{"id":3435,"date":"2026-10-02T13:53:39","date_gmt":"2026-10-02T04:53:39","guid":{"rendered":"https:\/\/forum-corp.co.jp\/eng\/?p=3435"},"modified":"2026-10-02T13:53:39","modified_gmt":"2026-10-02T04:53:39","slug":"article-23","status":"publish","type":"post","link":"https:\/\/forum-corp.co.jp\/eng\/article-23\/","title":{"rendered":"ChatGPT is Already in Your Office. Do You Have a Plan?"},"content":{"rendered":"\n
\n

It’s 5:40 on a Thursday. A sales coordinator at a mid-sized trading company in Nagoya has a complaint from an overseas client in the inbox, and the manager who usually checks English replies has gone home. So, the email goes into a free chatbot, along with the client’s name, the order number and a short instruction: make this polite, apologize, don’t admit fault.<\/p>\n\n\n\n

The reply is good, better than the coordinator would have managed alone. It goes out at 5:52.<\/p>\n\n\n\n

Nobody told them they could do that. Nobody told them they couldn’t, either. Whether the 2019 security rule about external cloud services covers a chatbot is anyone’s guess.<\/p>\n\n\n\n

This isn’t unusual anymore. According to the Ministry of Internal Affairs and Communications, personal use of generative AI in Japan more than doubled in a single year, to almost 60%. Most companies say they use AI somewhere. Far fewer can say what their people are actually typing into it.<\/p>\n\n\n\n

That gap ends up with HR. IT can block websites and legal can write rules, but who uses what, and whether it helps or hurts at evaluation time, are people questions.<\/p>\n<\/div>\n\n\n\n

\n

The<\/strong> Rule<\/strong> Nobody<\/strong> Follows<\/strong><\/em><\/h4>\n\n\n\n
\n
\n
\"\"<\/figure>\n<\/div>\n\n\n\n
\n

Surveys suggest only around one in ten Japanese workers use AI for work on most days, even where it has been rolled out. Tightly controlling how workers use AI, produces little advantage.<\/p>\n\n\n\n

The more telling finding: most respondents felt uneasy about putting confidential material into AI, and most of those uneasy people had done it anyway.<\/p>\n<\/div>\n<\/div>\n\n\n\n

So, awareness isn’t the issue. The deadline is real, the risk feels abstract, and nothing offers a safer way to get the same result. A rule that says don’t, with no approved alternative, mostly teaches people to stop mentioning it.<\/p>\n\n\n\n

There’s another reason people keep quiet. Revealing that a two-hour task now takes twenty minutes rarely goes well for them. More work, less credit, or a telling-off for a rule nobody explained. And many evaluation systems in Japan still reward visible effort, in practice if not on paper. The person who finishes the proposal by lunch with a chatbot’s help has a harder story to tell than the one who stayed late.<\/p>\n\n\n\n

For the company, that silence means the good discoveries stay in a few browser tabs and the risky habits stay invisible. Shadow AI is mostly a disclosure problem, and disclosure is shaped far more by HR systems than by IT controls.<\/p>\n<\/div>\n\n\n\n

\n

Handing<\/strong> Out<\/strong> Licenses isn’t a<\/strong> Plan<\/strong><\/em><\/h4>\n\n\n\n

The opposite mistake is just as common. Buy an AI enterprise package, send everyone a login and a thirty-minute video, call it adoption.<\/p>\n\n\n\n

Surveys suggest only around one in ten Japanese workers use AI for work on most days, even where it has been rolled out. Tightly controlling how workers use AI, produces little advantage.<\/p>\n\n\n\n

The companies getting results let people experiment and give them a way to share what works. The staff who benefit most aren’t the most technical. They’re the most curious.<\/p>\n\n\n\n

Panasonic Connect shows what that looks like at scale. Its in-house assistant saved close to 800,000 working hours last fiscal year, a little over 3% of the company’s total. The savings came from ordinary work like meeting minutes, code and first drafts of proposals. Nothing futuristic.<\/p>\n<\/div>\n\n\n\n

\n

What the Law Does and Doesn’t Require<\/strong><\/em><\/h4>\n\n\n\n
\n
\n

No law in Japan requires a company to have an AI policy. The AI Promotion Act of 2025 encourages AI use and carries no penalties for employers, and the government’s AI Guidelines for Business are voluntary.<\/p>\n<\/div>\n\n\n\n

\n
\"\"<\/figure>\n<\/div>\n<\/div>\n\n\n\n

The real exposure is personal data. The Personal Information Protection Commission has warned that putting personal data into an AI tool that uses inputs for training may breach the privacy law. Free consumer versions often do train on inputs. Enterprise versions usually don’t. That one distinction should shape much of the policy.<\/p>\n<\/div>\n\n\n\n

\n

Five Decisions HR Has to Make<\/strong><\/em><\/h4>\n\n\n\n

Approved tools.<\/strong>
Offer an approved option before restricting anything. In international surveys, over half of workers say they’d use AI tools anyway if their company didn’t provide them. A ban with no alternative just moves usage onto personal phones.<\/p>\n\n\n\n

What can go in.<\/strong>
This is the rule that matters most. Keep it to tiers people can remember:<\/p>\n\n\n\n

    \n
  • Public information: any approved tool.<\/li>\n\n\n\n
  • Internal information, such as drafts or meeting notes without names: approved tools only.<\/li>\n\n\n\n
  • Confidential information, such as client contracts: approved enterprise tools, for listed uses only.<\/li>\n\n\n\n
  • Personal data about employees, customers or candidates: not at all, unless specifically approved.<\/li>\n<\/ul>\n\n\n\n

