GPT-6.2 Astra vs GPT-6.2 Sol vs Opus 5.5 vs Fable 5.2Researching workplace and business-partner relationships in Japan
We compared GPT-6.2 Astra and GPT-6.2 Sol from OpenAI with Claude Opus 5.5 and Claude Fable 5.2 from Anthropic. At 10:40 JST on 18 November 2026, we sent the same research request, written in Japanese, to all four models within 40 seconds. Every model ran at the "high" reasoning setting. All four used the same research skill (d-search) with the instruction to keep costs down (go cheaper).
There is one conclusion. When you hit a wall, try another model and see whether a different direction comes out.
When you hit a wall, try another model and see whether a different direction comes out.
This time, the clear difference was where each model went looking. Three reached the same conclusion, and only Astra went looking for how the staff on the two sides of a business relationship behave toward each other. Adding one more model can bring out a different direction like this.
Honestly, there is no best model. Even pinning down what each one is like is hard. Most of the differences came from the conditions of each run. The token gap came from what each run used for its child tasks, and Fable's larger number of facts came from its extra research.
Naming the direction you want searched in the request is another step you can take before trying a different model.
Whatever the model, you still need to check the sources. In 3 of the 30 cited passages, the figures the model wrote under the quote did not match.
Brief · what to pick
Which run for which job.
Work time runs across, and tokens in the logs run up. A bigger point means more facts in the research note. Toward the lower left is faster and uses fewer tokens.
Astra
Sol
Opus
Fable
Fable first report
5 min10 min15 min20 min
02M4M6M8M
Work timeTokens used
Facts in the note102030Fable up to its first report (12 min 35 s, 4.58M tokens, 18 facts)
Astra7 min 30 s, 2.58M tokens, 10 facts in the noteSol8 min 24 s, 2.67M tokens, 7 facts in the noteOpus9 min 13 s (not counting 9 s waiting for our answer to its question), 4.56M tokens, 13 facts in the noteFable18 min 20 s (not counting 11 min 30 s waiting for our answer), 7.17M tokens, 27 facts in the note
Each model ran once.
Prompt
The prompt sent to all four runs.
As sent, in Japanese
日本の職場で発生する人間関係の問題について調査して 取引先との人間関係の問題も含む
ファクトとAI推論は分けて出力 アウトプットは日本語
$d-search go cheaper
English translation
Research the interpersonal problems that occur in Japanese workplaces Include interpersonal problems with business partners
Output facts and AI inferences separately Output in Japanese
$d-search go cheaper
"$d-search" calls Dorado's research skill. It quotes the original text of each source and keeps confirmed facts apart from the AI's inferences. "go cheaper" asks for cheaper research methods without changing the scope or depth.
Measured · elapsed time
Which model finished first?
7 min 30 sAstraanswered at 10:48
8 min 24 sSolanswered at 10:48
9 min 13 sOpusanswered at 10:49
18 min 20 sFableanswered at 11:09
Work time
Each bar is the work time from sending the prompt to the last answer. Time the model spent waiting for our reply is not in the bar and shows as a dotted line. In Fable's bar, the dark part runs up to its first report and the light part is its extra research.
Astra
7 min 30 s. Answered at 10:48
Sol
8 min 24 s. Answered at 10:48
Opus
waiting 9 s
9 min 13 s. Answered at 10:49. Right after starting, it asked, "The same request is running in two other threads. Is this for a comparison?" It got our reply 9 s later and carried on
Fable
To first report 12 min 35 s
Extra research 5 min 45 s
waiting 11 min 30 s
18 min 20 s. First report at 10:52 (18 facts). It had reached its 4M-token budget, so it asked whether to research 5 more items. It got our reply, "research the 5 items", 11 min 30 s later and sent its final report at 11:09
0 min5 min10 min15 min20 min25 min30 min
Fable: up to its first reportFable: extra researchTime spent waiting for our reply
The two GPT-6.2 models asked nothing and ran straight through to the end. The two Claude models each asked one question along the way. Opus asked to check for duplicates, and Fable asked about its budget. Counting the 11 min 30 s until we replied, Fable's final report came 29 min 50 s after we sent the prompt.
Measured · tokens
How many tokens did each run use?
2.58MAstrareported about 1.42M
2.67MSolreported about 2M
4.56MOpusreported 3.9M
7.17MFablereported about 6.9M
Tokens in the main session and the child tasks
Usage in the logs. The dark part is the main session, and the light part is the child tasks (helpers). The black tick is the figure the model itself reported last.
