TL;DR
Polimill says QommonsAI now serves about 1,050 Japanese municipalities and 550,000 public employees through specialized administrative tools built with OpenAI technology. The company plans to expand the platform with an application store and coordinating AI agent in fall 2026, although its adoption and performance figures have not been independently verified.
Polimill says its OpenAI-powered QommonsAI platform is now used by about 1,050 Japanese municipalities and 550,000 public employees, as the startup prepares to expand the service into a shared system connecting administrative AI tools and outside applications. The planned expansion could affect how local governments search records, prepare policy materials and preserve experienced employees’ working knowledge.
Released in October 2024, QommonsAI provides specialized tools for assembly responses, public services, social welfare and legal research. Polimill collected and standardized assembly minutes from across Japan, added AI-generated metadata and created a search layer designed to work across municipalities and time periods. The company later extended that structure to welfare, laws and other administrative information.
OpenAI GPT models power core parts of the platform. QommonsAI also gives administrators access to usage histories and controls for restricting available models under organizational policies. Polimill says GPT models are the most frequently selected option in the platform’s general conversation feature, though it did not publish usage shares or an independent comparison with other models.
Polimill also adopted Codex for requirements work, code consistency checks, implementation and testing. The company reports that Codex and OpenAI support raised development speed to three to five times its previous level. OpenAI did not provide the measurement period, underlying data or methodology behind that comparison.
Polimill Builds Japan’s Next-generation Public AI Infrastructure
QommonsAI brings municipal records, specialized administrative tools and OpenAI technology into one shared environment. Polimill’s next ambition is larger: a coordinating layer for public-sector applications, knowledge and workflows.
Japanese municipalities, according to Polimill.
Public employees covered by the platform’s stated reach.
Target window for the broader Qommons ONE rollout.
A shared layer for municipal work
QommonsAI is designed around recurring administrative tasks. Polimill standardized records from across Japan, added AI-generated metadata and built search tools that can work across jurisdictions and time periods.
Response preparation
Searches assembly minutes and supporting material to help officials research issues and prepare informed responses.
Public-service support
Organizes information used in everyday municipal workflows, reducing repeated searches across fragmented sources.
Welfare research
Extends structured search and generative assistance into complex welfare rules, records and administrative guidance.
Legal research
Helps staff navigate laws and related administrative information while keeping human review central to decisions.
Knowledge preservation
Aims to retain how veteran officials investigate questions and revise drafts before experienced employees leave.
Model controls
Provides usage histories and organizational controls that can restrict which AI models employees are allowed to use.
From scattered records to reusable knowledge
The strategic value is not only faster drafting. A common information layer could help municipalities with different staffing levels, budgets and technical capacity reuse knowledge across organizational boundaries.
Collect
Assembly minutes, laws, welfare records and administrative sources.
Standardize
Different formats are organized into a more consistent structure.
Enrich
AI-generated metadata makes material easier to discover and connect.
Search
Officials retrieve relevant information across places and time periods.
Apply
Evidence supports research, drafting, review and public-service work.
AI output depends on record quality, local context and careful review. Preserving expertise is valuable only when officials can inspect, challenge and correct the system’s conclusions.
Scale is promising; proof is incomplete
Polimill’s announcement describes significant reach and faster development, but does not disclose the definitions, measurement periods or independent evaluations needed to judge real-world performance.
| Claim | Reported result | Missing evidence | Status |
|---|---|---|---|
| Municipal adoption | About 1,050 municipalities | Active-use definition, frequency and independent verification | ~ Unclear |
| Employee reach | About 550,000 public employees | Registered, eligible or recurring-user breakdown | ~ Unclear |
| Development speed | Three to five times faster | Baseline, measurement period, data and methodology | ✗ Unverified |
| Policy-drafting support | Novices approached veteran evaluations | Sample size, scoring method, evaluators and full results | ✗ Unverified |
| Administrative functionality | Multiple specialized tools described | Independent outcome and error-rate data | ✓ Documented scope |
Qommons ONE aims to become a “public OS”
The phrase describes Polimill’s stated ambition, not an established government designation. The planned system would connect specialized AI tools and outside applications through a central coordinating agent.
The broader launch is expected to test whether a municipal productivity platform can evolve into shared public-sector infrastructure.