    Then give real examples, including awkward ones. Translating a customer’s email feels harmless, but it contains their name and complaint.<\/p>\n\n\n\n

    Who<\/strong> is responsible for<\/strong> the output.<\/strong>
    Whoever uses it. Write this down, because blaming the tool is tempting when something goes wrong in front of a client.<\/p>\n\n\n\n

    AI in HR’s own decisions.<\/strong>
    AI can draft, sort and summarize CVs and evaluation comments. A person makes and owns any decision about someone’s employment.<\/p>\n\n\n\n

    What happens to the time saved.<\/strong>
    Almost no policy addresses this, and it’s what employees care about most. If AI makes someone faster, do their targets go up? The recommendation is no, not at first, and say so openly. If a faster method just earns more work, people go back to hiding it.<\/p>\n\n\n\n

    Sample Policy Clauses<\/strong>
    A starting point only. Adapt<\/em> to<\/em> your compliance<\/em> policies.<\/em> 1.<\/strong> Use only the AI tools on the approved list for work involving company information.
    2.<\/strong> Do not enter personal data about employees, customers or candidates into any AI tool unless that use has been approved in writing.
    3.<\/strong> You are responsible for anything you produce with AI. Check facts, figures, names and tone before sending.
    4.<\/strong> AI must not be the sole basis for decisions on hiring, evaluation, pay or discipline.
    5.<\/strong> If information goes into a tool by mistake, report it to [contact] the same day. Early reporting is the right call and will be treated that way.
    Clause 5 matters more than it looks. If reporting a mistake feels dangerous, mistakes stop getting reported.<\/em><\/td><\/tr><\/tbody><\/table><\/figure>\n<\/div>\n\n\n\n
    \n

    Training That isn’t a Tool Demo<\/strong><\/strong><\/em><\/h4>\n\n\n\n
    \n
    \n
    \"\"<\/figure>\n<\/div>\n\n\n\n
    \n

    Most AI training so far has been a tour of features. That’s useful for about a week. The harder skill is judgment.<\/p>\n\n\n\n

    In one well-known experiment, consultants given AI did faster, better work on tasks that suited it. On a task just outside what the AI handled well, they were noticeably more likely to get it wrong than colleagues working without it. The answer sounded confident, so they accepted it.<\/p>\n<\/div>\n<\/div>\n\n\n\n

    That’s what training should teach: where AI fails in your own work. In Japanese offices, translation belongs near the top of the list. AI translation is fluent enough that mistakes are hard to spot, and the problems are about meaning. A cautious \u691c\u8a0e\u3057\u307e\u3059 (we\u2019ll consider it) can come out in English as a firm commitment. A polite refusal can read as a maybe.<\/p>\n\n\n\n

    Managers need specific training. Their job is to make AI use visible, which means asking how someone got a task done without it sounding like an accusation. One simple habit: in team meetings, ask everyone to share one thing they used AI for that week, and go first.<\/p>\n\n\n\n

    Skip the thirty-minute training video. Start with one willing department, use its real tasks, fix what doesn’t work, then move on. Twenty people trained well will spread more good practice than two thousand who clicked through a video.<\/p>\n<\/div>\n\n\n\n

    \n

    Where to Start<\/strong><\/em><\/h4>\n\n\n\n

    If only one thing happens, find out what’s actually going on. Run a short anonymous survey asking which tools people use, for what, and what information has gone into them. Make it clear the goal is a sensible policy, not catching anyone. A policy written blind tends to ban what’s already widespread, which is the fastest way to make a rule meaningless.<\/p>\n\n\n\n

    The coordinator in Nagoya didn’t do anything malicious. The client was satisfied. The problem is that the company has no idea it happened, or how often it happens, and no way to learn from the fact that it worked.<\/p>\n\n\n\n

    A good AI policy is one that coordinator would be comfortable mentioning to their manager the next morning. While using AI still feels like a confession, the policy isn’t doing its job.<\/p>\n<\/div>\n\n\n\n

    \"\"<\/figure><\/div>\n","protected":false},"excerpt":{"rendered":"

    It’s 5:40 on a Thursday. A sales coordinator at a mid-sized trading company in Nagoya has a complaint fr […]<\/p>\n","protected":false},"author":6,"featured_media":3436,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_uag_custom_page_level_css":"","iawp_total_views":2,"footnotes":""},"categories":[17],"tags":[],"class_list":["post-3435","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-article"],"uagb_featured_image_src":{"full":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero.jpg",1600,1100,false],"thumbnail":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero-150x150.jpg",150,150,true],"medium":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero-300x206.jpg",300,206,true],"medium_large":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero-768x528.jpg",768,528,true],"large":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero-1024x704.jpg",1024,704,true],"1536x1536":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero-1536x1056.jpg",1536,1056,true],"2048x2048":["https:\/\/forum-corp.co.jp\/eng\/wp-content\/uploads\/2026\/10\/22_Hero.jpg",1600,1100,false]},"uagb_author_info":{"display_name":"Forum Staff","author_link":"https:\/\/forum-corp.co.jp\/eng\/author\/forum\/"},"uagb_comment_info":0,"uagb_excerpt":"It’s 5:40 on a Thursday. A sales coordinator at a…","_links":{"self":[{"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/posts\/3435","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/comments?post=3435"}],"version-history":[{"count":2,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/posts\/3435\/revisions"}],"predecessor-version":[{"id":3442,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/posts\/3435\/revisions\/3442"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/media\/3436"}],"wp:attachment":[{"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/media?parent=3435"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/categories?post=3435"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forum-corp.co.jp\/eng\/wp-json\/wp\/v2\/tags?post=3435"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}