being a daily task for most lawyers in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 6 per day; with AI: 8% light, 25% typical, 89% heavy
About 6 legal questions per day, interpreting texts being a daily task in O*NET; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: O*NET
Write a page : 1 per day; with AI: 6% light, 17% typical, 59% heavy
About 1 page per day (note, opinion, clause), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Proofread a page : 3 per day; with AI: 7% light, 21% typical, 74% heavy
About 3 texts to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write an email : 20 per day; with AI: 3% light, 8% typical, 30% heavy
About 20 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 3 per week; with AI: 8% light, 25% typical, 89% heavy
About 3 decisions or reports to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Query a PDF : 3 per week; with AI: 7% light, 21% typical, 74% heavy
About 3 long case files to query per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Literature search : 1 per week; with AI: 7% light, 21% typical, 74% heavy
About 1 literature search per week (case law, legal scholarship), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 17% light, 51% typical, 100% heavy
About 2 pages to translate per week (contracts in English), assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Write meeting minutes : 2 per week; with AI: 8% light, 25% typical, 89% heavy
About 2 sets of minutes per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 1 per week; with AI: 6% light, 17% typical, 59% heavy
About 1 case strategy to prepare per week, preparing arguments being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Summarise a meeting : 0.6 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 3 meetings per week, assumption; with AI in the office suite, most are summarised without being asked, assumption; one summarised meeting serves its 5 participants, each carrying a fifth of it.
Researcher, 1 year13 assumptions
Query a PDF : 5 per week; with AI: 8% light, 25% typical, 88% heavy
About 5 scientific articles queried per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 3 per week; with AI: 10% light, 30% typical, 100% heavy
About 3 articles or reports to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Literature search : 1 per week; with AI: 8% light, 25% typical, 88% heavy
About 1 bibliographic search per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Deep research : 22 per year; with AI: 3% light, 10% typical, 35% heavy
About 2 quick literature reviews per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 1 per day; with AI: 7% light, 20% typical, 70% heavy
About 1 page per day (article, project, report), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Proofread a page : 2 per day; with AI: 8% light, 25% typical, 88% heavy
About 2 texts to rework per day (scientific English), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 20% light, 60% typical, 100% heavy
About 2 pages to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Analyse a spreadsheet : 3 per week; with AI: 5% light, 15% typical, 53% heavy
About 3 datasets to analyse per week, data analysis being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Generate a function : 1 per day; with AI: 10% light, 30% typical, 100% heavy
About 1 analysis script per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 5 per day; with AI: 10% light, 30% typical, 100% heavy
About 5 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Write an email : 12 per day; with AI: 3% light, 10% typical, 35% heavy
About 12 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 11 per year; with AI: 8% light, 25% typical, 88% heavy
About 1 presentation per month (seminar, conference), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 2 per week; with AI: 7% light, 20% typical, 70% heavy
About 2 methodology discussions per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Developer, 1 year11 assumptions
Code suggestion : 200 per day; with AI: 30% light, 80% typical, 100% heavy
About 200 suggestions offered per day when the assistant is active, assumption; share done with AI based on the half of developers who use AI every day (Stack Overflow, 2025). Source: Stack Overflow
Generate a function : 6 per day; with AI: 10% light, 35% typical, 80% heavy (if tooling: conversational assistant or AI in the office suite)
About 6 functions written per day, assumption; share done with AI based on the 59% of respondents who write code partly with AI (Stack Overflow, 2025); replaced by agent hours when there are agents. Source: Stack Overflow
Have code reviewed : 2 per day; with AI: 5% light, 25% typical, 70% heavy
About 2 code reviews per day, assumption; share done with AI based on the 23% of respondents who have their code partly reviewed by AI (Stack Overflow, 2025). Source: Stack Overflow
Quick question : 10 per day; with AI: 15% light, 45% typical, 100% heavy
About 10 technical questions per day, assumption; share done with AI set high, looking for an answer being the main stated use (Stack Overflow, 2025). Source: Stack Overflow
Conversation of 10 exchanges : 2 per day; with AI: 5% light, 20% typical, 60% heavy
About 2 debugging sessions per day, fixing code being weekly or more often in O*NET; share done with AI based on the 47% of respondents who debug partly with AI (Stack Overflow, 2025). Source: Stack Overflow
Query a PDF : 3 per week; with AI: 9% light, 26% typical, 92% heavy
About 3 long documentation sets queried per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 2 per week; with AI: 7% light, 21% typical, 74% heavy
About 2 pages of documentation per week, documentation being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Write an email : 8 per day; with AI: 4% light, 11% typical, 37% heavy
About 8 long emails or messages per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 2 per week; with AI: 11% light, 32% typical, 100% heavy
About 2 minutes or incident reports per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 3 per day; with AI: 7% light, 21% typical, 74% heavy
About 3 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Coding agent hour : 6 per day; with AI: 15% light, 45% typical, 90% heavy (if tooling: agents)
About 6 hours of development per day in which an agent can work, assumption; share done with AI based on the 31% who use AI more than 4 hours a day (Stack Overflow, 2026). Source: Stack Overflow
Data scientist, 1 year12 assumptions
Generate a function : 5 per day; with AI: 10% light, 35% typical, 80% heavy (if tooling: conversational assistant or AI in the office suite)
About 5 functions or queries written per day, assumption for lack of published frequencies for this occupation in O*NET; fallback as for developers (Stack Overflow, 2025); replaced by agent hours when there are agents. Source: Stack Overflow
Code suggestion : 150 per day; with AI: 30% light, 80% typical, 100% heavy
About 150 suggestions offered per day when the assistant is active, assumption; fallback as for developers (Stack Overflow, 2025). Source: Stack Overflow
Analyse a spreadsheet : 1 per day; with AI: 5% light, 25% typical, 60% heavy
About 1 dataset to explore per day, assumption; share done with AI based on the 16-fold gap between the most active users and the median for data analysis (OpenAI, 2025). Source: OpenAI
Quick question : 8 per day; with AI: 12% light, 36% typical, 100% heavy
About 8 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Conversation of 10 exchanges : 1 per day; with AI: 8% light, 24% typical, 84% heavy
About 1 methodology discussion per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Query a PDF : 2 per week; with AI: 10% light, 30% typical, 100% heavy
About 2 articles or documentation sets queried per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 1 per week; with AI: 12% light, 36% typical, 100% heavy
About 1 long document to summarise per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 2 per week; with AI: 8% light, 24% typical, 84% heavy
About 2 pages of reports per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 22 per year; with AI: 10% light, 30% typical, 100% heavy
About 2 presentations of results per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write an email : 10 per day; with AI: 4% light, 12% typical, 42% heavy
About 10 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Specialised task : 1,000 per week; with AI: 5% light, 30% typical, 100% heavy
About 1,000 items to classify per week (texts, reviews, tickets), assumption; share delegated to a specialised model set without a published survey.
Coding agent hour : 4 per day; with AI: 10% light, 40% typical, 90% heavy (if tooling: agents)
About 4 hours per day in which an agent can work, assumption; share done with AI based on the usage time reported by developers (Stack Overflow, 2026). Source: Stack Overflow
Graphic designer, 1 year11 assumptions
Visual in 4 attempts : 10 per week; with AI: 5% light, 20% typical, 60% heavy
About 10 visual leads searched for per week, creating mock-ups happening several times a day in O*NET; share done with AI based on the 31% of designers who use AI in their creative work (Figma, 2025). Source: Figma
Retouch a photo : 15 per week; with AI: 10% light, 35% typical, 80% heavy
About 15 photos to retouch per week (extension, cleaning, replacing an element), images being worked on every day in O*NET; fallback set higher, AI retouching being built into image software, assumption. Source: O*NET
Create an image : 5 per week; with AI: 5% light, 25% typical, 70% heavy
About 5 supporting images per week (background, texture, icon), assumption; share done with AI based on Figma (2025). Source: Figma
Create a large image : 3 per week; with AI: 2% light, 10% typical, 40% heavy
About 3 final high-definition visuals per week, assumption; share done with AI set low, the delivered visual still being mostly made by hand (Figma, 2025). Source: Figma
Clip of 5 seconds : 1 per week; with AI: 0% light, 5% typical, 30% heavy
About 1 visual to animate per week, assumption; share done with AI set low, generated video remaining rare at work, assumption.
