Google DeepMind and UK Government Build AI Planner to Halve Planning Decision Times

The UK government is partnering with Google DeepMind to cut householder planning application decision times by 50%, using an AI prototype built with Gemini that is being trialed in three local planning authorities before a planned national rollout from 2027.

The partnership targets a concrete bottleneck in the government's pledge to build 1.5 million new homes by 2029. Local planning authorities routinely process dense paperwork and administrative backlogs, and householder applications — projects like loft conversions and extensions — account for nearly 70% of planning applications each year. Officers can spend hours on a single case, cross-referencing policy documents, historical files and PDFs before reaching a decision.

Google described the goal in direct terms: "Our goal is to help officers cut application decision times by 50%, freeing up time for planners so that more homes can be built."

The stakes are both economic and administrative. Britain's housing target depends on councils processing applications faster than the current manual workflow allows. Google framed the work as part of its National Partnerships for AI, an initiative that, in the company's words, "seek to support reimagining of public services to create more resilient societies."

Who is building the tool?

Google DeepMind is developing the prototype alongside the UK government, Google Cloud, the AI firm Faculty, and the local planning authorities in Barnet, Dorset and Camden. The tool is designed to act, in Google's words, as "a highly skilled assistant for planning officers, handling the heavy lifting of data extraction and case analysis."

The co-development approach is deliberate. By embedding the build process with councils that handle live caseloads, the partners aim to produce a tool matched to, as Google puts it, "the unique needs and day-to-day challenges that planners contend with every day."

Following the early trials in Barnet, Camden and Dorset, the government plans to make the tool available to all councils nationally from 2027. That timeline places the national deployment two years before the 2029 housing target deadline.

What does the AI tool actually do?

The prototype streamlines routine tasks across four functions:

• Consolidating data: It pre-processes backlogs, highlights data gaps and extracts key site information so planners can review everything on one screen.
• Identifying local policies: It highlights relevant national and local policies, pre-assesses compliance and provides exact citations for the officer to verify.
• Summarizing feedback: It reviews individual consultation letters to identify key objections or precedents.
• Drafting assessments: It creates a first draft of the final report, including the rationale and proposed conditions.

The scope is limited to the heavy lifting, not the judgment. By compressing the time spent on straightforward cases like loft conversions or extensions, the prototype could, according to Google, "help planning officers focus more on complex applications for public benefit."

Does the planning officer stay in charge?

Yes — and the design makes that explicit. The planning officer remains the final decision-maker throughout. Google states that officers "review every line the tool generates, edit the reasoning and retain the authority to approve or reject an application."

Accountability is built into the system's architecture. The prototype records its work at every step, creating what Google calls "a clear chain of thought and a robust audit trail for every decision to support the planning officer." That audit trail matters for a public-sector deployment: planning decisions are subject to appeal and legal challenge, and an undocumented AI recommendation would be difficult to defend.

What is Extract, and why does it matter?

The new planning tool builds on an existing foundation: Extract, a tool built with Gemini by the UK government's Incubator for AI (i.AI). Extract addresses a structural problem in UK planning — critical information locked in unstructured PDFs — and can turn hundreds of pages of legacy planning documents into usable data in minutes.

Extract was made available to every council in England earlier this month. Trials across more than 20 local planning authorities have already produced measurable results: Google says the tool is expected to save the average council around 255 hours of manual work a year.

The Extract rollout matters beyond its own numbers. It gives every English council a working example of Gemini-based document processing in a live planning context, and it supplies the structured data foundation on which the more ambitious planning assistant will operate.

Why is this significant beyond the UK?

Governments worldwide are exploring how AI can deliver public services faster, and the UK project offers an early test of whether frontier models can be safely embedded in quasi-judicial decision processes. Planning is a demanding use case: decisions require policy reasoning, factual accuracy, public accountability and a documented rationale. The UK prototype's answer — AI drafts, human decides, system logs everything — is a template other public administrations will watch closely.

Google is explicit about that ambition. "We hope these developments will serve as a model for other governments around the world as they look to navigate the future of public service delivery," the company said.

For councils, the near-term calculus is practical. A 50% reduction in decision times on applications that make up roughly 70% of annual volume would free officer capacity at scale, and the 255-hour annual saving already attributed to Extract suggests the baseline efficiency gains are real. Whether the 2027 national rollout hits its schedule, and whether trial results in three authorities transfer to the hundreds of councils with different systems and caseloads, will determine whether AI meaningfully accelerates Britain's path to 1.5 million homes — or simply tidies the paperwork around it.