I’m optimistic about saving time and reducing risk. Previously, Pol.is‑style tools hit limits at scale because meaningful semantic aggregation was hard. With language models, summarization and topicalization are grounded enough that hallucination is negligible for these tasks—true only this year. Small pilots, like Bowling Green, Kentucky, already show good results. Jigsaw’s open‑source Sensemaker (currently relying on Gemini 2.5) works; we’re porting it to GPT‑OSS for more openness.