Employer rules on generative AI and how much workers pay out of pocket per month (employees and civil servants)

KEY FINDINGS

What this survey found

Percentages use the people who answered each question as the base. The breakdown of each base is shown above the chart for each question.

• What is your company's policy on using AI for work? 44.2% No policy 122 of 276 people (base: respondents) · See all results for this question
• Does your company provide its own AI (such as a company-contracted ChatGPT or Copilot plan)? 50.7% None 140 of 276 people (base: respondents) · See all results for this question
• Who pays for the paid plan of the AI tool you use most at work? 43.5% I use only the free plan 104 of 239 people (base: People who use AI for work) · See all results for this question

SAMPLE

From collected to valid responses

This shows how many of the responses received we excluded, and on what grounds. The exclusion rules were set before collection began and were not changed along the way.

Percentages below use the 276 valid responses as the base. The target was 300 responses.

RESULTS

Results by question

Each heading shows the most common answer. It is generated automatically from the tallies. The bars and the tables show the same figures. Click “View as a table” to open the table.

BREAKDOWN

By respondent profile

Where the base is small, we show neither the percentage nor the count. Otherwise they could be worked out by subtracting other cells.

Do you use AI (such as ChatGPT) for work? (by Your age group)

Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

What is your company's policy on using AI for work? (by Number of employees at your workplace)

Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

Do you use AI (such as ChatGPT) for work? (by Your job type)

Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

Who pays for the paid plan of the AI tool you use most at work? (by Do you use AI (such as ChatGPT) for work?)

Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

Who pays for the paid plan of the AI tool you use most at work? (by Does your company provide its own AI (such as a company-contracted ChatGPT or Copilot plan)?)

Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

VOICES

What respondents wrote

The last question was free text. From the answers, we picked ones that contain nothing that could identify a person or their employer, and show them exactly as written. They are not included in the aggregate data.

• There are no clear rules, so I don't know if using AI is actually a good thing. Some people use AI on their personal phones to look things up (like how to fill out documents), but they keep it secret from their boss that they used AI to create or research something. I'd like the company to make clear the environment and rules for when AI use is allowed.

  40s, Retail/Wholesale, 29 or fewer employees, Administrative (contract/temp staff)

• The criteria for how much company information can be entered into AI are a bit hard to understand, so I'd feel more confident using it for work if more specific guidelines and examples were provided.

  40s, Manufacturing, 30–99 employees, Administrative

• Since this is a medical workplace, I understand that content involving personal information can't be used with AI. There are many other areas where it could improve efficiency, but right now each person uses AI on their own. Efficiency would improve further with a paid plan, but the pricing structure is hard to understand, and I don't know how to encourage adoption.

  40s, Healthcare/Welfare, 300–999 employees, Section chief/Supervisor

• The company recommends using AI to improve work efficiency, but since it doesn't provide specific methods, only the people who already know how to use AI are seeing efficiency gains. If this were just for personal use, that would be understandable, but if the company is going to recommend its use, it would help beginners more to have a manual with a collection of effective prompts.

  20s, Real Estate/Construction, 1,000 or more employees, Administrative

METHODOLOGY

Methodology

We show everything: who we asked, how we asked and how we counted.The figures are produced by a program only. We do not let AI do any counting.

Conducted by

AInformation

Survey period

September 18, 2026 (Asia/Tokyo)

How responses came in

300 responses from 10:11 a.m. to 10:13 p.m..The 10 a.m. hour had the most, with 106 responses.We closed data collection once we reached 300 responses

Target respondents

Respondents to an open online call for participants who said they are employees (including contract and temporary staff) or civil servants

Survey method

Internet survey (web questionnaire) / Respondents recruited online (first-come, one response per person, limited to respondents with a good track record on past surveys)

Eligibility requirements

Company employee (including contract and temp staff) or public servant

How people took part

Optional. There is no penalty for not answering. Answers come from people who signed up and answered on their own. This is not a randomly selected sample.

