Speeding up planning decisions with AI

Planning backlogs touch ordinary people. A loft conversion or an extension can sit for months waiting on an officer's assessment. PlanAssist, built with Gemini by Google DeepMind with the UK government, Google Cloud and Faculty, takes on the routine checks for householder applications and drafts cited reports that officers edit before deciding.

In pilots at Barnet, Camden and Dorset, DeepMind says it cut average officer assessment time on these tasks by up to 58%, and more than 70% of its generated reports were approved with no edits. Now twelve more councils are joining a 12-month Beta on live applications, bringing the total to 15, with a government plan to offer it to every council in England in 2027 if testing succeeds.

We think this is close to the right shape for public-sector AI. It targets the dull, high-volume work, document validation, checking against heritage constraints and policy, surfacing prior objections, and leaves the decision with a qualified officer. DeepMind says householder applications can make up to 70% of officers' workloads, so freeing time there is a real gain for residents and staff alike.

But let's be careful with the numbers. They come from DeepMind's own account of a pilot across three councils since June. Encouraging, yes. Independent, no. The Beta on live applications across metropolitan and rural districts is where we'll learn whether those results hold.

The statistic I keep coming back to is the one framed as good news: more than 70% of reports approved with no edits. That can mean the drafts are excellent. It can also mean officers, under pressure to clear backlogs, start approving what the machine wrote. A human in the loop only protects people if that human has the time and incentive to disagree. Automation bias is the quiet failure mode here, and the residents who lodge objections are the ones who lose if their concerns get summarized away.

There's also the dependency question. If this reaches every council in England, a core local-government function runs on one vendor's model. That's not a reason to stop. It's a reason to insist on transparency about errors, override rates, and what happens when the tool gets a heritage or policy check wrong.

Our read: useful, well-scoped, worth backing, and worth watching closely as it scales.

PlanAssist is part of Google DeepMind's broader partnership with the UK government on AI in public services. The government has set a goal of building 1.5 million new homes by 2029, and planning delays are one of the bottlenecks it's trying to clear.