Unlocking UK house-building with AI-accelerated planning

The UK government is partnering with Google DeepMind to develop an AI-accelerated planning tool designed to halve the time it takes to process homeowner planning applications, the company announced [1]. The prototype, built with Google's Gemini large language model, is being co-developed with Google Cloud, the AI firm Faculty, and local planning authorities in Barnet, Dorset, and Camden [1]. The government aims to make the tool available to all councils nationally from 2027 [1]. The initiative targets a persistent bottleneck: householder applications account for nearly 70% of all planning applications each year, and officers currently spend hours cross-referencing policy documents, historical files, and PDFs for each case [1]. The new tool is designed to consolidate data, identify relevant local policies, summarize public consultation feedback, and draft assessment reports, while the planning officer remains the final decision-maker [1]. The planning tool builds on Extract, a separate application built with Gemini by the UK government's Incubator for AI (i.AI) and made available to every council in England earlier this month [1]. Extract converts hundreds of pages of legacy planning documents into structured, usable data in minutes and is expected to save the average council around 255 hours of manual work per year, based on trials across more than 20 local planning authorities [1]. Google DeepMind, a British-American AI research laboratory and Alphabet subsidiary headquartered in London, has been responsible for developing the Gemini family of large language models since its merger with Google Brain in April 2023 [7]. The Gemini architecture is trained natively on multiple data types, allowing it to process and generate text, code, images, audio, and video simultaneously [5]. The 1.5 and 3 model generations introduced extended context windows that enable the analysis of large datasets, such as extensive document archives, in a single prompt [5]. The UK government's push to deploy AI in planning arrives alongside its target of building 1.5 million new homes by 2029 [1]. The construction sector has seen limited real-world adoption of automation technologies. A 2021 survey indicated that 55% of construction companies in the United States, Europe, and China reported using robots in some form, though real-world deployment remains confined largely to pilot projects and specific tasks such as surveying, inspection, and progress monitoring using drones or ground robots [3]. The unstructured and variable nature of construction sites has posed a fundamental challenge for deploying autonomous systems outside controlled factory settings [3]. The UK planning tool represents a different category of AI deployment—one focused on administrative decision support rather than physical automation. The prototype records its work at every step, creating a chain of thought and an audit trail for each decision, a design choice that echoes the human-in-the-loop frameworks adopted in other government AI programs [1]. The US Department of Defense's Project Maven, launched in 2017 to accelerate machine learning adoption for intelligence workflows, was similarly framed by its leaders as human-in-the-loop decision support rather than an autonomous platform [2]. Google withdrew from Project Maven in 2018 after internal employee protests [2]. Google DeepMind said it hopes the planning developments will serve as a model for other governments navigating the future of public service delivery [1].