Astra
Main 1.76M
0.81M
2.58M (main 1.76M, 0.81M in a child task on Sol low). Reported about 1.42M
Sol
Main 2.29M
2.67M (main 2.29M, 0.38M in a child task on Sol low). Reported about 2M
Opus
Main 2.75M
1.8M
4.56M (main 2.75M, 1.8M in a child task on Sonnet 5.5). Reported 3.9M
Fable
Main 2.69M
7.17M (main 2.69M, 4.48M in 4 child tasks on Sonnet 5.5). Reported about 6.9M
02M4M6M8M
Main sessionChild tasks (Astra, Sol: GPT-6.2 Sol low. Opus, Fable: Sonnet 5.5)The model's own report
All four reported less than the logs show. The gap was largest for Astra, which reported 55% of the logged total. Astra itself wrote that "the delegate's usage is not included". Sol reported 75% of its logged total, Opus 86%, and Fable 96%.
Both GPT-6.2 models set a budget of 1.5M tokens before they started researching, and both used 1.7 to 1.8 times that. Opus used 4.56M against a 4M budget. Fable stopped once at 4M, and then used 7.17M against a budget of 5.5M with the extra 1.5M added.
The two GPT-6.2 models used GPT-6.2 Sol at the low setting for their child tasks. The two Claude models used Sonnet 5.5 for theirs.
Measured · link check of the answers
Does each chat answer hold up as written?
After the runs, we opened every cited link in the answers and compared each page with the passage that cites it. A passage counts as backed when every figure, name, date, and condition in it is on the page.
Of the 30 cited passages in the four answers, 27 were fully backed. The other 3 were partly backed. None were unbacked, and none were dead links.
7 of 7Astrafully backed
6 of 6Solfully backed
5 of 6Opusfully backed
9 of 11Fablefully backed
How well the cited passages are backed
For each model, the share of its cited passages that were fully backed, partly backed, or not backed or unverifiable.
Astra
7
7 of 7
Sol
6
6 of 6
Opus
5
1
5 of 6
Fable
9
2
9 of 11
0%25%50%75%100%
BackedPartly backedNot backed or unverifiable
In all 3 partly backed passages, the error was not in the quote itself. It was in the note the model added under the quote.
Fable: in its note on the number of workers' compensation awards, it wrote that "the top three were all about relationships with people". The third item in the quote is "a major change in the content or amount of work", which is not about relationships with people.
Fable: it gave the number of companies answering the company survey as "1,880". The figure in the chart is 1,870.
Opus: it gave the respondents to the power harassment question as "n=7,774". 7,774 is the number of companies answering the company survey. The worker survey had 8,000 respondents.
The dates and source locations also had small errors.
Seven passages dated the survey report "17 May 2024" (Fable 3, Astra 2, Sol 2). The PDF cover is dated March 2024 (Reiwa 6), and 17 May appears nowhere in the PDF.
Fable wrote "no publication date given" for the Industrial Safety and Health Survey. Page 1 of the PDF gives 7 August 2026 (Reiwa 8).
Opus wrote that "the Ministry of Health, Labour and Welfare's own page still described the effective date in old wording" and cited a regional labour bureau's page instead. On the day we checked, the ministry's own page gave the effective date of 1 October 2026 (Reiwa 8).
The Ministry of Economy, Trade and Industry page returned a 403 error on the first automated fetch. It opened when we fetched it again with the same settings as a browser, so we count it as backed.
Compared · what each answer found
Where the four answers agree, and where they part.
Facts in the answers
The facts in the four answers, grouped as "inside the company", "business partners", and "terms of trade". A dot means that model wrote the fact in its answer.
Fact written in the answer
Astra
Sol
Opus
Fable
Total
Inside the company
Used the Survey on Workplace Harassment (MHLW-commissioned, FY2023)
4
32.9% named "interpersonal relationships" as a cause of strong stress (Industrial Safety and Health Survey, 2025)
3
19.3% had experienced power harassment (same harassment survey)
3
65.7% of power harassers were bosses
2
36.9% "did nothing" after power harassment
2
Mental-disorder compensation awards: power harassment by a boss most common, 222 cases (FY2025)
2
"Workplace relationships" as a reason for leaving the previous job (Survey on Employment Trends, 2025)
1
Top real reason for leaving is "bad relationships" (En Japan, 2024)
1
Consulted someone about a relationship problem 20.3%, did not 14.6% (JILPT)
1
27.7% have no one nearby to consult (Job Soken)
1
Sexual harassment: sexual jokes 49.7%
1
Business partners
22.6% named people at business partners as perpetrators of nuisance behavior
2
Customer harassment measures an employer obligation from 1 October 2026
2
The MHLW counts a business partner's staff among the "customers and others" in customer harassment
1
Nuisance behavior by customers 46.8% (UA Zensen, 2024)
1
Mainly business-to-business dealings: 16 cases of verbal abuse or intimidation, 5 of being kept for long periods (Hamamatsu CCI)
1
Cases of being unable to refuse a business partner's request (Food Rengo, UA Zensen)
1
67.0% sent staff with future business in mind (JFTC, 2012)
1
Terms of trade
Price negotiations 90.7% or pass-through rate 54.2% (SME Agency, March 2026)
3
6.9% held back from price negotiations for fear of fewer orders (same survey)
Freelancers' fees reduced 28.1% (JFTC and MHLW, 2024)
1
1 fact used by all four, 3 facts used by three, 5 facts used by two, and 13 facts used by only one (Astra 6, Fable 4, Sol 3, Opus 0).
