DeepMind extends Gemini-built PlanAssist to 12 more English councils
Google DeepMind has announced that PlanAssist, an AI planning tool built with Gemini, will be made available to an additional 12 local planning authorities across England. The lead of the post says the tool has shown it can cut the time planning officers spend assessing householder applications by up to 58%.
The problem it targets is routine work. According to the post, householder applications (the examples given are a loft conversion, a conservatory or an extension) account for up to 70% of planning officers' workloads. They are often simple to approve but take a disproportionate share of officers' time. DeepMind ties the delays to the UK government's goal of building 1.5 million new homes by 2029, and says families can wait months for approval of something as simple as a loft conversion.
PlanAssist was developed by Google DeepMind in collaboration with the UK government, Google Cloud and Faculty. It is part of DeepMind's broader partnership with the UK government to explore how AI can help transform planning and other public services. The tool works in four steps. First, it validates application documents and architectural drawings, checking that the required information has been submitted so the application can proceed. Next, it checks the application against heritage constraints (such as tree preservation orders) and local and national planning policies. It also checks whether any issues or objections have previously been raised, for example by local residents. Finally, it drafts fully cited reports that planning officers can edit before they reach a final decision.
Since June, the tool has been piloted in Barnet, Camden and Dorset councils. In early pilot testing it brought the average time an officer spent on householder assessment tasks down by 58%, which the post describes as more than halving the assessment time per application. In addition, more than 70% of the reports the tool generated were approved by officers with no edits required. Those reports cover things like whether an application complies with local planning rules or whether the applicant has submitted all required documentation. DeepMind calls the early results encouraging.
PlanAssist is now moving into a 12-month Beta phase using live applications, starting with a subset of councils. Twelve more councils will be onboarded in three cohorts, bringing the total involved to 15. The post says these include metropolitan and rural districts, so the tool can be tested across a diverse range of communities. Subject to successful testing, the government plans for the tool to be made available to all councils in England in 2027, with every assessment reviewed and approved by a qualified planning officer before any decision is made.
Throughout, the post stresses that the tool is designed to support human judgement and that planning officers retain the final say on every application.
Key facts
- PlanAssist, built with Gemini, is being made available to an additional 12 local planning authorities across England, taking the total involved to 15.
- In early pilot testing in Barnet, Camden and Dorset, it cut the average time an officer spent on householder assessment tasks by 58%; the post's lead says "up to 58%".
- More than 70% of the reports the tool generated in the pilot were approved by officers with no edits required.
- The next stage is a 12-month Beta using live applications, with the twelve new councils joining in three cohorts.
- Subject to successful testing, the government plans to offer the tool to all councils in England in 2027; officers keep the final say on every application.
Why it matters
Routine householder applications make up as much as 70% of planning officers' workloads, according to the post, and DeepMind links the resulting delays to the UK government's goal of 1.5 million new homes by 2029. This is a move from pilot to live use: the tool goes from three councils to 15, working on live applications for 12 months. It is also a concrete example of DeepMind's wider partnership with the UK government on AI in public services.
Who it affects
Planning officers at the councils involved, who get automated help with processing applications and drafting assessment recommendations. Applicants, such as families waiting on a loft conversion, conservatory or extension, are the ones who would feel any speed-up. The 12 new councils are described as a mix of metropolitan and rural districts. Their names and the dates of the three cohorts are not given.
How to use it
There is no public product to try. Access is through the government's rollout: the 12 additional councils are being onboarded in three cohorts during the 12-month Beta, and the government plans to offer the tool to all councils in England in 2027, subject to successful testing. Officers use it for four jobs: validating documents and drawings, checking heritage constraints and local and national policies, checking earlier issues or objections, and drafting cited reports they can edit.
How solid is it
The figures come from Google DeepMind's own blog post and describe early pilot testing. The 58% figure is for the average time on householder assessment tasks; the lead says "up to 58%". The source does not give absolute hours or minutes, sample size, or number of applications. The more-than-70% figure refers to reports approved without edits; the source does not say how many reports or what share of the whole decision was automated. No independent evaluation of the results is cited.
Risks and caveats
The 2027 national availability is a government plan subject to successful testing, not a commitment. The post says officers retain the final say and that every assessment is reviewed and approved by a qualified planning officer; the source does not say whether the tool makes or recommends approval decisions itself beyond drafting assessment recommendations and reports. No cost, pricing or funding amount is given. The Beta will show whether the pilot results hold across a wider and more varied set of councils.