Collected and valid responses

Of the 300 responses collected, we excluded 24 responses from the tally and counted 276 responses as valid. The target was 300 responses.

Weighting

None (unweighted). Because respondents took part voluntarily, there is no basis for adjusting the results to represent Japan as a whole. We do not add margins of error (±X%) either.

Funding

Costs are covered by AInformation (Shunta Fujii). There is no sponsor.

Questionnaire and quality check versions

Questionnaire v1 / Quality check rules v1 / Published version v2

Questions and answer options

These are the questions and answer options as we presented them. For questions with branching, we also note who was meant to answer.

• 1. Choose the option that best describes how you work Choose one
  Company employee (full-time) / Company employee (contract or temporary staff) / Public servant / Business owner or executive / Self-employed or freelance / Part-time worker / Student / Other See the results
• 2. Your age group Choose one
  20s / 30s / 40s / 50s / 60s or older See the results
• 3. Number of employees at your workplace Choose one
  29 or fewer / 30–99 / 100–299 / 300–999 / 1,000 or more / Don't know See the results
• 4. Industry of your employer Choose one
  Manufacturing / IT and telecom / Retail and wholesale / Finance / insurance / Real estate and construction / Healthcare / welfare / Education / Services / Public sector / Other See the results
• 5. Your job type Choose one
  Sales / Administration and management / Planning or marketing / Engineering and development / Design and production / HR and general affairs / Accounting and finance / Customer support / Manufacturing and on-site work / Other See the results
• 6. Your job title Choose one
  General staff / Chief / section chief / Department manager / Division manager or above / Executive See the results
• 7. Do you use AI (such as ChatGPT) for work? Choose one
  Almost every day / A few times a week / A few times a month / Have used it before but not now / Never used it See the results
• 8. When did you start using AI for work? Choose one; asked of People who use AI for work (239 people)
  Within the last 6 months / 6 months to 1 year ago / 1 to 2 years ago / 2 or more years ago / Don't use it See the results
• 9. How much time do you spend using AI for work each day? Choose one; asked of People who use AI for work (239 people)
  Under 10 minutes / 10 to 30 minutes / 30 minutes to 1 hour / 1 to 3 hours / 3 hours or more / Don't use it See the results
• 10. Which AI do you use most for work? Choose one; asked of People who use AI for work (239 people)
  ChatGPT / Gemini / Copilot (Microsoft) / Claude / Perplexity / NotebookLM / The company's own in-house AI / Other / Don't use it See the results
• 11. What do you use AI for at work? (select all that apply) Choose all that apply; asked of People who use AI for work (239 people)
  Drafting emails and documents / Summaries and meeting minutes / Translation / Creating materials and slides / Spreadsheets and data organization / Programming / Image creation / Research / Brainstorming ideas / Other / Don't use it See the results
• 12. Why don't you use AI for work, or why did you stop? (Select all that apply) Choose all that apply; asked of People who do not use AI for work (37 people)
  No need for it / The company prohibits it / I don't know how to use it / Worried about entering company information / I don't trust the answers / It costs money / No one around me uses it / Other / I use it See the results
• 13. What is your company's policy on using AI for work? Choose one
  Only AI provided by the company / Specific approved AI tools only / Tools you subscribe to personally can also be used freely / Use is prohibited / No policy / Don't know the policy See the results
• 14. Which of the following are included in your company's policy? (select all that apply) Choose all that apply
  Information that must not be entered is specified / Specific approved AI tools only / AI-generated content must be checked by a person / AI-generated content must be labeled as such / Can't use it without training / Can use it after applying / No particular policy / Don't know the policy See the results
• 15. Do you follow your company's policy? Choose one
  I follow it / I sometimes don't follow part of it / I don't follow it / No policy / Don't know the policy See the results
• 16. Does your company provide its own AI (such as a company-contracted ChatGPT or Copilot plan)? Choose one
  Yes (I use it) / Yes (I don't use it) / None / Don't know See the results
• 17. Are you satisfied with your company's AI policy? Choose one
  Satisfied / Somewhat satisfied / Somewhat dissatisfied / Dissatisfied / No policy / don't know See the results
• 18. What problems do you have when using AI for work? (select all that apply) Choose all that apply
  It takes time to check if the answer is correct / Not sure if it's okay to enter internal company information / Not sure which AI tools are allowed / No place to learn how to use it / Few people around me use it / The fee is too high / Hard to admit that I used AI / No particular problems / Don't use it See the results
• 19. What do you want from your company? (select all that apply) Choose all that apply
  Company covers the cost of paid plans / Clearly define the scope of acceptable use / Training on how to use it / Introduce AI tools employees can use at work / Have someone available for consultation / Reflect it in performance evaluations / Nothing in particular See the results
• 20. Who pays for the paid plan of the AI tool you use most at work? Choose one; asked of People who use AI for work (239 people)
  The company pays / I pay for it myself / I use only the free plan / Don't know / Don't use it See the results
• 21. If you pay for it yourself, how much do you pay per month? Choose one; asked of People who pay for AI tools used at work out of their own pocket (12 people)
  Under ¥1,000 / ¥1,000–2,999 / ¥3,000–4,999 / ¥5,000–9,999 / ¥10,000 or more / I don't pay for any myself See the results
• 22. Why do you pay for it yourself? (select all that apply) Choose all that apply; asked of People who pay for AI tools used at work out of their own pocket (12 people)
  The company's AI isn't enough / Applying for it is a hassle / I don't want the company to know / I also use it outside of work / The company won't cover it / Other / I don't pay for any myself See the results
• 23. Have you ever asked your company to cover the cost of a paid plan? Choose one; asked of People who use AI for work (239 people)
  Asked and they covered it / Asked but was refused / Haven't asked (hard to bring up) / Haven't asked (not necessary) / The company provided it from the start / Don't use it See the results
• 24. If you paid for work AI yourself, how much would you be willing to pay per month? Choose one
  ¥0 (would not pay) / Under ¥1,000 / ¥1,000–2,999 / ¥3,000–4,999 / ¥5,000 or more See the results
• 25. How much has using AI reduced your work time per day? Choose one; asked of People who use AI for work (239 people)
  Almost no change / Less than 30 minutes / 30 minutes to 1 hour / 1 to 2 hours / 2 hours or more / Don't use it See the results
• 26. How is using AI viewed in your workplace? Choose one
  Using it is valued / It doesn't really matter / There is an atmosphere where using it is looked down on / Don't know See the results
• 27. If there is anything you disagree with or struggle with regarding your company's AI rules or how AI is used, please tell us (optional) Open-ended See the results

