Polimill builds Japan's next-generation public AI infrastructure

OpenAI customer Polimill says about 1,050 municipalities and 550,000 public employees use QommonsAI for assembly responses, public services, welfare, and legal search. It standardized assembly minutes and added metadata so records can be searched across organizations and time periods.

Polimill says Codex and OpenAI support cut development time three to five times. QommonsAI ONE, planned for fall 2026, would combine specialized AI and outside applications.

This is an OpenAI customer case study, so adoption and speed claims are self-reported. Its architecture identifies the public-sector bottleneck: fragmented records, inconsistent workflows, and expert knowledge absent from databases.

Signal Desk's read: the public-OS thesis is plausible only if governance grows alongside retrieval. Common search can reduce duplication, but bad metadata and stale policy then have a wider blast radius.

Polimill’s validation suggests augmentation and knowledge transfer, not replacement: less-experienced staff approached veteran work, while veterans still scored highest.

QommonsAI began after Polimill’s civic platform Surfvote exposed how daily operations crowded out citizen input. It launched in October 2024. Polimill links a shared foundation to Japan’s labor shortage.