Google Just Gave UK Councils an AI Tool to Cut Planning Times in Half. Here’s What’s Actually Happening.

A quick recap: why does UK house building keep stalling?

The UK government has a target: 1.5 million new homes built by 2029. That is an enormous number. And yet, one of the biggest bottlenecks is not land, not money, and not builders. It is paperwork.

Local planning authorities — the councils that approve or reject planning applications — are drowning in it. Every application for a loft conversion, an extension, or a new home means officers manually cross-referencing policy documents, historical files, and dense PDFs. It takes hours, sometimes days, per case. And householder applications alone account for nearly 70% of all planning applications every year.

The result? A backlog that slows down everything — and a housing shortage that keeps getting worse.

So what did Google DeepMind actually build?

On June 16, 2026, Google DeepMind announced a partnership with the UK government to build an AI-powered planning tool for councils. It is being tested right now in three local authorities: Barnet, Camden, and Dorset.

Think of it as a very capable assistant sitting alongside a planning officer. The tool does the heavy lifting before the officer even opens a case file. Here is what it handles:

Consolidating data — It pre-processes application backlogs, flags missing information, and pulls key site details onto a single screen. No more toggling between twelve PDFs.

Identifying relevant policies — It reads through national and local planning policy, highlights what applies to this specific application, and cites exactly where those policies appear so the officer can verify them.

Summarising consultation responses — When residents or stakeholders send in objection letters, the tool reads through all of them and identifies the key themes and concerns.

Drafting the assessment report — It produces a first draft of the officer’s final report, including proposed conditions and the rationale behind them.

The planning officer still makes every decision. They review every line the tool generates, edit the reasoning, and retain full authority to approve or reject. The tool also keeps an audit trail of its work at every step, which matters enormously for accountability in public decision-making.

The goal is to cut decision times by 50%.

What is Extract, and why does it matter here?

Before this new planning tool, Google DeepMind and the UK government’s AI incubator (called i.AI) built something called Extract.

The problem Extract solves is simple but stubborn: critical planning information is often buried in old, unstructured PDFs. Documents that have never been digitised properly. Extract can turn hundreds of pages of legacy planning documents into usable, structured data — in minutes.

It was rolled out to every council in England earlier this month, and trials across more than 20 local authorities suggest it saves the average council around 255 hours of manual work per year.

The new planning tool builds on top of what Extract started.

What is the invisible thing happening here?

Here is what the press release does not say loudly enough.

This is not a small efficiency tweak. This is a government making a deliberate bet that AI can handle a substantial chunk of routine public administration — and doing it openly, with named local authorities, published timelines, and a stated accountability model.

That is actually unusual. Most government AI projects stay vague. This one has specifics: councils in Barnet, Camden, Dorset. A 50% target on decision times. A rollout to all councils nationally from 2027.

The political stakes are real too. The UK government is under enormous pressure on housing. If this works, it becomes a template. Not just for planning, but for every other area of public administration buried under paperwork.

And Google DeepMind knows this. They are not just selling a product here. They are building a reference case they can take to every other government in the world.

What is happening now?

The tool is in active trials across Barnet, Camden, and Dorset as of June 2026.

If those trials go well, the UK government plans to make it available to all English councils by 2027.

Google DeepMind has framed this explicitly as a model for other governments — part of what they are calling National Partnerships for AI, with similar deals already signed with Singapore, South Korea, and India.

Whether it delivers the 50% cut in decision times remains to be seen. But for the first time in a long time, Britain’s housing bottleneck has an AI-shaped attempt at a solution.

Sources including image: Google DeepMind Blog, UK Government i.AI , UK government harnesses Gemini to support faster planning decisions