Quick question : 4 per day; with AI: 10% light, 29% typical, 100% heavy
About 4 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Write an email : 10 per day; with AI: 3% light, 10% typical, 33% heavy
About 10 emails per day, exchanges with the client being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Proofread a page : 1 per day; with AI: 8% light, 24% typical, 84% heavy
About 1 mock-up text to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Conversation of 10 exchanges : 2 per week; with AI: 6% light, 19% typical, 67% heavy
About 2 briefs to dig into per week, research on the target audience being weekly or more often in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Web search with an AI : 2 per day; with AI: 6% light, 19% typical, 67% heavy
About 2 reference searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Presentation of 10 slides : 22 per year; with AI: 8% light, 24% typical, 84% heavy
About 2 presentations of design directions per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Videographer, 1 year9 assumptions
Summarise a meeting : 3 per week; with AI: 20% light, 60% typical, 100% heavy
About 3 hours of rushes or interviews to transcribe per week, assumption; share done with AI set high, automatic transcription being built into editing software, assumption.
Clip of 5 seconds : 10 per week; with AI: 1% light, 5% typical, 30% heavy
About 10 illustration shots per week, editing being done several times an hour in O*NET; share done with AI set low, the Microsoft study finding little use among editors. Source: Microsoft Research
Clip of 10 seconds : 5 per week; with AI: 0% light, 3% typical, 20% heavy
About 5 long shots per week, assumption; share done with AI set very low, a clip of 10 seconds remaining expensive and rarely usable, assumption. Source: Microsoft Research
Visual in 4 attempts : 3 per week; with AI: 5% light, 20% typical, 60% heavy
About 3 thumbnails or storyboards per week, assumption; fallback as for graphic designers (Figma, 2025). Source: Figma
Retouch a photo : 5 per week; with AI: 5% light, 20% typical, 60% heavy
About 5 images to rework per week, assumption; fallback set as for graphic designers' retouching, a little lower.
Write a page : 2 per week; with AI: 5% light, 16% typical, 57% heavy
About 2 pages of script or synopsis per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 16% light, 48% typical, 100% heavy
About 2 pages of subtitles to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Quick question : 3 per day; with AI: 8% light, 24% typical, 85% heavy
About 3 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Write an email : 8 per day; with AI: 3% light, 8% typical, 28% heavy
About 8 emails per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Customer service officer, 1 year8 assumptions
Handle a customer request : 40 per day; with AI: 3% light, 12% typical, 50% heavy
About 40 requests handled per day, exchanges with customers being hourly or more often in O*NET; fallback set below the grid, the tool often being supplied and restricted by the employer, assumption. Source: O*NET
Write an email : 15 per day; with AI: 4% light, 13% typical, 45% heavy
About 15 written replies per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Quick question : 6 per day; with AI: 13% light, 38% typical, 100% heavy
About 6 questions per day (procedure, product), assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Proofread a page : 5 per day; with AI: 11% light, 32% typical, 100% heavy
About 5 replies to reword per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 2 per week; with AI: 26% light, 77% typical, 100% heavy
About 2 exchanges to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Web search with an AI : 2 per day; with AI: 9% light, 26% typical, 89% heavy
About 2 web searches per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write meeting minutes : 30 per day; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 30 call summaries per day, a record of each exchange being kept several times an hour in O*NET; with AI in the tools, most are written without being asked, assumption. Source: O*NET
Automated workflow : 30 per day; with AI: 20% light, 50% typical, 90% heavy (if tooling: agents)
About 30 simple requests per day, assumption; with agents, a share is handled by an automated workflow, assumption with no published survey.
Salesperson, 1 year12 assumptions
Write an email : 25 per day; with AI: 4% light, 11% typical, 39% heavy
About 25 emails per day (prospecting, follow-ups), replies to customers being a daily task in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: O*NET
Quick question : 5 per day; with AI: 11% light, 33% typical, 100% heavy
About 5 questions per day, assumption; share done with AI based on the Crédoc Baromètre du numérique (2026 edition, 2025 data), adjusted to the occupation. Source: Crédoc
Presentation of 10 slides : 2 per week; with AI: 9% light, 28% typical, 97% heavy
About 2 sales proposals as slides per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Write a page : 3 per week; with AI: 7% light, 22% typical, 77% heavy
About 3 tender pages per week, price and lead-time quotation being daily in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: O*NET
Write meeting minutes : 6 per week; with AI: 11% light, 33% typical, 100% heavy
About 6 meeting reports per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 3 per day; with AI: 7% light, 22% typical, 77% heavy
About 3 searches per day to prepare a meeting, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Deep research : 22 per year; with AI: 4% light, 11% typical, 39% heavy
About 2 key accounts to study in depth per month, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Proofread a page : 2 per day; with AI: 9% light, 28% typical, 97% heavy
About 2 texts to rework per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation using the Microsoft study of 200,000 conversations. Source: Microsoft Research
Voice mode, 1 minute : 10 per week; with AI: 6% light, 17% typical, 58% heavy
About 10 minutes of dictation per week (in the car, between two meetings), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 1 per week; with AI: 22% light, 66% typical, 100% heavy
About 1 page to translate per week, assumption; share done with AI set high, translation being among the tasks most often delegated to AI (Microsoft, 2025). Source: Microsoft Research
Summarise a meeting : 0.8 per week; with AI: 30% light, 70% typical, 100% heavy (if tooling: AI in the office suite)
About 4 video meetings per week, assumption; with AI in the office suite, most are summarised without anyone asking, assumption; one summarised meeting serves its 5 participants, each carrying one fifth of it.
Browser agent : 5 per week; with AI: 5% light, 20% typical, 60% heavy (if tooling: agents)
About 5 client file updates or prospect searches per week, assumption; share handed to an agent set without a published survey.
Field occupation (technician, care worker), 1 year8 assumptions
Quick question : 4 per day; with AI: 7% light, 22% typical, 76% heavy
About 4 questions per day (procedure, breakdown, instruction), assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025), adjusted to the occupation, low for occupations where the work is mainly physical (Microsoft, 2025). Source: Microsoft Research
Write meeting minutes : 3 per day; with AI: 7% light, 22% typical, 76% heavy
About 3 intervention reports or handovers per day, record keeping being several times a day for nurses in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: O*NET
Write an email : 3 per day; with AI: 2% light, 7% typical, 25% heavy
About 3 emails or messages per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Voice mode, 1 minute : 5 per week; with AI: 4% light, 11% typical, 38% heavy
About 5 minutes of dictation per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Query a PDF : 2 per week; with AI: 6% light, 18% typical, 64% heavy
About 2 long manuals or protocols to query per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 1 per week; with AI: 15% light, 44% typical, 100% heavy
About 1 exchange to translate per week (patient, client, manual), assumption; high share done with AI, translation being among the tasks most often handed to AI (Microsoft, 2025). Source: Microsoft Research
Proofread a page : 2 per week; with AI: 6% light, 18% typical, 64% heavy
About 2 texts to rework per week, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Web search with an AI : 1 per day; with AI: 5% light, 15% typical, 51% heavy
About 1 web search per day, assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Teacher, 1 year11 assumptions
Prepare a lesson : 8 per week; with AI: 3% light, 15% typical, 60% heavy
About 8 lessons prepared or redone per week, for 18 hours of teaching and a quarter of working time spent on preparation (DEPP, 2025); share done with AI based on the 37% of teachers who prepare their lessons with AI at least once a month (Gallup, 2025). Source: Gallup
Write a page : 4 per week; with AI: 3% light, 15% typical, 60% heavy
About 4 exercise sheets or assessments per week, assumption; share done with AI based on the 33% who make their sheets with AI at least once a month (Gallup, 2025). Source: Gallup
Proofread a page : 60 per week; with AI: 0.3% light, 2% typical, 15% heavy
About 60 marked papers per week, marking being weekly or more often in O*NET; very low share done with AI, only 16% marking with AI at least once a month (Gallup, 2025). Source: Gallup
Quick question : 1,000 per year; with AI: 5% light, 17% typical, 60% heavy
About 660 questions and 360 report comments per year, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025), adjusted to the occupation. Source: Crédoc
Write an email : 5 per day; with AI: 3% light, 8% typical, 29% heavy
About 5 emails per day (families, colleagues), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Create an image : 2 per week; with AI: 1% light, 5% typical, 30% heavy
About 2 lesson illustrations per week, assumption; low share done with AI, assumption.