The only source all four answers used was the "Survey on Workplace Harassment" (FY2023), commissioned by the Ministry of Health, Labour and Welfare. Three other facts appeared in three or more answers: 32.9% of workers named interpersonal relationships as a cause of strong stress, 19.3% had experienced power harassment, and the figures on price negotiation and passing on costs.
On relationships with business partners, the answers split in two.
Opus, Fable, and Sol looked for government statistics. They concluded that problems with business partners are counted in two separate groups: "nuisance behavior by people at a business partner (customer harassment)" and "terms of trade such as prices and fees". Opus wrote, "We did not find any government statistic that counts strained relationships with business partners themselves (things that are hard to say, awkwardness)."
Astra looked for sources that directly show how the staff on the two sides of a business relationship behave toward each other. It found verbal abuse, intimidation, and being kept for long periods in business-to-business dealings (a members' survey by the Hamamatsu Chamber of Commerce and Industry), a case where staff could not refuse a purchase request that cited last year's figures (a survey by Food Rengo and UA Zensen), and the share of suppliers that sent staff to a retailer with future business in mind (Japan Fair Trade Commission, 2012). For every one of these sources, Astra itself noted the limits of its scope in the table.
Sol picked out one figure from the trade-terms statistics that is close to relationships between people. 6.9% of answers said the company did not ask to negotiate prices because it feared fewer orders or the end of the business relationship.
13 facts appeared in only one answer: 6 from Astra, 4 from Fable, 3 from Sol, and none from Opus. Every fact in Opus's answer also appears in Fable's or Sol's.
The answers also differed in how they wrote their inferences.
Opus: sorted out what the government statistics count and what they do not.
Fable: wrote 10 inferences. One example: "The side with more power may count what it does at a lower rate than the other side does."
Sol: wrote only 2 inferences. Both are cautions on how to read in-house consultation counts and how business partners react.
Astra: wrote them as what readers should look at in their own workplace. One example: "The fact that someone accepted a request does not show that they agreed with it."
The main result · research notes
The research note each run wrote.
The answer each run returned in chat is a summary of the research note it wrote. For each fact, the note holds the source's original text and where in the source it appears. You can read all four notes in full under "Actual results" below.
Each run's research note and final answer are shown unedited. Switch between the models, and between "Research note" and "Final answer", to compare them. Fable's answer joins its three messages: the first report (sent in two parts) and the final report.
In Japanese workplaces, the reported problems include having no one to consult when needed, getting no advice that would help improve the work, and verbal abuse or exclusion by bosses and colleagues. With business partners, the research confirmed verbal abuse and being kept for long periods, and being unable to refuse a partner's request with future business in mind.
