English councils to trial Google AI planning tool to halve approval times

TL;DR:

• A Google-built AI system, the Augmented Planning Decision Tool, will be piloted by English councils this month to generate grant-or-refuse recommendations on building applications, with officers retaining final sign-off.
• Google won the £8.3m contract in February to build a tool that produces “reasoned, verifiable recommendations” and writes draft reports.
• Ministers want to halve council planning decision times to support the parliament’s target of 1.5 million new homes; the tool also targets the rise in AI-generated planning objections.

The pilot, reported by the Financial Times, is the most concrete deployment yet of generative AI in English local government decision-making. James Murray, chief secretary to the Treasury, told the FT the largest gains would come from “thinking about the end-to-end process and how you can make things quicker or even automated”. The system will be combined with a separate ongoing project to digitise property boundaries, conservation areas and other planning data so it can apply local plan policies to applications.

Context

The push sits inside a broader UK government effort to use AI for public-service productivity. The 1.5 million homes target has slipped against benchmark expectations, and ministers have framed planning system delays as the binding constraint. Council planners have separately reported a rise in AI-generated objections — boilerplate text submitted in volume — that adds workload without adding substance; a system trained to triage these is one of the pilot’s quieter wins.

Concerns from the planning profession are not about whether AI can do the work, but how. Daniel Slade of the Royal Town Planning Institute warned ministers not to exclude “human experience and values”, noting “no value in processing applications more quickly if the developments that follow are low quality”. Matthew Spry of consultancy Lichfields raised public-confidence questions and warned about training-data drift: “If they extrapolate judgments made in light of old policy or practice, or systemic biases, this might be at odds with the kind of more positive policy judgments that government wants to secure growth.”

Compared with earlier UK government AI procurement — the Home Office’s contested algorithmic visa system, DWP’s fraud-detection models — the planning tool has been launched with more visible public-sector framing of the human-in-the-loop principle. It also follows a £200m DSIT framework agreement with hyperscaler vendors signed last year that is making it easier for departments to deploy off-the-shelf foundation-model-based systems on UK government data.

Looking forward

The pilot’s outcome — whether processing time genuinely halves, whether refusals stand up at appeal, and whether public objections fall or rise — will set the political terms for any nationwide rollout. For UK SaaS vendors selling into local government, the FT’s reporting confirms that hyperscalers are now the default partners for high-stakes public-sector AI. The opening for smaller specialist suppliers is narrowing to evaluation, audit and bias-testing services rather than the front-line decision tooling itself.