Presentation of 10 slides : 1 per week; with AI: 2% light, 10% typical, 50% heavy
About 1 slide deck per week, assumption; share done with AI based on Gallup (2025). Source: Gallup
Conversation of 10 exchanges : 1 per week; with AI: 6% light, 17% typical, 59% heavy
About 1 sequence to adapt per week, adapting to pupils' needs being daily in O*NET; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: O*NET
Write meeting minutes : 1 per week; with AI: 8% light, 25% typical, 88% heavy
About 1 report per week (class council, meeting), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Summarise 30 pages : 10 per year; with AI: 8% light, 25% typical, 88% heavy
About 10 long texts to read per year (curriculum, report), assumption; share done with AI taken from the value common to all occupations for this task, then adjusted to the occupation from the Microsoft study of 200,000 conversations. Source: Microsoft Research
Translate a page : 0.5 per week; with AI: 17% light, 50% typical, 100% heavy
About 1 page to translate every 2 weeks, assumption; high share done with AI, translation being among the tasks most often handed to AI (Microsoft, 2025). Source: Microsoft Research
Secondary school pupil, 1 year9 assumptions
Quick question : 10 per week; with AI: 5% light, 25% typical, 70% heavy
About 10 exercise questions per week, assumption; share done with AI based on the 68% of teenage users who use it for homework (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Conversation of 10 exchanges : 3 per week; with AI: 5% light, 20% typical, 60% heavy
About 3 concepts to understand or revise per week, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Write a page : 1 per week; with AI: 5% light, 25% typical, 70% heavy
About 1 essay per week, assumption; share done with AI based on the quarter of American teenagers who use ChatGPT for homework (Pew, 2025). Source: Pew Research Center
Summarise 30 pages : 0.5 per week; with AI: 5% light, 20% typical, 60% heavy
About 1 chapter or book to summarise every 2 weeks, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Translate a page : 1 per week; with AI: 5% light, 20% typical, 60% heavy
About 1 language exercise to translate per week, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Presentation of 10 slides : 6 per year; with AI: 10% light, 30% typical, 70% heavy
About 6 presentations per year, assumption; use set without a reviewed French survey.
Create an image : 1 per week; with AI: 5% light, 20% typical, 60% heavy
About 1 image per week for a presentation or homework, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Web search with an AI : 5 per week; with AI: 10% light, 30% typical, 70% heavy
About 5 searches per week, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
AI overview of a search : 10 per week; with AI: 30% light, 30% typical, 30% heavy
About 10 searches per week on a search engine that displays an AI-generated overview, assumption; the overview is imposed by the engine and does not depend on the pupil's use.
Student, 1 year13 assumptions
Quick question : 15 per week; with AI: 10% light, 35% typical, 80% heavy
About 15 questions per week, assumption; share done with AI based on the 74% of French students who use generative AI for their studies (Eurostat, 2025). Source: Eurostat
Conversation of 10 exchanges : 4 per week; with AI: 10% light, 30% typical, 70% heavy
About 4 concepts to understand or revise per week, assumption; share done with AI based on explaining concepts, the first use reported by British students (HEPI, 2025). Source: HEPI
Summarise 30 pages : 2 per week; with AI: 10% light, 35% typical, 80% heavy
About 2 articles or lectures to summarise per week, assumption; share done with AI based on summarising articles, among the first uses reported (HEPI, 2025). Source: HEPI
Query a PDF : 2 per week; with AI: 5% light, 25% typical, 60% heavy
About 2 long documents to query per week, assumption; share done with AI based on HEPI (2025). Source: HEPI
Write a page : 3 per week; with AI: 5% light, 25% typical, 60% heavy
About 3 pages of coursework or dissertation per week, assumption; share done with AI based on the 88% of British students who use AI for their assessed work (HEPI, 2025). Source: HEPI
Proofread a page : 3 per week; with AI: 10% light, 40% typical, 90% heavy
About 3 pages to rework per week, assumption; high share done with AI, rephrasing being the most common writing use (OpenAI, 2025, and HEPI, 2025). Source: HEPI
Translate a page : 1 per week; with AI: 20% light, 50% typical, 90% heavy
About 1 page to translate per week, assumption; high share done with AI, assumption based on Eurostat (2025). Source: Eurostat
Literature search : 9 per year; with AI: 10% light, 30% typical, 70% heavy
About 9 literature searches per year (one per month of study), assumption; share done with AI based on HEPI (2025). Source: HEPI
Deep research : 9 per year; with AI: 0% light, 10% typical, 50% heavy
About 9 topics per year that lend themselves to in-depth research, assumption; low share done with AI, the free plan limiting the number.
Presentation of 10 slides : 8 per year; with AI: 10% light, 30% typical, 70% heavy
About 8 defences or presentations per year, assumption; use set without a reviewed French survey.
Generate a function : 1 per week; with AI: 5% light, 25% typical, 70% heavy
About 1 programming exercise per week on average across all fields of study, assumption; share done with AI based on Eurostat (2025). Source: Eurostat
Create an image : 1 per week; with AI: 2% light, 10% typical, 40% heavy
About 1 image per week for a dissertation or presentation, assumption.
Web search with an AI : 10 per week; with AI: 10% light, 30% typical, 70% heavy
About 10 searches per week, assumption; share done with AI based on Eurostat (2025). Source: Eurostat
Free personal use, 1 year13 assumptions
Quick question : 10 per day; with AI: 3% light, 10% typical, 40% heavy
About 10 information needs per day; share done with AI based on the 21% of users who look for information with AI every day and the 28% who do so 1 to 2 times a week (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Conversation of 10 exchanges : 3 per week; with AI: 5% light, 30% typical, 80% heavy
About 3 pieces of practical advice per week (health, cooking, travel, sport), assumption; share done with AI based on practical advice, about 29% of messages (OpenAI, 2025). Source: OpenAI
Write an email : 10 per week; with AI: 1% light, 5% typical, 20% heavy
About 10 personal messages or emails per week, assumption; low share done with AI, based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Proofread a page : 1 per week; with AI: 5% light, 30% typical, 70% heavy
About 1 important text to rework per week (letter, CV, message), assumption; share done with AI based on the 38% of users who write or improve a text with AI at least once a week (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Translate a page : 1 per week; with AI: 10% light, 30% typical, 70% heavy
About 1 page to translate per week, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Create an image : 2 per week; with AI: 2% light, 10% typical, 50% heavy
About 2 image requests per week (map, illustration, portrait), assumption; share done with AI based on the 26% of users who create images or videos at least once a week (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Retouch a photo : 2 per week; with AI: 1% light, 5% typical, 30% heavy
About 2 photos to retouch per week, assumption; share done with AI based on the Crédoc Baromètre du numérique (edition 2026, data 2025). Source: Crédoc
Clip of 5 seconds : 1 per week; with AI: 0% light, 4% typical, 40% heavy
About 1 video request per week, assumption; low share done with AI, about 2 clips per year at the typical level, the free plan limiting the number, assumption. Source: Crédoc
Voice mode, 1 minute : 5 per week; with AI: 2% light, 10% typical, 50% heavy
About 5 minutes per week during which one could talk to an assistant, assumption; share done with AI based on the 41% of users who chat with AI (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Query a PDF : 12 per year; with AI: 5% light, 20% typical, 60% heavy
About 1 long document per month (lease, manual, contract), assumption.
Deep research : 12 per year; with AI: 0% light, 10% typical, 50% heavy
About 1 topic per month that lends itself to in-depth research, assumption; low share done with AI, the free plan limiting the number.
Generate a function : 0.5 per week; with AI: 0% light, 10% typical, 50% heavy
About 1 small program every 2 weeks, assumption; share done with AI based on the 20% of users who program with AI at least once a week (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
AI overview of a search : 3 per day; with AI: 30% light, 30% typical, 30% heavy
About 3 searches per day on a search engine that displays an AI-generated overview, assumption; the overview is imposed by the engine and does not depend on the person's use.