How we tallied

• The percentages use the number of people who answered that question as the denominator (non-responses are excluded from the denominator).
• The results are not weighted. They come from a group of respondents who volunteered to take part and do not represent Japan as a whole.
• Percentages shown are rounded to one decimal place. They may not add up to 100%.
• For questions that allow multiple answers, the totals can add up to more than 100%.
• In some multiple-answer questions, respondents selected “no use” together with a specific option. These counts are included in the checks for each question.
• Results with a base of fewer than 30 people are for reference only, and results with fewer than 10 are not shown. This is an operating rule of this site, not a standard that guarantees statistical significance or anonymity.

What this survey cannot tell us

• This survey was answered by people who responded to an open online call for participants. It does not represent all company employees and public servants in Japan.
• All answers are self-reported. We did not independently verify actual usage or workplace rules.
• Answers about workplace rules reflect only what the respondent knows. They may differ from what the company has actually set.
• “Time saved by AI” is the amount of time respondents felt they saved. It is not a measured value.
• We limited responses to one per person, but this cannot fully prevent the same person from answering again under a different registration.
• We received 300 responses on the day the call for participants was posted. Respondents are those who noticed this call early and applied.
• Answers come from people who signed up and answered on their own. Because this is not a random sample, the figures cannot be read as percentages for Japan as a whole.
• Time spent and effects are self-reported by respondents. We did not measure them.