Facts
Relationship or situation
What the surveys confirmed
Not consulting anyone when a problem arises at work
When Job Soken asked 578 working registered users about stress from relationships, 27.7% said "there is no one nearby I can consult", 26.8% said "the line on harassment is hard to judge", and 26.5% said "there is no constructive feedback". 2025 Workplace Stress Survey, 17 March 2025
Power harassment by bosses and colleagues
In the Ministry of Health, Labour and Welfare survey, of 1,546 people who said they had experienced power harassment in the past three years, 65.7% named a boss other than an executive as the harasser, and 20.2% named a colleague. The behavior was threats, insults, severe verbal abuse, and the like in 48.5% of cases, excessive demands in 38.8%, and isolation, exclusion, or being ignored in 27.8%. All are multiple-answer questions. MHLW, harassers, Figure 149, 17 May 2024 / types of behavior, Figure 137, same day
Sexual conduct at work
In the same MHLW survey, of 503 people who said they had experienced sexual harassment in the past three years, 49.7% named "sexual jokes or teasing" and 26.2% named "unnecessary physical contact". Multiple answers. MHLW, Survey on Workplace Harassment, Figure 138, 17 May 2024
Verbal abuse, intimidation, and being kept for long periods in business-to-business dealings
Unable to refuse a business partner's purchase request
A survey of union members working in sales in food manufacturing and in distribution and retail includes this case: "A sales representative at a business partner cites last year's figures and asks me to report how many seasonal products we will buy, and I can't refuse." The text does not show whether the partner said outright that it would cut business if refused. Food Rengo and UA Zensen, "Survey on Trading Practices", result 10, February 2025
Sending staff with future business in mind
In a Japan Fair Trade Commission survey, of 581 suppliers that answered the question after large retailers and others asked them to send employees, 389 (67.0%) answered that "considering future business, there were times we had no choice but to agree to the request". It is a 2012 survey of retail transactions, and the current share across all industries is unknown. JFTC, Survey on Transactions between Large Retailers and Suppliers, Figure 35, 11 July 2012
AI inferences
Judging relationships by the number of in-house consultations alone will mislead you. Some people did not consult anyone even when a problem actually arose. Others named the lack of anyone nearby to consult as a source of stress in itself. However, these are two separate surveys, so they cannot be linked to explain why people did not consult anyone.
Narrowing in-house problems to "how the boss talks" misses being ignored by colleagues and how work is handed out. Besides who said what, look separately at whether the person was left out of places they needed to be and whether they were asked for work they could not do. That makes it easier to see what the person is struggling with.
In a business relationship, the fact that someone accepted a request does not show that they agreed with it. Asking how many seasonal products the other side will buy, or asking for extra staff, looks like normal business to the side making the request. For the side receiving it, whether it can keep trading on the same terms after refusing may decide the answer.
The measured use of the parent thread was about 1.42M tokens, 19 steps, about 6 minutes. The budget was 1.5M tokens, and the delegate's usage is not included in the measurement.
In Japanese workplaces, interpersonal relationships are reported as a cause of strong stress and a reason for leaving jobs. With business partners, a partner's staff can commit customer harassment, and the surveys also confirmed cases where the receiving side held back from price negotiations for fear of fewer orders.
Facts
32.9% of workers with strong work stress named interpersonal relationships.In the 2025 survey, the overall figure for "interpersonal relationships (including sexual harassment and power harassment)" was 32.9%. The question asks for up to three main reasons. Published 7 August 2026. MHLW "Industrial Safety and Health Survey" › Stress › Table 17
People who changed jobs named relationships as a reason for leaving their previous job.In 2025, "workplace relationships were not good" was given by 8.9% of men and 11.5% of women. The survey covers people who were employed in their previous job and were working at their new employer at the time of the survey. Published 31 August 2026. MHLW "Survey on Employment Trends" › Reasons people who changed jobs left their previous job › Table 5
In a survey of 8,000 workers, 19.3% had experienced power harassment in the past three years, and 6.3% sexual harassment.10.8% had experienced serious nuisance behavior by customers and others. The most common type of power harassment was psychological attacks such as threats, insults, and severe verbal abuse. "Customers and others" includes ordinary customers, and the rate for corporate business partners alone could not be confirmed. Published 17 May 2024, corrected edition of 15 August 2025. MHLW "Survey on Workplace Harassment" › Workers' experience of harm
36.9% of people who experienced power harassment answered that they "did nothing" afterward.Of those who did nothing, 65.6% gave as a reason that they thought nothing they did would solve it. Published 17 May 2024, corrected edition of 15 August 2025. MHLW "Survey on Workplace Harassment" › Actions after harm and the reasons
The MHLW includes a business partner's staff among those who can commit customer harassment.The examples in the document are "a business partner's representative" and "a representative negotiating toward a contract between companies". To count, the conduct must go beyond what is socially acceptable and harm the worker's working environment. Prepared July 2026. MHLW "Pamphlet on Harassment Measures Employers Must Take" › Examples of "customers and others"
6.9% of answers held back from price negotiations for fear of fewer orders or the end of business with the ordering company.The answer item reads: "Costs rose, but the ordering company made no offer, and fearing fewer orders or the end of business, we did not ask to negotiate." The denominator is 65,311 company answers in total, excluding answers where price negotiation was not needed. In the same tally, 90.7% negotiated prices. Published 26 June 2026. SME Agency "Results of the Follow-up Survey on the Price Negotiation Promotion Month" › Price negotiations in the last six months
AI inferences
Judging the situation by in-house consultation counts alone may underestimate the problem. Some people took no action after experiencing power harassment. Some people left jobs because of relationships, but the surveys are separate, so it is unknown whether people who did not consult anyone went on to quit. Based on 2 and 4.