ChatGPT Plus, 1 year12 assumptions
Question to ChatGPT : 15 per day; with AI: 10% light, 35% typical, 90% heavy
About 15 occasions to ask a question per day; use higher than on the free plan, paying users being the most intensive users (Crédoc, Baromètre du numérique, edition 2026, data 2025); usage cap of the plan not verified. Source: Crédoc
Conversation of 10 exchanges : 1 per day; with AI: 10% light, 40% typical, 100% heavy
About 1 piece of practical advice per day, assumption; share done with AI based on OpenAI (2025). Source: OpenAI
Proofread a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 text to rework per day, assumption; half of daily users of AI for writing pay (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Write a page : 3 per week; with AI: 10% light, 30% typical, 80% heavy
About 3 pages to write per week, assumption.
Query a PDF : 2 per week; with AI: 10% light, 40% typical, 90% heavy
About 2 long documents to query per week, assumption.
Translate a page : 1 per week; with AI: 20% light, 50% typical, 90% heavy
About 1 page to translate per week, assumption.
Create an image : 3 per week; with AI: 5% light, 30% typical, 90% heavy
About 3 image requests per week, assumption; 55% of daily users of AI for creating visuals pay (Crédoc, Baromètre du numérique, edition 2026, data 2025). Source: Crédoc
Deep research : 1 per week; with AI: 2% light, 20% typical, 80% heavy
About 1 topic per week that lends itself to in-depth research, assumption; usage cap of the plan not verified.
Clip of 5 seconds : 1 per week; with AI: 0% light, 5% typical, 40% heavy
About 1 video request per week, assumption; usage cap of the plan not verified.
Voice mode, 1 minute : 10 per week; with AI: 2% light, 15% typical, 60% heavy
About 10 minutes per week during which one could talk to the assistant, assumption.
Browser agent : 1 per week; with AI: 0% light, 10% typical, 50% heavy
About 1 online procedure per week that can be entrusted to the agent, assumption.
Coding agent hour : 5 per week; with AI: 10% light, 30% typical, 80% heavy (if tooling: agents)
About 5 hours per week during which a coding agent can work, for those who code, assumption; usage cap of the plan not verified.
ChatGPT Pro, 1 year10 assumptions
Question to ChatGPT : 30 per day; with AI: 20% light, 50% typical, 100% heavy
About 30 occasions to ask a question per day for professional use, assumption; usage cap of the plan not verified.
Conversation of 10 exchanges : 3 per day; with AI: 20% light, 50% typical, 100% heavy
About 3 long reasoning sessions per day, assumption.
Query a PDF : 1 per day; with AI: 20% light, 50% typical, 100% heavy
About 1 long document per day, assumption.
Proofread a page : 2 per day; with AI: 20% light, 50% typical, 100% heavy
About 2 texts to rework per day, assumption.
Write a page : 1 per day; with AI: 20% light, 40% typical, 90% heavy
About 1 page to write per day, assumption.
Create an image : 1 per day; with AI: 10% light, 40% typical, 100% heavy
About 1 image request per day, assumption.
Deep research : 1 per day; with AI: 10% light, 30% typical, 80% heavy
About 1 topic per day that lends itself to in-depth research, assumption; usage cap of the plan not verified.
Clip of 10 seconds : 3 per week; with AI: 2% light, 15% typical, 60% heavy
About 3 video requests per week, assumption; usage cap of the plan not verified.
Browser agent : 3 per week; with AI: 10% light, 30% typical, 80% heavy
About 3 online procedures per week that can be entrusted to the agent, assumption.
Coding agent hour : 10 per week; with AI: 20% light, 50% typical, 100% heavy
About 10 hours per week during which a coding agent can work, the plan being aimed at professional use, assumption; usage cap of the plan not verified.
Claude Pro, 1 year10 assumptions
Quick question : 10 per day; with AI: 10% light, 35% typical, 90% heavy
About 10 occasions to ask a question per day, assumption; the provider publishes no quantified cap. Source: Anthropic
Conversation of 10 exchanges : 1 per day; with AI: 10% light, 40% typical, 100% heavy
About 1 long reasoning session per day, assumption.
Write a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 page to write per day, assumption.
Proofread a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 text to rework per day, assumption.
Query a PDF : 3 per week; with AI: 10% light, 40% typical, 90% heavy
About 3 long documents to query per week, assumption.
Summarise 30 pages : 2 per week; with AI: 10% light, 40% typical, 90% heavy
About 2 documents to summarise per week, assumption.
Generate a function : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 function to write per day, assumption.
Analyse a spreadsheet : 1 per week; with AI: 10% light, 30% typical, 80% heavy
About 1 spreadsheet per week, assumption.
Deep research : 1 per week; with AI: 2% light, 20% typical, 80% heavy
About 1 topic per week that lends itself to in-depth research, assumption; the provider publishes no quantified cap. Source: Anthropic
Coding agent hour : 5 per week; with AI: 10% light, 40% typical, 100% heavy (if tooling: agents)
About 5 hours per week during which a coding agent can work, assumption; the provider publishes no quantified cap. Source: Anthropic
Claude Max, 1 year8 assumptions
Coding agent hour : 15 per week; with AI: 20% light, 50% typical, 100% heavy
About 15 hours per week during which a coding agent can work; the plan gives 5 times the usage of Claude Pro, with no quantified cap (Anthropic, 2026). Source: Anthropic
Quick question : 20 per day; with AI: 10% light, 35% typical, 90% heavy
About 20 occasions to ask a question per day, assumption. Source: Anthropic
Conversation of 10 exchanges : 2 per day; with AI: 10% light, 40% typical, 100% heavy
About 2 long reasoning sessions per day, assumption.
Write a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 page to write per day, assumption.
Query a PDF : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 long document per day, assumption.
Generate a function : 3 per day; with AI: 10% light, 40% typical, 90% heavy
About 3 functions per day outside the agent, assumption.
Have code reviewed : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 code review per day, assumption.
Deep research : 3 per week; with AI: 10% light, 30% typical, 80% heavy
About 3 topics per week that lend themselves to in-depth research, assumption.
Google AI Pro, 1 year9 assumptions
Question to Gemini : 15 per day; with AI: 10% light, 35% typical, 90% heavy
About 15 occasions to ask a question per day, assumption; the plan gives 4 times the limits of the free plan, with no quantified cap (Google, 2026). Source: Google
Conversation of 10 exchanges : 1 per day; with AI: 10% light, 40% typical, 100% heavy
About 1 piece of practical advice per day, assumption.
Deep research : 2 per week; with AI: 5% light, 25% typical, 80% heavy
About 2 topics per week that lend themselves to in-depth research, assumption; the plan gives 4 times the limits of the free plan, with no quantified cap (Google, 2026). Source: Google
Create an image : 1 per day; with AI: 5% light, 25% typical, 80% heavy
About 1 image request per day, assumption; the plan gives 4 times the limits of the free plan, with no quantified cap (Google, 2026). Source: Google
Retouch a photo : 3 per week; with AI: 5% light, 25% typical, 70% heavy
About 3 photos to retouch per week, assumption.
Clip of 5 seconds : 1 per week; with AI: 0% light, 5% typical, 40% heavy
About 1 video request per week, assumption; the plan gives only a limited trial of the video generator (Google, 2026). Source: Google
Query a PDF : 2 per week; with AI: 10% light, 40% typical, 90% heavy
About 2 long documents per week, the plan reading 1 million tokens (Google, 2026). Source: Google
Proofread a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 text to rework per day, assumption.
Voice mode, 1 minute : 10 per week; with AI: 2% light, 15% typical, 60% heavy
About 10 minutes per week during which one could talk to the assistant, assumption.
Mistral Vibe Pro, 1 year8 assumptions
Quick question : 15 per day; with AI: 10% light, 35% typical, 90% heavy
About 15 occasions to ask a question per day, i.e. at the typical level about 3% of the 150 quick answers per day of the plan (Mistral AI, 2026); calculated as a quick question, the Mistral AI life cycle assessment giving only carbon. Source: Mistral AI
Conversation of 10 exchanges : 1 per day; with AI: 10% light, 40% typical, 100% heavy
About 1 piece of practical advice per day, assumption.
