Google DeepMind's PlanAssist expands to 12 more English planning authorities after 58% time cut

Google DeepMind is scaling a public-sector deployment. PlanAssist, built on Gemini, helps UK government planning officers assess building and renovation applications. After pilots in Barnet, Camden and Dorset, it is being made available to 12 more local planning authorities across England.

The headline result is a reduction of up to 58% in the time officers spend on routine householder applications, the high-volume cases such as extensions and loft conversions that clog planning queues. The tool assists with assessment rather than making decisions: officers retain judgment while Gemini handles document review and checks against policy.

This is a concrete example of the deployment story that frontier labs increasingly need. While Gemini 4 Carbon took benchmark headlines, PlanAssist shows a measurable productivity gain in a regulated, accountable setting. Planning delays are a political issue in the UK, so a tool that speeds housing approvals has obvious appeal to government.

Caveats: 'up to 58%' is a ceiling, routine applications are the easiest cases, and public-sector AI carries transparency and appeal concerns. Watch whether the rollout reaches complex applications, how decisions are audited, and whether results hold at the wider set of councils.