Google DeepMind's PlanAssist Cuts Planning Officer Workload by 58%

An AI planning tool built by Google DeepMind cut the time planning officers spend assessing householder applications by up to 58% in early pilots. Now it's scaling up. PlanAssist is being made available to 12 additional local planning authorities across England, moving from a three-council pilot into a 12-month Beta phase using live applications.

The pilot results give the expansion its momentum. Since June, PlanAssist has run across Barnet, Camden and Dorset councils, and the early numbers are striking: average officer time on householder assessment tasks dropped by 58%, and more than 70% of the reports the tool generated were approved by officers with no edits required.

Those figures matter well beyond the councils involved. Planning bottlenecks carry real costs, and the UK government has attached a hard number to them: a goal of building 1.5 million new homes by 2029. Slow approvals slow that goal. For individual families, the same bottlenecks can mean months of waiting on approval for something as routine as a loft conversion, a conservatory, or an extension.

What is PlanAssist and who built it?

PlanAssist is an AI planning tool designed to handle the heavy lifting of routine applications, so officers can focus their expertise where it matters most. Google DeepMind developed it in collaboration with the UK government, Google Cloud, and Faculty. It's built with Gemini, and it forms part of Google DeepMind's broader partnership with the UK government to explore how AI can transform planning and other public services.

The design principle behind the tool is explicit: support human judgement, don't replace it. The tool speeds up routine checks while ensuring planning officers retain the final say on every application. No report becomes a decision without a human signing off.

The workload problem it targets is well defined. Routine householder applications account for up to 70% of planning officers' workloads. These applications are often simple to approve, yet they consume a disproportionate share of officer time. That mismatch — high volume, low complexity, heavy administrative burden — is exactly the gap PlanAssist is built to close.

How does the tool actually work?

PlanAssist gives planning officers automated support with two things: processing applications and drafting assessment recommendations. The workflow runs through several distinct stages.

• Document validation. The tool first validates application documents and architectural drawings, checking that all the required information has been submitted so the application can proceed.
• Heritage and policy checks. It then checks the application against heritage constraints — such as tree preservation orders — and against local and national planning policies.
• Prior objections. It also checks whether any issues or objections have previously been raised, including objections from local residents.
• Draft reporting. Finally, it drafts fully cited reports that planning officers can edit before reaching a final decision.

The 70% no-edit figure from the pilots speaks to the quality of that final stage. Reports covering whether an application complies with local planning rules, or whether the applicant has submitted all required documentation, were frequently accepted as-is by the officers reviewing them.

What changes in the Beta phase?

The move from pilot to scale is structured and deliberate. PlanAssist enters 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 number involved to 15.

The selection isn't accidental. The participating councils include metropolitan and rural districts across the country, a spread chosen to test PlanAssist's capabilities across a diverse range of communities. What works for a dense London borough may not work the same way for a rural district, and the Beta is designed to surface those differences before any wider rollout.

The end state, if testing goes well, is national. Subject to successful testing, the government plans to make the tool available to all councils in England in 2027. Even at full scale, the safeguard stays in place: every assessment will be reviewed and approved by a qualified planning officer before any decision is made.

Why does this matter for UK housing policy?

The stakes sit at the intersection of two pressures. On one side is the government's target of 1.5 million new homes by 2029 — an ambition that depends on the planning system moving faster than it currently does. On the other is the everyday experience of applicants, where a decision on a home extension can stretch across months.

Planning decisions shape everyday lives, as Google notes — from a family adding a nursery to a homeowner preparing for winter. By handling time-consuming administrative tasks, and by keeping planning officers in the decision loop, PlanAssist points to a model where AI helps cut backlogs and approve applications faster without removing accountability from the process.

The partnership framing matters too. PlanAssist is part of Google DeepMind's broader collaboration with the UK government on transforming public services with AI. A successful planning tool could become a template for other backlogged, document-heavy government functions. A failed one would sharpen questions about how much automation belongs in quasi-judicial decisions.

For now, the evidence base is three councils and a handful of months. The Beta's 12-month run across 15 councils, spanning metropolitan and rural districts, will test whether a 58% time saving in pilots holds up under live conditions at scale.

Google's stated intent is to keep going regardless of the answer's difficulty. Providing access to the new councils is, in the company's words, the natural next step, and it says it looks forward to continuing to support the UK government as it seeks to build more homes and unlock economic growth across the country. If the 2027 England-wide rollout happens on schedule, PlanAssist will become one of the most widely deployed AI decision-support tools in British public administration — with every final call still resting, by design, with a human planning officer.