Create an image : 3 per week; with AI: 5% light, 30% typical, 90% heavy
About 3 image requests per week, assumption; the plan gives up to 40 times the images of the free plan (Mistral AI, 2026). Source: Mistral AI
Deep research : 1 per week; with AI: 2% light, 20% typical, 80% heavy
About 1 topic per week that lends itself to in-depth research, assumption; cap not quantified by the provider. Source: Mistral AI
Proofread a page : 1 per day; with AI: 10% light, 40% typical, 90% heavy
About 1 text to rework per day, assumption.
Translate a page : 2 per week; with AI: 20% light, 50% typical, 90% heavy
About 2 pages to translate per week, assumption.
Generate a function : 3 per week; with AI: 5% light, 30% typical, 90% heavy
About 3 functions to write per week, assumption.
Query a PDF : 2 per week; with AI: 10% light, 40% typical, 90% heavy
About 2 long documents per week, assumption.
Perplexity Pro, 1 year6 assumptions
Web search with an AI : 15 per day; with AI: 20% light, 50% typical, 100% heavy
About 15 searches per day, the use the plan is built for, assumption; the provider's help page quantifies no cap. Source: Perplexity
Deep research : 1 per day; with AI: 3% light, 15% typical, 60% heavy
About 1 topic per day that lends itself to in-depth research, assumption; the provider's help page quantifies no cap. Source: Perplexity
Quick question : 3 per day; with AI: 5% light, 20% typical, 60% heavy
About 3 questions per day outside search, assumption.
Literature search : 1 per week; with AI: 10% light, 40% typical, 90% heavy
About 1 literature search per week, assumption.
Query a PDF : 1 per week; with AI: 5% light, 25% typical, 70% heavy
About 1 long document per week, assumption.
Create an image : 1 per week; with AI: 2% light, 10% typical, 50% heavy
About 1 image request per week, assumption.
Microsoft 365 Copilot licence, 1 year9 assumptions
Summarise a meeting : 1.2 per week; with AI: 30% light, 70% typical, 100% heavy
About 6 meetings per week, most summarised by the suite's AI without anyone asking, assumption; one summarised meeting serves its 5 participants, each carrying one fifth of it. Source: Microsoft Research
Write an email : 15 per day; with AI: 5% light, 15% typical, 40% heavy
About 15 emails per day, some written or summarised in the mail client, assumption.
Summarise 30 pages : 3 per week; with AI: 10% light, 35% typical, 90% heavy
About 3 long threads or documents to summarise per week, assumption.
Write a page : 1 per day; with AI: 5% light, 20% typical, 50% heavy
About 1 page written per day in the word processor, assumption.
Analyse a spreadsheet : 3 per week; with AI: 5% light, 20% typical, 60% heavy
About 3 spreadsheets per week, assumption.
Presentation of 10 slides : 22 per year; with AI: 10% light, 30% typical, 70% heavy
About 2 presentations per month, assumption.
Quick question : 8 per day; with AI: 10% light, 30% typical, 90% heavy
About 8 questions per day in the assistant, assumption.
Proofread a page : 2 per day; with AI: 5% light, 25% typical, 70% heavy
About 2 texts to rework per day, assumption.
Deep research : 88 per year; with AI: 2% light, 12% typical, 60% heavy
About 2 in-depth questions per week; at the typical level about 1 deep research per month, against a cap of 25 per month (Microsoft, 2026). Source: Microsoft
GitHub Copilot licence, 1 year6 assumptions
Code suggestion : 200 per day; with AI: 50% light, 85% typical, 100% heavy
About 200 suggestions offered per day; they are unlimited and outside credits in all plans (GitHub, 2026); use is an assumption. Source: GitHub
Generate a function : 6 per day; with AI: 10% light, 35% typical, 80% heavy
About 6 functions per day, assumption; use as for developers (Stack Overflow, 2025). Source: Stack Overflow
Have code reviewed : 2 per day; with AI: 10% light, 30% typical, 80% heavy
About 2 code reviews per day, assumption; share done with AI based on Stack Overflow (2025). Source: Stack Overflow
Quick question : 8 per day; with AI: 15% light, 40% typical, 100% heavy
About 8 technical questions per day, assumption; share done with AI based on Stack Overflow (2025). Source: Stack Overflow
Conversation of 10 exchanges : 1 per day; with AI: 5% light, 20% typical, 60% heavy
About 1 debugging session per day, assumption; share done with AI based on Stack Overflow (2025). Source: Stack Overflow
Coding agent hour : 4 per day; with AI: 2% light, 10% typical, 50% heavy
About 4 hours per day during which the licence's agent can work, assumption; the enterprise licence gives 1,900 credits per month, with no published equivalent in hours (GitHub, 2026). Source: GitHub
ChatGPT Business or Enterprise licence, 1 year11 assumptions
Question to ChatGPT : 10 per day; with AI: 10% light, 35% typical, 90% heavy
About 10 occasions to ask a question per working day, assumption; at the intensive level, several times the typical level, like the most active employees, who send 6 times the messages of the median employee (OpenAI, 2025). Source: OpenAI
Write an email : 15 per day; with AI: 3% light, 10% typical, 30% heavy
About 15 emails per day; without integration with the mail client, a small share is written or reworked in the assistant, assumption.
Proofread a page : 2 per day; with AI: 5% light, 25% typical, 70% heavy
About 2 texts to rework per day, assumption.
Write a page : 1 per day; with AI: 5% light, 20% typical, 50% heavy
About 1 page to write per day, assumption.
Summarise 30 pages : 3 per week; with AI: 10% light, 30% typical, 80% heavy
About 3 long documents or threads to summarise per week, assumption.
Query a PDF : 2 per week; with AI: 10% light, 30% typical, 80% heavy
About 2 long documents to query per week, assumption.
Analyse a spreadsheet : 3 per week; with AI: 5% light, 20% typical, 60% heavy
About 3 spreadsheets per week, assumption; 19% of monthly users have never run data analysis, against 3% of daily users (OpenAI, 2025). Source: OpenAI
Web search with an AI : 3 per day; with AI: 5% light, 20% typical, 60% heavy
About 3 web searches per day, assumption; 12% of monthly users have never run a search, against 1% of daily users (OpenAI, 2025). Source: OpenAI
Deep research : 88 per year; with AI: 2% light, 12% typical, 60% heavy
About 2 in-depth questions per week, as for the Microsoft 365 Copilot licence; at the typical level about 1 deep research per month, plan cap not reviewed.
Create an image : 1 per week; with AI: 2% light, 15% typical, 50% heavy
About 1 image need per week for a presentation or document, assumption.
