Google Cloud AI to automate English council planning

UK ministries have begun using Google Cloud generative AI to automate local planning work. At the Google Cloud Summit in London, the Ministry of Housing, Communities and Local Government and the Department for Science, Innovation and Technology announced that Extract will be made available to every local planning authority in England and that an Augmented Planning Decisions prototype is in alpha in the London Borough of Barnet, Dorset Council and the London Borough of Camden.

Extract, developed by the government’s applied AI team the Incubator for AI and built with Gemini foundation models, converts unstructured data held in legacy PDF records into structured digital datasets within minutes. Officials reported trials in more than 20 councils and trial figures indicate the tool removes about 255 hours of manual data entry per council each year by parsing hundreds of pages of historical planning records into searchable records.

The APD prototype is described as an analytical assistant rather than an automated decision-maker. The software consolidates incoming documentation and flags missing information, extracts core geographic site details, identifies relevant national and local planning rules with linked policy citations, summarises public consultation responses and past precedents, and drafts initial evaluation reports with suggested conditions. Protocols require planning officers to review and approve every element of AI-generated material, and the prototype records its internal processing steps to create an auditable trail for each application.

The Gemini models are hosted on Google Cloud in an environment designed to protect civic data and preserve data sovereignty. Officials noted active controls to block malicious inputs and measures intended to reduce the risk of prompt injection. Google Cloud also supplies the computing capacity needed to run inference at daily operational scale.

The government said the APD alpha phase will test the software against different local policies and datasets before wider deployment. The plan is to offer APD to more than 300 English local authorities by 2027. Officials estimate automation could halve decision times for routine householder applications, which make up nearly 70% of planning submissions, and free officers to concentrate on larger infrastructure and commercial projects linked to the government’s target of building 1.5 million homes by 2029.

The programme brings together central ministries and the applied AI team with private partners including Google Cloud, Google DeepMind and Faculty, with ministers setting policy boundaries and private partners responsible for engineering and deployment. Lila Ibrahim, Chief AI Readiness Officer at Google DeepMind, commented that councils face a “mountain of paperwork” and that the tool was developed with councils to address those bottlenecks. Paul Maltby, Director of Public Services at Faculty, described the system as intended to cut time on simple applications so officers can focus on major developments. Naisha Polaine, Executive Director for Growth at Barnet Council, noted the prototype’s ability to gather information, make provisional assessments and draft reports could save officer time and speed decisions for residents.