DeepMind’s PlanAssist cut UK planning-officer assessment time by up to 58% in pilots
Watch
Primary
On October 31, 2026, Google DeepMind said PlanAssist, built with the UK government, Google Cloud, and Faculty, cut the time planning officers spent assessing householder applications by up to 58% in early pilots at Barnet, Camden, and Dorset. A 12-month live Beta starts with more councils; every assessment still needs a qualified officer’s approval.
At a glance
Pilot figures are from Google DeepMind’s October 31, 2026 post.
Officer assessment time
Up to 58% lower in early pilots (DeepMind)
Householder share of workload
Up to 70% (DeepMind)
Pilot councils
Barnet, Camden, Dorset
Next door
12-month live Beta; 12 more councils (15 total)
Human in the loop
Qualified planning officer approves every assessment
How to read this claim
Measurement — PlanAssist cut officer assessment time up to 58% in pilots — DeepMind’s figure; a qualified officer still approves every decision.
Bill — List input and output per million tokens on the vendor page — a loop pays both again.
Door — PlanAssist is not described as a public developer API.
DeepMind published pilot results for PlanAssist and a path to a wider UK council Beta, with a concrete 58% assessment-time claim.
What it introduces
PlanAssist, an AI drafting and validation aid for UK householder planning applications, moving from three-council pilots into a wider Beta.
Who benefits Anyone quoting DeepMind applied-AI results into a product plan, and anyone who assumed a planning copilot replaces the officer’s sign-off.
Try this
Treat the 58% as DeepMind’s pilot claim on householder assessment time. Do not equate it with a public model rank. If a slide says PlanAssist is live nationwide, the post still says Beta and a 2027 goal for England.
What we do not know
The post does not publish the raw timing study or the officer survey instrument behind the 58% and 70% figures.
PlanAssist is not described as a public developer API.
Nationwide English rollout is a 2027 plan subject to successful testing, not a ship date on the post.
Evidence
PrimaryBlog. Figures are from Google DeepMind's October 31, 2026 PlanAssist post about UK local-authority pilots with Google Cloud and Faculty.
Independent The 58% and the officer survey figures are DeepMind’s. No independent council evaluation is linked on the post.
Google DeepMind published Speeding up planning decisions with AI on October 31, 2026. PlanAssist validates application documents and drawings, checks heritage and other constraints, and drafts assessment text so officers can focus on judgment. DeepMind says early pilot testing across Barnet, Camden, and Dorset cut average officer time on householder assessment tasks by up to 58%, and that more than 70% of officers said the tool improved their work. Routine householder applications account for up to 70% of planning officers’ workloads on DeepMind’s account. The tool moves into a 12-month Beta on live applications; twelve more councils join in three cohorts for 15 total. Subject to testing, the UK government plans availability to all English councils in 2027, with every assessment reviewed and approved by a qualified planning officer.
Why it matters
DeepMind is publishing a timed, named pilot result on a government workflow, not a leaderboard score. The so-what for a newcomer is that the 58% is DeepMind’s pilot figure for assessment time, and the door is a council Beta — not a model you can call.
DeepMind frames the UK government’s target of 1.5 million new homes by 2029 as context for planning bottlenecks; that goal is not a PlanAssist score.
The post credits collaboration with the UK government, Google Cloud, and Faculty.