AIToday
Large Language ModelsAI Business & IndustryITmedia AI+Published: Oct 27, 2026, 10:01 JST

Kyoto Prefecture finds Copilot Chat RAG insufficient for 7,000 staff

Send on LINE
Kyoto Prefecture finds Copilot Chat RAG insufficient for 7,000 staff

3 Key Points

  1. What happened

    Kyoto Prefecture rolled out Copilot Chat across all departments in January 2025 for about 7,000 staff, but the tool's RAG function was not sufficient to load large volumes of internal manuals.

  2. Why it matters

    Staff still have to page through the common 'tebiki' manual one page at a time to check accounting and document rules, and queries that miss the manual go to the responsible division, which is buried in emails and calls.

WHO IT HITSGovernment back-office and administrative staff who must verify procedures against internal rulebooks stand to benefit most, along with the policy divisions that field repeated questions by email and phone.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Kyoto Prefecture's population fell from about 2.65 million in 2005 to about 2.5 million in 2025, and with fewer working-age residents expected, the government needed to handle growing administrative demand with limited staff. The heaviest burden was checking whether accounting and document work matched internal rules, which meant paging through a common manual and, when the answer was not found, waiting on the division that owns the rules. That division was itself swamped by daily emails and phone calls. The prefecture first turned to RPA, but found it effective only for routine work and too demanding in specialist knowledge for all staff to use daily. It then looked to generative AI, with Takashi Ito of the Digital Policy Promotion Division describing how RAG could let staff ask questions in natural language and pull up the needed information and source materials. The January 2025 Copilot Chat rollout was the first step, but the tool's RAG capability at the time could not handle the large volume of manuals, leaving the prefecture to look for a way to make up for that shortfall.

FAQ
When did Kyoto Prefecture introduce Copilot Chat?
It introduced the conversational AI Copilot Chat across the entire prefecture in January 2025.
Why wasn't Copilot Chat enough on its own?
At the time, its RAG (retrieval-augmented generation) function was not sufficient, making it hard to load and use enormous materials such as the internal manuals.
Why didn't RPA solve the problem?
RPA is effective for routine tasks, but building and operating it requires specialist knowledge, which was a hurdle for everyday use by all staff.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleAnthropic to pay Microsoft at least $44.9 billion through 2033