In business relationships, it is easier to see what is happening if you look separately at the other side's conduct and at the fear of losing out by asking for better terms. For example, being insulted in a talk about delivery dates and not daring to ask for a later delivery date for fear of losing the next order call for different checks. The delivery date is an example for explanation. What the survey confirmed goes only as far as the receiving side's worry about price negotiations. Based on 5 and 6.
The measured result was about 2M tokens, 23 steps, about 7 minutes. Because the run read excerpts from the sources over and over, the measurement went over the 1.5M-token budget. Usage by other tasks is not included in this measurement.
Done: I saved and committed a note that keeps 13 facts apart from the AI's inferences.
Answer (AI): In government statistics, the largest problem inside companies is power harassment by bosses and others. For business partners, the government does not count "relationships" as such. It counts two kinds of problem: customer harassment (serious nuisance behavior by customers or business partners) and terms of trade, such as passing on prices and paying fees.
Facts (the 6 main ones)
MHLW survey of workers: experience of harm is highest for power harassment, then customer harassment, then sexual harassment
"On the experience of each type of harassment at work and similar places in the past three years, power harassment was 19.3% (-12.1%), sexual harassment 6.3% (-3.9%), and serious nuisance behavior by customers and others 10.8% (-4.2%), (omitted)."
Survey on Workplace Harassment, summary of results (MHLW commissioned project, FY2023) › p.12 · published May 2024 (Reiwa 6) · source
AI: The respondents are workers aged 20 to 64, and n=7,774 for the power harassment question. The figures in brackets are the change from FY2020, and all three fell. I found no edition newer than FY2023. In the chat I shortened the quote. The full text is in the note.
Same survey: 22.6% of customer harassment came from people at business partners
"On who committed serious nuisance behavior by customers and others, "customers and others (including patients or their families)" (82.3%) was higher than "employees or executives of other companies such as business partners" (22.6%)."
Same as above › Content and perpetrators of serious nuisance behavior by customers and others (survey of workers) › p.14 · source
AI: The base is only the people who experienced customer harassment, not all workers.
Workers' compensation awards for mental disorders: power harassment most common, customer harassment and sexual harassment tied for second
"By number of benefit decisions, the most common were "received power harassment such as physical or psychological attacks from a boss or others" with 222, then "received serious nuisance behavior from customers, business partners, facility users, or others" and "received sexual harassment" with 127 each, (omitted)."
Announcing the FY2025 "Status of Workers' Compensation for Karoshi and Related Cases" › Key points › (7) Trends by event · 15 July 2026 · source
AI: These are counts of recognized cases, not of incidents. Business partners and ordinary customers fall in the same category and cannot be separated.
Industrial Safety and Health Survey: 32.9% of workers named interpersonal relationships as a cause of strong stress
"(omitted) followed by "interpersonal relationships (including sexual harassment and power harassment)" at 32.9% [previous survey 26.1%]."
Overview of the 2025 (Reiwa 7) "Industrial Safety and Health Survey (Fact-finding Survey)" › [Individual survey] › p.16 · source
AI: The previous survey (2024) gave 26.1%. The 2024 survey changed the form of the question in part, so the size of the increase cannot be compared directly.
SME Agency survey: the receiving side's cost pass-through rate was 54.2%
"The price pass-through rate was 54.2%, up 0.7 points from the previous survey (previous 53.5%)."
Efforts toward passing on prices and fair trade (August 2026, SME Agency, document 3) › Summary of results of the March 2026 "Price Negotiation Promotion Month" follow-up survey › p.7 · source
AI: The answers come from 69,625 small and medium-sized companies on the receiving side. It is a figure about terms of trade, not about harassment.
Customer harassment prevention measures become an employer obligation
"With this amendment, measures to prevent customer harassment and sexual harassment of job seekers and others become an obligation for employers. (Effective 1 October 2026 (Reiwa 8))"
[Amendment] Comprehensive Labor Policy Promotion Act... special page (Fukushima Labour Bureau) › About the amended law · source
AI: The MHLW's own page still described the effective date in old wording, so I cited the labour bureau's page.
The note has 7 more facts. They include relationships (11.5%) as the second most common reason women left their previous job in the Survey on Employment Trends, 55,614 consultations about "bullying and harassment", and measures taken against violations of the Freelance Act.
AI inferences (no source states these directly)
The core of in-house problems is the relationship with the boss: in 1 and 3 above and F2 in the note, power harassment is the largest figure in all three surveys.
We did not find any government statistic that counts strained relationships with business partners themselves (things that are hard to say, awkwardness): problems with business partners are counted separately as customer harassment (2, 3) and terms of trade (5, and F9, F10, and F13 in the note).
