Polimill scales QommonsAI across Japanese municipalities
Polimill says its QommonsAI generative AI platform is now used by about 1,050 municipalities and roughly 550,000 public employees across Japan. Released in October 2024 and built with OpenAI technology, the platform provides specialised support for assembly responses, public services, social welfare and legal search.
The startup began with Surfvote, a forum for citizens to exchange views on politics and society. Its work with local governments highlighted a constraint: public-sector teams were heavily occupied with daily operations, leaving limited time to incorporate citizens’ views into policy. Polimill’s response was to make government workflows more efficient while building a shared foundation for municipal work.
Turning fragmented records into shared knowledge
Municipal data is often fragmented across local workflows, document formats and historical records. Polimill cites assembly-response preparation as one example: officials may need to review years of minutes to ensure an answer remains consistent with prefectural policy. The company collected and standardised assembly minutes from across Japan, then used AI to add metadata and create a search foundation spanning municipalities and time periods.
Polimill extended that structure to welfare, laws and other administrative domains. Bringing specialised data and AI into a common, searchable interface is intended to make distributed administrative information usable in day-to-day work. The company describes this as the basis for QommonsAI to develop from a productivity tool into a public operating system for municipal work.
Security controls alongside familiar models
OpenAI’s GPT models are central to QommonsAI. For public-sector deployments, the platform includes controls allowing administrators to review feature-use history and limit available models in line with organisational policy. Masahiro Wakabayashi, Polimill’s CAIO, also points to the broad capability of GPT models and public familiarity with ChatGPT as factors that can reduce the initial barrier for employees adopting a new system.
Polimill says GPT models are the most frequently selected option in QommonsAI’s general conversation feature. The company uses them for tasks ranging from file reading to routine dialogue, seeking to provide one model that can respond to a wide range of work requests.
Faster development and plans for Qommons ONE
Polimill adopted Codex throughout its development workflow, including requirements definition, checks against existing GitHub code, implementation and testing. Engineers review AI-generated plans and make higher-level decisions while AI performs more implementation work. The company reports that Codex and OpenAI’s hands-on support increased development speed by three to five times, accelerating prototype validation and feedback cycles with municipal teams.
In fall 2026, Polimill plans a full rollout of Qommons ONE, a store through which external companies can offer applications to municipalities. Its proposed super agent would combine specialised AI systems and private-sector apps to produce deliverables such as research and presentations from a stated goal. For organisations considering AI in regulated workflows, Polimill’s approach underlines the practical need to combine standardised information, auditable controls, iterative validation and accountable human review.