DATA

Published aggregate data

The files below contain the same figures as this page. The JSON and CSV files have the same content. We do not delete versions (v1, v2 and so on). That way, links from people who cited them keep working.

• CSV Latest version (Japanese) https://ainformation.jp/data/research/worker-ai-rules-2026-09/latest.csv (118KB)
• JSON Latest version https://ainformation.jp/en/data/research/worker-ai-rules-2026-09/latest.json (91KB)
• CSV v2 (as published) (Japanese) https://ainformation.jp/data/research/worker-ai-rules-2026-09/v2.csv (118KB)
• JSON v2 (as published) https://ainformation.jp/en/data/research/worker-ai-rules-2026-09/v2.json (91KB)
• CSV v1 (as published) (Japanese) https://ainformation.jp/data/research/worker-ai-rules-2026-09/v1.csv (118KB)
• JSON v1 (as published) https://ainformation.jp/en/data/research/worker-ai-rules-2026-09/v1.json (90KB)

The CSV is UTF-8 with a BOM, so it opens directly in Excel. The files do not include individual responses, the original free-text answers, names, email addresses, account IDs on the recruiting platform or response dates and times.

Data field descriptions

How to cite

Source: AInformation, “Generative AI rules at work and out-of-pocket monthly spending (company employees and public servants)” (https://ainformation.jp/research/worker-ai-rules-2026-09), as of September 18, 2026

Copy citation

Terms of use

• The published aggregate data and the tables and bar charts on this page are available under CC BY 4.0. As long as you credit the source, you may quote them in articles, documents, research and apps, adapt them and redistribute them. Commercial use is allowed.
• Article text, images and chart designs are not covered by this license. If you want to use them beyond what quotation allows, please contact us via Contact.
• Free-text answers quoted in the article were written by respondents. They are not included in the aggregate data. Please use them only to the extent that quotation allows.
• If you reproduce the figures as they are, please state when the survey was conducted and that they are “shares among respondents.” The full terms are on the About us page.

FAQ

FAQ

Who runs these surveys?

AInformation (run by Shunta Fujii) plans the surveys and pays for them itself. No one commissions or sponsors them. We never create numbers to serve anyone's advertising.

How many people do you survey?

We set a target number of respondents for each survey before recruiting. Surveys published so far: 3, with 602 valid responses in total. Each survey's page shows how many responses we collected and how many we removed in quality checks. The base for each question is on the same page.

Can I use these figures as “X% of people in Japan”?

No. The survey asked people who answered an open online call for respondents, and the figures are shares of those respondents. The sample cannot be used to estimate figures for Japan as a whole, so we do not give a margin of error (±X%) either.

Do you use AI to tally the results?

All tallying is done by a program, not by AI. We use AI only for groundwork when reading open-ended answers and picking candidates for articles, and a person makes the final decision.

Can I use the data in my own materials or articles?

Yes. Please cite the AInformation survey name and URL as the source. The tallied figures are available as JSON and CSV. You may modify and redistribute them.

Could anyone tell which answers are mine?

No. We never publish anything in a form that could identify individuals. We do not publish the original open-ended answers either. Anything used in articles is chosen by a person after removing details that could identify an individual or a company.

Can I withdraw my answers?

Yes. Contact us through the contact form. If we have not tallied the results yet, we will remove your response. We keep responses for 2 years and delete them after that.

What will you survey next, and when?

Next, we will start a recurring survey that asks the same questions every time. Asking the same questions over time lets us show changes from the previous round in percentage points. We will post the topics and timing on this page when recruitment starts.

HISTORY

Corrections and updates

• 2026-09-23T16:02:12+09:00 · We had missed some free-text answers, so we collected them again and re-entered them

Survey and tallying by

AInformation

Reviewed by

Shunta Fujii (About us)

Publication date

September 23, 2026

Published version

v2

Related pages

• List of Original AI SurveysSee other surveys
• AI Model BenchmarkReal measurements: the same questions sent to each AI