Stress from interpersonal relationships may have increased: the evidence is the figure in 4. But the form of the question changed, so this is not certain.
Companies' duty to respond now reaches people outside the company: after the amended law on power harassment came fully into force in 2022, the customer harassment obligation took effect in October 2026, and the Act on Proper Trade (formerly the Subcontract Act) in January 2026 (6, and F13 in the note).
Open points
In the workers' compensation figures, the count for "violence, bullying, or harassment by colleagues and others" could not be read because the text in the attached table was garbled.
I have not read the text of the Tokyo ordinance. I confirmed its effective date and that it has no penalties in Tokyo's Q&A.
I checked every quote against the saved page text with a script, and all 13 matched. The output is in Japanese, so no English translations are attached.
Actual token usage: 3.9M (3% less than planned)
Planned: 4M
First report (message 1) · 10:52
Not done: I researched both in-house relationships and business partners, checked 18 facts against the original text, and saved them in the note. Five items, such as private surveys focused on business partners, are not researched, because I reached the 4M-token budget. Please choose whether to continue in the question that follows.
Answer (AI)
Inside companies, verbal attacks by bosses on their subordinates show up most in the statistics, and about 40% of those who receive them do nothing. Problems with business partners are counted in two separate groups: nuisance behavior by people at the partner, and terms of trade such as cutting payments. Laws covering both came in during 2026, but I have not found a survey that measures the change after they took effect.
Facts (1/2)
MHLW Industrial Safety and Health Survey: "interpersonal relationships" are 32.9% of the causes of strong stress
"Among workers who have matters they feel as strong stress, looking at what they are (up to 3 main items), "amount of work" 39.5% [previous 43.2%] and "failures at work, arising of responsibility, etc." 39.5% [previous 36.2%] were the most common, followed by "interpersonal relationships (including sexual harassment and power harassment)" at 32.9% [previous 26.1%]."
Overview of the 2025 (Reiwa 7) "Industrial Safety and Health Survey (Fact-finding Survey)" › (1) Stress at work and in working life · no publication date given · source
AI: About 1 in 3 workers with strong stress (67.5% of all workers) named interpersonal relationships. The values in [ ] are for 2024. The survey does not separate whether the other person is a boss, a colleague, or a customer.
MHLW-commissioned survey: "boss (not an executive)" was the harasser in 65.7% of power harassment
"As for who committed power harassment, "boss (not an executive)" (65.7%) was the most common, followed by "company executive (officer)" (24.7%)."
Survey on Workplace Harassment, report › Harassers · 2024-05-17 · source
AI: A FY2023 survey of 8,000 workers. 19.3% had experienced power harassment in the past three years, and these are their answers. The most common type was psychological attacks such as verbal abuse, at 48.5%. The shares for colleagues and subordinates are not in this sentence.
Same survey: 36.9% "did nothing" after experiencing power harassment
"As for actions after experiencing power harassment or sexual harassment, "did nothing" (36.9% and 51.7% respectively) was the most common, followed for power harassment by "consulted family or friends outside the company" (19.3%), and for sexual harassment by "consulted a colleague at the company" (14.7%)."
Survey on Workplace Harassment, report › Actions after experiencing harassment · 2024-05-17 · source
AI: The most common reason for doing nothing was "I thought nothing I did would solve it" (65.6% for power harassment). What happened afterward is unknown.
MHLW workers' compensation figures: "power harassment by a boss or others" is the most common mental-disorder award, at 222
"By number of benefit decisions, the most common were "received power harassment such as physical or psychological attacks from a boss or others" with 222, "received serious nuisance behavior from customers, business partners, facility users, or others" and "received sexual harassment" with 127 each, and "an event occurred that caused a major change in the content or amount of work" with 113."
Announcing the FY2025 "Status of Workers' Compensation for Karoshi and Related Cases" › p.27, Table 2-8 · 2026-07-15 · source
AI: The top three were all about relationships with people. Nuisance behavior by customers, business partners, and others was second with 127. These are counts of serious recognized cases, not incidence rates. There is no count for business partners alone.
First report (message 2) · 10:52
Facts (2/2)
Same survey: 22.6% of workers named people at a business partner as perpetrators of nuisance behavior
"For serious nuisance behavior by customers and others, "customers and others (including patients or their families)" was the most common at 82.3% (see Figures 149 to 151)."
Survey on Workplace Harassment, report › Harassers › p. 132-133 · 2024-05-17 · source
AI: In the same chart, "employees or executives of other companies such as business partners" is 22.6% (861 people who experienced nuisance behavior, multiple answers). In the answers from 1,880 companies, behavior by business partners toward the company's employees was 10.9%, and behavior by the company's employees toward business partners was 4.3%. I read these three figures from the order of the items and values in the chart.