Application store
Outside companies would be able to offer purpose-built municipal applications through the platform.
Coordinating AI agent
A central agent would select specialized AI systems or applications for tasks from research to presentations.
Shared operational layer
Tools, institutional knowledge and third-party services would become accessible through a common environment.
What to watch next
Practical value needs auditable evidence
The opportunity: organizing fragmented records and making institutional knowledge reusable could reduce duplicated work and narrow capability gaps between municipalities.
The caution: vendor-reported reach and speed do not establish service quality, safe operation or public benefit. Stronger confidence requires active-use definitions, error data, evaluation methods, procurement transparency and security documentation.
The balanced view: municipalities facing labor shortages have good reasons to test useful tools before evidence is perfect—but those experiments should produce results that citizens, officials and independent reviewers can audit.
Japan’s municipalities face uneven staffing, fragmented records and different document formats. A common platform could reduce repeated work by helping officials search information across jurisdictions and reuse administrative knowledge. It could also narrow gaps between governments that have different budgets and technical capacity, if access and performance remain consistent.
The stakes extend beyond employee productivity. Polimill wants QommonsAI to capture how veteran officials research issues and revise drafts, preserving tacit administrative knowledge that may disappear when experienced workers leave. That objective matters because AI output still depends on the quality of records, local judgment and human review.
AI-powered municipal record search software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Civic Platform to Public OS
Polimill began with Surfvote, a platform for public discussion about politics and society. Its work with local governments led the company to focus on administrative workloads, arguing that officials consumed by routine duties have less time to incorporate residents’ views into policy.
QommonsAI was built as Polimill’s response to that problem. The company now describes the service as moving toward a “public OS” for municipal work. That remains Polimill’s stated ambition, rather than an established government designation or independently evaluated national infrastructure program.
public sector AI administrative tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Adoption and Performance Need Scrutiny
The announcement does not explain how Polimill counted the reported 1,050 municipalities or 550,000 public employees. It is unclear whether those figures represent active recurring users, eligible employees, registered accounts or organizations with some level of access.
Polimill says less-experienced employees produced AI-assisted policy proposals that received evaluations close to those written by veteran officials. No sample size, scoring method, evaluator information or full results were disclosed. Details about error rates, privacy safeguards, procurement terms and responses to inaccurate model output also remain unavailable in the announcement.
OpenAI GPT model for government use
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Qommons ONE Faces Its Rollout
Polimill plans a full rollout of Qommons ONE in fall 2026. The service is intended to include a store where outside companies can offer municipal applications and a coordinating agent that selects specialized AI systems or apps for tasks ranging from research to presentations. The next evidence to watch will be rollout timing, participating vendors, security documentation and independently measured outcomes from municipal deployments.
AI legal research tools for government
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Where I land
This is my interpretation: I see real public value in organizing fragmented municipal records and making institutional knowledge easier to reuse. The strongest part of Polimill’s case is the practical focus on administrative work. Still, vendor-reported reach and speed do not establish service quality, public benefit or safe operation.
The strongest counterargument is that municipalities facing labor shortages cannot wait for perfect evidence before testing useful tools. I accept that point, but experimentation should produce auditable results. I would view the project more favorably if Polimill released active-use definitions, evaluation methods, error data and security documentation, along with independent evidence that QommonsAI improves outcomes without weakening accountability. Evidence of persistent errors, poor adoption or inadequate oversight would move my assessment in the opposite direction.
Source: OpenAI
Key Questions
What is QommonsAI?
QommonsAI is Polimill’s generative AI platform for Japanese public-sector work. It supports tasks including record searches, assembly responses, welfare research and legal research.
How widely is the platform used?
Polimill reports use across about 1,050 municipalities and access involving about 550,000 public employees. The announcement does not define active use or provide independent verification.
What role does OpenAI play?
GPT models provide core AI capabilities, while Codex is used in Polimill’s software-development workflow. OpenAI also provides technical guidance, validation support and development examples.
Will AI replace municipal employees?
Polimill presents QommonsAI as a tool for supporting staff and preserving expertise, not replacing experienced officials. Its own evaluation found that veteran employees’ proposals still received higher ratings.
What is Qommons ONE?
Qommons ONE is the planned next stage of the platform. Polimill says it will combine specialized AI tools and private-sector applications through a central coordinating agent.
Source: OpenAI