Coding agent hour : 5 per week; with AI: 5% light, 20% typical, 60% heavy (if tooling: agents)
About 5 hours per week during which a coding agent can work, for those who code, assumption; the coding agent uses more credits than the other functions (OpenAI, 2025). Source: OpenAI
Autonomous agent working alone, 1 months5 assumptions
The hours the agent works: 6 active hours per day over 30 days, waiting the rest of the time; the coding agent hour serves as the agent hour Source: Hausfather
About 20 web searches per day
About 5 browser tasks per day
One wake-up per hour, checking whether there is work
About 10 messages per day to report back
Agents talking to each other, 1 months1 assumption
Several agents talking to each other use about 15 times the tokens of a conversation, a single agent about 4 times (Anthropic, 2025): the same month of work costs 15/4 times the single agent Source: Anthropic
Application built with a coding agent6 assumptions
About 40 agent hours for a small web application, i.e. 2 weeks of supervised work; a researcher gave 1,138 instructions to an agent over 8 weeks Source: Hausfather
About 60 questions on scoping and technical choices
About 10 design sessions
About 20 code reviews
About 10 images for the interface
About 5 pages of documentation
Book written with AI6 assumptions
The first draft of each page: a book of 80,000 words makes about 200 pages of 400 words
Two rewrites per page
About 20 readings of the manuscript in blocks of 30 pages, for consistency
About 40 sessions on the outline, the characters or the chapters
About 5 literature searches
About 3 cover options of 4 attempts each
3-minute video produced with AI5 assumptions
18 shots of 10 seconds for 3 minutes, 3 attempts per shot kept
One starting image per shot
The script, about 2 pages
The synthetic voice: 3 minutes, 2 takes
The translated subtitles, about 1 page
Study carried out end to end with AI11 assumptions
About 5 in-depth searches
About 3 literature searches of 10 sources
About 20 long documents queried
About 10 documents summarised
About 5 spreadsheets analysed
About 10 one-hour interviews or meetings transcribed
Writing the 30-page report
Two rewrites per page
Two presentations
About 100 questions over the course of the study
About 15 thinking sessions
Automated monitoring, 1 year6 assumptions
One run per day
Each day, 50 articles sorted by a specialised model
Each day, 10 assisted searches
Each day, 5 selected articles summarised in one page
Each day, one summary page
One deep research per week
Organisations
196 assumptions · 18 entries
Micro-enterprise of 5 people, 1 year17 assumptions
Manager or executive, 1 year (if sector: market services)
Executives and managers, 1 out of 5; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: market services)
Office employees and technicians, 1 out of 5; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Skilled field occupations (care workers, technicians, social work), 1 out of 5; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Service and sales occupations, 1 out of 5; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Manual workers, farmers and manual occupations, 1 out of 5; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: industry)
Executives and managers, 1 out of 5; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: industry)
Office employees and technicians, 1 out of 5; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Manual workers, farmers and manual occupations, 3 out of 5; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: public administration)
Executives and managers, 1 out of 5; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: public administration)
Office employees and technicians, 1 out of 5; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Skilled field occupations (care workers, technicians, social work), 1 out of 5; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Service and sales occupations, 1 out of 5; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Manual workers, farmers and manual occupations, 1 out of 5; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: education)
Executives and managers, 1 out of 5; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: education)
Office employees and technicians, 1 out of 5; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Teacher, 1 year (if sector: education)
Teachers, 2 out of 5; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Service and sales occupations, 1 out of 5; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
SME of 50 people, 1 year51 assumptions
Manager or executive, 1 year (if sector: market services)
Executives and managers, 5 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: market services)
Engineers, consultants and research officers, 3 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: market services)
Lawyers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: market services)
Communication and marketing officers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: market services)
Human resources officers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: market services)
Accountants and finance professionals, 3 out of 50; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: market services)
Developers, 2 out of 50; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: market services)
Graphic designers, architects and designers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Videographer, 1 year (if sector: market services)
Videographers and image professionals, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: market services)
Office employees and technicians, 5 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: market services)
Assistants and secretaries, 1 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Customer service officer, 1 year (if sector: market services)
Customer account officers, 1 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: market services)
Sales representatives and buyers, 2 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Skilled field occupations (care workers, technicians, social work), 4 out of 50; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Service and sales occupations, 8 out of 50; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Manual workers, farmers and manual occupations, 11 out of 50; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: industry)
Executives and managers, 3 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: industry)
Engineers, consultants and research officers, 5 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: industry)
Researchers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: industry)
Accountants and finance professionals, 2 out of 50; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: industry)
Developers, 1 out of 50; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: industry)
Graphic designers, architects and designers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: industry)
Office employees and technicians, 3 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: industry)
Assistants and secretaries, 1 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: industry)
Sales representatives and buyers, 1 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Skilled field occupations (care workers, technicians, social work), 4 out of 50; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Service and sales occupations, 2 out of 50; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Manual workers, farmers and manual occupations, 26 out of 50; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: public administration)
Executives and managers, 4 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: public administration)
Engineers, consultants and research officers, 3 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: public administration)
Lawyers, 2 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: public administration)
Communication and marketing officers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: public administration)
Human resources officers, 1 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: public administration)
Accountants and finance professionals, 1 out of 50; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: public administration)
Developers, 1 out of 50; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: public administration)
Office employees and technicians, 14 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: public administration)
Assistants and secretaries, 2 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Skilled field occupations (care workers, technicians, social work), 4 out of 50; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Service and sales occupations, 7 out of 50; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Manual workers, farmers and manual occupations, 10 out of 50; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: education)
Executives and managers, 2 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: education)
Researchers, 3 out of 50; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: education)
Office employees and technicians, 7 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: education)
Assistants and secretaries, 1 out of 50; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Teacher, 1 year (if sector: education)
Teachers, 26 out of 50; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Skilled field occupations (care workers, technicians, social work), 4 out of 50; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Service and sales occupations, 4 out of 50; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Manual workers, farmers and manual occupations, 3 out of 50; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Invoice and email automation: 31% of organisations of this size use AI, 38% of the staff of companies that use AI work in a company that uses it for administration Source: Insee
Meeting transcription (if tooling: conversational assistant or agents)
Meeting transcription for 40% of staff: 31% of organisations use AI, 34% of these use speech recognition; with AI in the suite, meetings are already covered in the occupations. Source: Insee
Automated customer service (if sector: market services)
Automated customer service, 100 conversations per year per employee: 31% of organisations use AI, 31% of the staff of companies that use AI work in a company that uses it for marketing or sales Source: Insee
Mid-cap company of 500 people, 1 year55 assumptions
Manager or executive, 1 year (if sector: market services)
Executives and managers, 52 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: market services)
Engineers, consultants and research officers, 31 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: market services)
Lawyers, 6 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: market services)
Researchers, 5 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: market services)
Communication and marketing officers, 7 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Journalist, 1 year (if sector: market services)
Journalists, authors and translators, 4 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: market services)
Human resources officers, 7 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: market services)
Accountants and finance professionals, 30 out of 500; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: market services)
Developers, 22 out of 500; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Data scientist, 1 year (if sector: market services)
Data scientists, 5 out of 500; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: market services)
Graphic designers, architects and designers, 11 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Videographer, 1 year (if sector: market services)
Videographers and image professionals, 6 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: market services)
Office employees and technicians, 47 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: market services)
Assistants and secretaries, 13 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Customer service officer, 1 year (if sector: market services)
Customer account officers, 7 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: market services)
Sales representatives and buyers, 21 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Skilled field occupations (care workers, technicians, social work), 39 out of 500; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Service and sales occupations, 76 out of 500; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Manual workers, farmers and manual occupations, 111 out of 500; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: industry)
Executives and managers, 27 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: industry)
Engineers, consultants and research officers, 53 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: industry)
Researchers, 15 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: industry)
Accountants and finance professionals, 22 out of 500; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: industry)
Developers, 10 out of 500; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: industry)
Graphic designers, architects and designers, 5 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: industry)
Office employees and technicians, 29 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: industry)
Assistants and secretaries, 5 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: industry)