JFTC: 8,261 guidance cases under the Act on Proper Trade (formerly the Subcontract Act) in FY2025
"By type of violation subject to recommendations, there were 31 cases of requests to provide unjust economic benefits, 6 of reductions in manufacturing contract payments (note), 6 of returns, 1 of unjust changes to the content of delivery, and 1 of beating down prices."
Enforcement of the Act on Proper Trade in FY2025 and efforts toward fair trade for small and medium-sized businesses › (1) Number of recommendations · 2026-06-10 · source
AI: The same page says "There were 39 recommendations in FY2025." and "There were 8,261 guidance cases in FY2025." These are cases where the ordering side put small and medium-sized companies on the receiving side at a disadvantage. They do not count how the staff behave toward each other. From 1 January 2026, setting payments one-sidedly without agreeing to discuss them is also prohibited (JFTC briefing materials).
MHLW: measures against nuisance behavior by customers and business partners are an obligation for all employers from 1 October 2026
"From 1 October 2026 (Reiwa 8), measures against customer harassment and sexual harassment of job seekers and others have become an obligation for all employers!"
For the prevention of harassment at work › top of page · no publication date given · source
AI: The obligation took effect about 7 weeks ago. As of FY2023, 55.8% of companies had "no particular" measures in place. There are no figures from after the law took effect yet.
AI inferences (readings the sources do not state directly)
The core of in-house problems is words from bosses to subordinates. Based on 2, 4, and psychological attacks at 48.5%.
The problems rarely go through workplace procedures. 36.9% do nothing (3). Meanwhile, in a different survey, 94.8% have someone to talk to. This reads as people having someone to talk to without it leading to a solution at work. The two are different surveys, and no source followed the same people.
Statistics count only serious problems that have a name. Awkwardness or sharp remarks that fall short of a definition show up only in the 32.9% in 1, and what they consist of is unknown.
Problems with business partners are counted separately as conduct toward people (5) and terms of trade (6). Both may come from the power gap between the ordering side and the receiving side. I found no source linking the two.
When the other person is outside the company, the company's own systems reach them less easily. People who experienced nuisance behavior responded by "handing it over to a boss" (37.4%) and "kept apologizing" (32.1%). The two 2026 laws are placed to fill this gap.
The same company can be both the side receiving such behavior and the side doing it. Based on 10.9% and 4.3% in 5.
Actual token usage: about 4.4M (10% more than planned)
Planned: 4M
Final report · 11:09
Done (one item not obtained): I researched the 5 extra items, and the note now has 27 facts and 10 AI inferences. Only for the share of freelancers who experienced harassment could I not get a source with figures.
These are the 4 main facts added to the 7 sent earlier.
En Japan survey: the top real reason for leaving that people did not tell the company was "bad relationships"
"★ The top reason for leaving that people told the company: "I want to try a different type of job". The top real reason for leaving that people did not tell the company: "bad relationships"."
"Real reasons for leaving" survey (2024) › Summary of results · 2024-07-31 · source
AI: Answers from 5,168 users of the company's own job site, so they do not represent all workers. Persol Research and Consulting also estimates that of about 865,000 people who left because of harassment, 573,000 did not tell the company (2021 figures).
JFTC and MHLW freelance survey: fee reductions reported by 3.0% of clients and 28.1% of freelancers
"The share answering that they had reduced or received a reduction in fees was 3.0% for clients and 28.1% for freelancers. (See graph)"
Results of the fact-finding survey on freelance transactions (survey of conditions before the law took effect), summary › Fee reductions › p. 14 · 2024-10-18 · source
AI: Answers that fees were set without enough discussion were also 22.2% for clients against 67.1% for freelancers. Clients and freelancers answered separately, so the figures do not compare the same transactions.
UA Zensen survey of union members: 46.8% experienced nuisance behavior by customers, against 56.7% in 2020
"The answer "I have experienced nuisance behavior within the last 2 years" was 46.8% (56.7% in the 2020 survey), a downward trend. *We presume this is the result of public awareness-raising and labor and management efforts over this period"
Press lecture held on the results of the customer harassment questionnaire · 2024-06-05 · source
AI: 33,133 answers from union members in retail and services. In the same survey, 33.7% felt nuisance behavior was "increasing" and 6.7% felt it was "decreasing". The question covers only behavior by customers and does not separate out business partners.
SME Agency survey: price negotiations took place in 90.7% of cases, and the pass-through rate was 54.2%
"The share where "price negotiations took place" rose slightly from the previous survey to 90.7%."
