From 3 Councils to 15: DeepMind PlanAssist Enters a 12-Month Beta After Cutting UK Planning Assessment Time by 58 Percent

PlanAssist, the AI tool that prepares routine planning cases for council officers in England, is moving from a three-council pilot into a 12-month beta. The Ministry of Housing, Communities and Local Government announced on October 30 that the tool, developed with Google DeepMind, Google Cloud and the consultancy Faculty, will roll out to 12 more local planning authorities, bringing the total to 15, with the aim of offering it to every council in England by the end of 2027.

The tool targets the most common and least contentious part of a planner's job: householder applications such as extensions, loft conversions and conservatories, which the government says account for nearly 70 percent of applications each year. Built on Gemini, PlanAssist first checks that submitted documents and architectural drawings are complete. It then checks the proposal against heritage designations, tree preservation orders, local and national planning policy and any objections already on file, and finally drafts a fully cited report that the officer can edit before deciding.

The numbers come from the June pilot in Barnet, Camden and Dorset. DeepMind says the average time an officer spent on householder assessment tasks fell by 58 percent, and that more than 70 percent of the reports the tool generated were approved by officers with no edits. The government frames the same data as helping to halve processing times for routine householder applications from eight weeks to four. All of these are government and DeepMind figures measured on three councils; no independent evaluation has been published.

The 12 new councils span metropolitan and rural England: Ashford, Lambeth, Newcastle, Nottingham, Bristol, Cumberland, North Lincolnshire, Rochdale, Redbridge, South Gloucestershire, Stoke-on-Trent and Westminster. They will join in three cohorts, working on live applications rather than test cases, a mix DeepMind says is meant to show how the tool handles different kinds of communities.

The safeguards are conventional for public-sector AI: officers keep the final decision on every case, councils are not required to use the tool, and DeepMind says every assessment will be reviewed and approved by a qualified planning officer. That mirrors the line UK regulators have taken, that existing law already covers AI agents and that responsibility stays with people.

The expansion arrives inside a broader planning overhaul. From October 31, a new National Scheme of Delegation means straightforward applications are decided by planning officers rather than waiting for a committee slot. The government says 420 planners have been recruited through its programs, beating a manifesto pledge of 300, and it has committed to requiring key planning documents to be digitised and published as open data by the end of 2027. The backdrop is a target of 1.5 million new homes by 2029, which planning delays currently threaten.

The open question is whether the pilot numbers survive contact with twelve more councils. The 58 percent covers assessment tasks, not the time from submission to decision, and the halving of processing time is something the tool is helping with alongside new staff and new delegation rules, making the attribution hard. And there is a subtler risk the beta will have to measure: when more than 70 percent of drafts are accepted unedited, the system depends on officers staying critical of documents they have learned to trust.