Sales representatives and buyers, 12 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Skilled field occupations (care workers, technicians, social work), 43 out of 500; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Service and sales occupations, 19 out of 500; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Manual workers, farmers and manual occupations, 260 out of 500; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: public administration)
Executives and managers, 41 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: public administration)
Engineers, consultants and research officers, 28 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: public administration)
Lawyers, 19 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: public administration)
Communication and marketing officers, 9 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: public administration)
Human resources officers, 9 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: public administration)
Accountants and finance professionals, 9 out of 500; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: public administration)
Developers, 9 out of 500; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: public administration)
Office employees and technicians, 144 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: public administration)
Assistants and secretaries, 17 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Skilled field occupations (care workers, technicians, social work), 42 out of 500; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Service and sales occupations, 72 out of 500; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Manual workers, farmers and manual occupations, 101 out of 500; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: education)
Executives and managers, 20 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: education)
Researchers, 30 out of 500; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: education)
Office employees and technicians, 75 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: education)
Assistants and secretaries, 5 out of 500; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Teacher, 1 year (if sector: education)
Teachers, 255 out of 500; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Skilled field occupations (care workers, technicians, social work), 40 out of 500; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Service and sales occupations, 43 out of 500; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Manual workers, farmers and manual occupations, 32 out of 500; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Invoice and email automation: 58% of organisations of this size use AI, 38% of the staff of companies that use AI work in a company that uses it for administration Source: Insee
Meeting transcription (if tooling: conversational assistant or agents)
Meeting transcription for 40% of staff: 58% of organisations use AI, 34% of these use speech recognition; with AI in the suite, meetings are already covered in the occupations. Source: Insee
Automated customer service (if sector: market services)
Automated customer service, 100 conversations per year per employee: 58% of organisations use AI, 31% of the staff of companies that use AI work in a company that uses it for marketing or sales Source: Insee
Assistant on internal documentation, 40 pages per employee: 58% of organisations use AI, 43% of these generate text Source: Insee
Large company of 5,000 people, 1 year55 assumptions
Manager or executive, 1 year (if sector: market services)
Executives and managers, 516 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: market services)
Engineers, consultants and research officers, 313 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: market services)
Lawyers, 61 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: market services)
Researchers, 51 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: market services)
Communication and marketing officers, 68 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Journalist, 1 year (if sector: market services)
Journalists, authors and translators, 36 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: market services)
Human resources officers, 69 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: market services)
Accountants and finance professionals, 300 out of 5000; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: market services)
Developers, 220 out of 5000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Data scientist, 1 year (if sector: market services)
Data scientists, 52 out of 5000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: market services)
Graphic designers, architects and designers, 106 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Videographer, 1 year (if sector: market services)
Videographers and image professionals, 57 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: market services)
Office employees and technicians, 474 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: market services)
Assistants and secretaries, 132 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Customer service officer, 1 year (if sector: market services)
Customer account officers, 74 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: market services)
Sales representatives and buyers, 211 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Skilled field occupations (care workers, technicians, social work), 386 out of 5000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Service and sales occupations, 763 out of 5000; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: market services)
Manual workers, farmers and manual occupations, 1111 out of 5000; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: industry)
Executives and managers, 273 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: industry)
Engineers, consultants and research officers, 535 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: industry)
Researchers, 146 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: industry)
Accountants and finance professionals, 219 out of 5000; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: industry)
Developers, 97 out of 5000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designer, 1 year (if sector: industry)
Graphic designers, architects and designers, 49 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: industry)
Office employees and technicians, 294 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: industry)
Assistants and secretaries, 50 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Salesperson, 1 year (if sector: industry)
Sales representatives and buyers, 121 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Skilled field occupations (care workers, technicians, social work), 433 out of 5000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Service and sales occupations, 187 out of 5000; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: industry)
Manual workers, farmers and manual occupations, 2596 out of 5000; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: public administration)
Executives and managers, 406 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Consultant, 1 year (if sector: public administration)
Engineers, consultants and research officers, 279 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyer, 1 year (if sector: public administration)
Lawyers, 186 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year (if sector: public administration)
Communication and marketing officers, 93 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources, 1 year (if sector: public administration)
Human resources officers, 93 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountant, 1 year (if sector: public administration)
Accountants and finance professionals, 93 out of 5000; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developer, 1 year (if sector: public administration)
Developers, 93 out of 5000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: public administration)
Office employees and technicians, 1436 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: public administration)
Assistants and secretaries, 170 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Skilled field occupations (care workers, technicians, social work), 418 out of 5000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Service and sales occupations, 718 out of 5000; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: public administration)
Manual workers, farmers and manual occupations, 1015 out of 5000; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Manager or executive, 1 year (if sector: education)
Executives and managers, 201 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researcher, 1 year (if sector: education)
Researchers, 300 out of 5000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office worker, 1 year (if sector: education)
Office employees and technicians, 745 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Executive assistant, 1 year (if sector: education)
Assistants and secretaries, 49 out of 5000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Teacher, 1 year (if sector: education)
Teachers, 2553 out of 5000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Skilled field occupations (care workers, technicians, social work), 401 out of 5000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Service and sales occupations, 435 out of 5000; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year (if sector: education)
Manual workers, farmers and manual occupations, 316 out of 5000; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Invoice and email automation: 58% of organisations of this size use AI, 38% of the staff of companies that use AI work in a company that uses it for administration Source: Insee
Meeting transcription (if tooling: conversational assistant or agents)
Meeting transcription for 40% of staff: 58% of organisations use AI, 34% of these use speech recognition; with AI in the suite, meetings are already covered in the occupations. Source: Insee
Automated customer service (if sector: market services)
Automated customer service, 100 conversations per year per employee: 58% of organisations use AI, 31% of the staff of companies that use AI work in a company that uses it for marketing or sales Source: Insee
Assistant on internal documentation, 40 pages per employee: 58% of organisations use AI, 43% of these generate text Source: Insee
Team of 12 people with an assistant, 1 year1 assumption
One office employee-year per person on the team
Coding agents for a development team, 1 year1 assumption
One developer-year working with coding agents
Content production (writing and images), 1 year2 assumptions
The articles written with AI, every week of the year
The images generated, attempts included, every week of the year
Assistant licences for all employees, 1 year1 assumption
One licence for an assistant built into the office suite per employee
Assistant on internal documentation, 1 year2 assumptions
The documentation indexed, twice a year (full update)
The questions put to the documentation: 30% of employees, 2 questions per working day, 6 retrieved passages read for each question
Automated customer service, 1 year1 assumption
One customer service conversation handled by AI
Process automation (invoices, emails), 1 year3 assumptions
The invoices read and entered by an automated flow, 40 per year per employee
The emails of shared mailboxes sorted by a specialised model, 200 per year per employee
The draft replies written by AI, for 30% of the emails sorted
Data analysis with AI, 1 year1 assumption
One spreadsheet analysis per week per analyst, 44 weeks
Machine translation of a website, 1 year1 assumption
Each page of the site translated once a year into each language
Meeting transcription, 1 year1 assumption
The hours of meetings transcribed, once for all participants
In-house training with an AI tutor, 1 year1 assumption
Two conversations of ten exchanges per hour of training
AI feature in a product, 100,000 users, 1 year1 assumption
The queries of active users each day, one page read and one short text written
Open model hosted on one's own servers, 1 year1 assumption
Two GPUs switched on all year, at 60% of the consumption of an hour of computing on average
Retraining a model, 1 year1 assumption
Four retrainings per year, of 24 hours on 8 GPUs Source: Dettmers et al.
Territories
131 assumptions · 14 entries
Île-de-France, use by residents, 1 year28 assumptions
12 to 15 years: 656,000, French share increased by 15% for a highly urban region Source: Insee
16 to 24 years: 1.46 million Source: Insee
25 to 64 years: 6.58 million Source: Insee
65 to 89 years: 2.0 million Source: Insee
The 8% of personal users who pay for a subscription Source: Crédoc
Secondary school pupil, 1 year
Lower secondary pupils in Île-de-France: 663,000 Source: DEPP
Secondary school pupil, 1 year
Upper secondary pupils in Île-de-France: 450,000 Source: DEPP
Students in Île-de-France: 793,000, about 26% of students in France Source: SIES
Executives and managers in Île-de-France: 647,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Engineers, consultants and research officers in Île-de-France: 399,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Lawyers in Île-de-France: 77,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Researchers in Île-de-France: 65,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Communications officer, 1 year
Communication and marketing officers in Île-de-France: 87,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Journalists, authors and translators in Île-de-France: 46,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Human resources officers in Île-de-France: 87,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Accountants and finance professionals in Île-de-France: 241,000; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Insee