Announcing the results of the follow-up survey on the Price Negotiation Promotion Month (March 2026) · 2026-06-26 · source
AI: The same page says "The price pass-through rate was 54.2%." It is a figure showing better trading relationships, but it does not measure how the staff behave toward each other.
I added 4 AI inferences.
The side with more power may count what it does at a lower rate than the other side does. Based on 2 above, and on the company survey, where behavior by the company toward business partners was only 4.3%.
The real reasons do not reach the company. Three sources point the same way: the MHLW (36.9% did nothing), En Japan, and Persol Research and Consulting.
Apart from the MHLW-commissioned survey, I found no source that counts conduct by people at business partners on its own. UA Zensen, Teikoku Databank, Rengo, and Tokyo Shoko Research either do not separate customers from business partners or ask only about customers.
Figures showing improvement and figures showing decline are mixed. This reads as more chances to talk, with results that have not caught up yet.
Two points about the figures:
The 62 workers' compensation cases for "violence or bullying by colleagues and others" and the 8.9% of men who left because of "workplace relationships" were read by an agent from page images in the PDF. They were not checked with a script.
All other quotes matched the saved original text.
All 27 facts are in japan-workplace-relationship-problems-2026-11-18.md, committed and pushed. The same request is also running in Workplace and business-partner relationships in Japan and Workplace and business-partner relationship problems in Japan.
Actual token usage: about 6.9M (25% more than planned)
Planned: 5.5M (4M at first + 1.5M added)
The research note link at the end of each answer opens that note's download. Links to other chat threads keep only their text.
The answers are AI translations of the Japanese originals, which are the record. The research notes are shown in the original Japanese.
原文の英訳 Asked which workplace relationships caused them stress, the 578 respondents most often cited having nobody nearby to consult (27.7%), followed by difficulty judging what constitutes harassment (26.8%) and a lack of constructive feedback (26.5%).
原文の英訳 Over the preceding three years, 19.3% experienced power harassment at work, 6.3% sexual harassment, and 10.8% serious misconduct by customers or others.
原文の英訳 In the fiscal 2023 survey, 27.8% of those reporting power harassment cited isolation, exclusion or being ignored. Among those reporting sexual harassment, 49.7% cited sexual jokes or teasing.
原文の英訳 Among 861 people reporting serious misconduct by customers or others, 22.6% identified an employee or executive of another company, such as a business partner.
原文の英訳 Of the companies in this question, 10.9% reported an incident involving someone at another company acting toward their employee, and 4.3% reported someone at their company acting toward another company's employee.
原文の英訳 There were 57 responses from business-to-business companies. In the category primarily serving corporate clients (1-2), 16 cited verbal abuse or intimidation and five cited being kept for a long time.
原文の英訳 A sales representative at a business partner cited last year's figures and asked us to report how many seasonal products we would buy. We feel unable to refuse.
原文の英訳 Of the 581 suppliers answering this question, 67.0% said there had been occasions when they felt obliged to agree because of future business dealings.
Translation (English): Rows “2025, men” and “2025, women” / column “Interpersonal relationships at work were unsatisfactory” / values “8.9” and “11.5”.
Translation (English): The percentages who experienced power harassment, sexual harassment or seriously disruptive conduct by customers or others at their place of employment over the past three years were 19.3%, 6.3% and 10.8%, respectively.
AI: 2024年1月に、全国の企業・団体に勤務する20~64歳の労働者8,000人に聞いた。被害の内容では、パワハラ経験者は脅迫・侮辱・ひどい暴言などの精神的な攻撃を多く挙げた。セクハラ経験者は性的な冗談やからかいが最多だった。顧客等から著しい迷惑行為を受けた人は、頻繁なクレームなどの執拗な言動を最も多く挙げた。同資料 p. 22–23。
Translation (English): Consultations on “bullying and harassment” totalled 55,614, up 1.1% from the previous fiscal year and the largest category for the 14th consecutive year.
Translation (English): Representatives of business counterparties / representatives involved in negotiations aimed at concluding a contract between companies.
Translation (English): Costs rose, but the buyer did not initiate talks, and the supplier did not request negotiations for fear of fewer orders or an end to the business relationship.
AI: 2025年10月から2026年3月末までの取引を受注側に聞いた。価格交渉が不要だった回答を除く延べ65,311社分では、この回答が6.9%。同じ集計で90.7%は交渉を実施した。不要だった回答も含む延べ91,233社分では、それぞれ5.0%、64.9%。後者の集計では28.4%が交渉不要だった。同資料 p. 2–4。