Developers in Île-de-France: 290,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Insee
Data scientists in Île-de-France: 66,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Insee
Graphic designers, architects and designers in Île-de-France: 148,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Videographers and image professionals in Île-de-France: 81,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Insee
Office employees and technicians in Île-de-France: 431,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Insee
Assistants and secretaries in Île-de-France: 121,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Insee
Customer service officer, 1 year
Customer account officers in Île-de-France: 65,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Insee
Sales representatives and buyers in Île-de-France: 179,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Insee
Teachers in Île-de-France: 272,000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Insee
Field-based occupation, 1 year
Skilled field occupations (care workers, technicians, social work) in Île-de-France: 719,000; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Insee
Field-based occupation, 1 year
Service and sales occupations in Île-de-France: 760,000; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Insee
Field-based occupation, 1 year
Manual workers, farmers and manual occupations in Île-de-France: 992,000; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Insee
France, use by residents, 1 year21 assumptions
Executives and managers in France: 2.15 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Engineers, consultants and research officers in France: 1.19 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyers in France: 231,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researchers in France: 193,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year
Communication and marketing officers in France: 261,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Journalists, authors and translators in France: 138,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources officers in France: 261,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountants and finance professionals in France: 1.29 million; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developers in France: 864,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Data scientists in France: 198,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designers, architects and designers in France: 442,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Videographers and image professionals in France: 243,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office employees and technicians in France: 2.31 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Assistants and secretaries in France: 650,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Customer service officer, 1 year
Customer account officers in France: 346,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Sales representatives and buyers in France: 957,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Teachers in France: 1.36 million; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year
Skilled field occupations (care workers, technicians, social work) in France: 3.77 million; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker Source: Eurostat
Field-based occupation, 1 year
Service and sales occupations in France: 4.21 million; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker Source: Eurostat
Field-based occupation, 1 year
Manual workers, farmers and manual occupations in France: 7.97 million; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker Source: Eurostat
Pupils and students, studies, counted once only
European Union, use by residents, 1 year29 assumptions
12 to 15 years: 18.9 million, 38% use AI for themselves, 92% without a paid subscription Source: Eurostat
The 8% of personal users who pay for a subscription, as in France Source: Crédoc
The 8% of personal users who pay for a subscription, as in France Source: Crédoc
The 8% of personal users who pay for a subscription, as in France Source: Crédoc
65 to 89 years: 94.5 million, from 2 to 6% depending on age, 92% without a paid subscription Source: Eurostat
The 8% of personal users who pay for a subscription, as in France Source: Crédoc
Secondary school pupil, 1 year
12 to 15 years who use AI for their studies, assumption of 30% Source: Eurostat
Secondary school pupil, 1 year
16 to 19 years who use AI for their studies, 50%, of whom 60% at upper secondary school Source: Eurostat
16 to 24 years in higher education who use AI for their studies Source: Eurostat
Executives and managers in the European Union: 10.49 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Engineers, consultants and research officers in the European Union: 6.86 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Lawyers in the European Union: 1.53 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Researchers in the European Union: 1.11 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Communications officer, 1 year
Communication and marketing officers in the European Union: 1.43 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Journalists, authors and translators in the European Union: 919,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Human resources officers in the European Union: 1.43 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Accountants and finance professionals in the European Union: 8.26 million; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Developers in the European Union: 5.31 million; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Data scientists in the European Union: 1.13 million; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Graphic designers, architects and designers in the European Union: 2.27 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Videographers and image professionals in the European Union: 1.28 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Office employees and technicians in the European Union: 15.97 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Assistants and secretaries in the European Union: 4.60 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Customer service officer, 1 year
Customer account officers in the European Union: 3.48 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Sales representatives and buyers in the European Union: 5.81 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Teachers in the European Union: 10.69 million; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Field-based occupation, 1 year
Skilled field occupations (care workers, technicians, social work) in the European Union: 23.39 million; share who use AI at work according to adoption data, AI use 0.65 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Field-based occupation, 1 year
Service and sales occupations in the European Union: 31.61 million; share who use AI at work according to adoption data, AI use 0.45 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
Field-based occupation, 1 year
Manual workers, farmers and manual occupations in the European Union: 59.23 million; share who use AI at work according to adoption data, AI use 0.27 times that of the average worker; × 0.95 to come back to the 21.8% from Eurostat for the Union Source: Eurostat
World, use by residents, 1 year4 assumptions
The 15 to 64 year olds worldwide who have used generative AI (18.8%), of whom 78% for themselves Source: Microsoft
The same, of whom 47% at work, counted as office employees Source: Microsoft
Secondary school pupil, 1 year
The same, of whom 28.5% for their studies, half at lower and upper secondary school Source: Microsoft
The other half in higher education Source: Microsoft
Under 25 years in France, personal use, 1 year3 assumptions
12 to 15 years: 3.31 million, 40% use AI for themselves, 92% without a paid subscription Source: Eurostat
12 to 15 years, the 8% of personal users who pay for a subscription Source: Crédoc
16 to 24 years, the 8% of personal users who pay for a subscription Source: Crédoc
25 to 64 years in France, personal use, 1 year1 assumption
25 to 64 years, the 8% of personal users who pay for a subscription Source: Crédoc
65 years and over in France, personal use, 1 year2 assumptions
75 to 89 years: 6.7 million, 3% use AI for themselves Source: Eurostat
65 years and over, the 8% of personal users who pay for a subscription Source: Crédoc
Pupils and students in France, studies, 1 year2 assumptions
Secondary school pupil, 1 year
Lower secondary pupils: 3.35 million, 40% get help from AI with their homework Source: Crédoc
Secondary school pupil, 1 year
Upper secondary pupils: 2.27 million, 66% of 16 to 19 year olds use AI for their studies Source: Eurostat
Office employees in France, work, 1 year16 assumptions
Executives and managers in France: 2.15 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Engineers, consultants and research officers in France: 1.19 million; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Lawyers in France: 231,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Researchers in France: 193,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Communications officer, 1 year
Communication and marketing officers in France: 261,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Journalists, authors and translators in France: 138,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Human resources officers in France: 261,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Accountants and finance professionals in France: 1.29 million; share who use AI at work according to adoption data, AI use 1.6 times that of the average worker Source: Eurostat
Developers in France: 864,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Data scientists in France: 198,000; share who use AI at work according to adoption data, AI use 2.5 times that of the average worker Source: Eurostat
Graphic designers, architects and designers in France: 442,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Videographers and image professionals in France: 243,000; share who use AI at work according to adoption data, AI use 2 times that of the average worker Source: Eurostat
Office employees and technicians in France: 2.31 million; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Assistants and secretaries in France: 650,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Customer service officer, 1 year
Customer account officers in France: 346,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Sales representatives and buyers in France: 957,000; share who use AI at work according to adoption data, AI use 1.5 times that of the average worker Source: Eurostat
Public servants in France, work, 1 year14 assumptions
Executives and managers in the civil service: 110,000; AI use 2 times that of the average worker Source: Insee
Engineers, consultants and research officers in the civil service: 200,000; AI use 2 times that of the average worker Source: Insee
Lawyers in the civil service: 50,000; AI use 2 times that of the average worker Source: Insee
Researchers in the civil service: 100,000; AI use 2 times that of the average worker Source: Insee
Communications officer, 1 year
Communication and marketing officers in the civil service: 30,000; AI use 2 times that of the average worker Source: Insee
Human resources officers in the civil service: 80,000; AI use 2 times that of the average worker Source: Insee
Accountants and finance professionals in the civil service: 100,000; AI use 1.6 times that of the average worker Source: Insee
Developers in the civil service: 70,000; AI use 2.5 times that of the average worker Source: Insee
Office employees and technicians in the civil service: 850,000; AI use 1.5 times that of the average worker Source: Insee
Assistants and secretaries in the civil service: 150,000; AI use 1.5 times that of the average worker Source: Insee
Teachers in the civil service: 1.00 million; AI use 0.65 times that of the average worker Source: Insee
Field-based occupation, 1 year
Skilled field occupations (care workers, technicians, social work) in the civil service: 1.30 million; AI use 0.65 times that of the average worker Source: Insee
Field-based occupation, 1 year
Service and sales occupations in the civil service: 950,000; AI use 0.45 times that of the average worker Source: Insee
Field-based occupation, 1 year
Manual workers, farmers and manual occupations in the civil service: 860,900; AI use 0.27 times that of the average worker Source: Insee
ChatGPT, 1 day4 assumptions
76% of messages: an ordinary exchange, one tenth of a ten-exchange conversation Source: OpenAI
10% of messages with reasoning Source: OpenAI
10% of messages that launch a web search Source: OpenAI
4% of messages that create an image, out of the 7% of "multimedia" messages Source: OpenAI
Mistral Vibe, 1 day1 assumption
Le Chat answers over one day, counted as a question to an average model with an answer of 400 tokens, that of the Mistral AI analysis, which publishes only carbon and water; assumed volume, Mistral AI does not publish it Source: Arcom
Gemini app, questions and images, 1 day2 assumptions
The questions asked in one day by the billion monthly users, 0.7 per user Source: Google
More than 150 million images generated every day, according to Google; the electricity of a Gemini image is not published, that of an open generator of 2023 stands in for it Source: Google
AI assistants worldwide, people's requests, 1 day4 assumptions
Gemini
Le Chat
The AI-generated overviews at the top of Google results, the search itself not counted
The other assistants (Meta AI, Copilot, Claude, Grok, Perplexity, Chinese assistants), counted as a share of ChatGPT Source: Arcom
