UK Government Deploys AI Planning Tools Nationwide on Google Cloud

The UK government has reached a pivotal stage in its effort to digitally transform local planning systems, with two AI-powered tools now moving from trial phases toward nationwide deployment. The Ministry of Housing, Communities and Local Government (MHCLG) and the Department for Science, Innovation and Technology (DSIT) announced the progress at the Google Cloud Summit London, marking a significant step in how England's 300-plus local authorities will process planning applications and policy documents.

Two distinct tools, one infrastructure

The Extract tool, developed internally by MHCLG and the government's Incubator for AI (i.AI), has completed its rollout to all councils across England. The system automates the conversion of complex planning documents into digital formats, a task that previously consumed substantial staff time. According to the announcement, Extract is projected to save the average council approximately 255 hours of manual document processing work.

Running in parallel is the Augmented Planning Decisions (APD) prototype, a more complex collaboration involving MHCLG, Google Cloud, Google DeepMind, and delivery partner Faculty. Currently in alpha testing with three local authorities—the London Borough of Barnet, Dorset Council, and the London Borough of Camden—the tool is designed to help planning officers interpret and apply local policies. The government aims to make APD available nationally starting in 2027, though officials emphasized that current results remain early and experimental.

Both systems operate on Google Cloud infrastructure and leverage Gemini AI models, but their development paths differ significantly. Extract represents an in-house government build, while APD involves direct collaboration with Google's research and cloud teams.

Security and scale requirements

The choice of Google Cloud as the foundation reflects specific technical and regulatory demands. Processing sensitive government data at national scale requires enterprise-grade security controls, including protections against prompt injection attacks and assurances of data sovereignty. By deploying Gemini models within Google Cloud's protected environment, the government maintains control over how AI systems access and process citizen information.

The infrastructure must also support elastic scaling. If MHCLG proceeds beyond the alpha phase, the APD system will need to serve hundreds of local planning authorities simultaneously while maintaining consistent performance for daily public-facing operations.

Why it matters

These deployments represent one of the largest practical applications of generative AI in UK public services to date. Planning systems are a persistent bottleneck in housing development and local governance—automating document processing and policy interpretation could materially accelerate decision timelines. The dual-track approach also offers a test case for how governments can balance internal AI development with external partnerships, particularly when working with frontier models that require specialized infrastructure. Success or failure at this scale will likely influence procurement decisions across other UK government departments.

Looking ahead

The 2027 target for national APD availability gives the government roughly two years to move from alpha testing with three councils to production readiness across England. That timeline will depend on evaluation results from the current pilot sites and decisions about whether to proceed with Google Cloud as the long-term infrastructure provider.

Details of the deployments were first reported by Google in a company blog post announcing the updates at the Google Cloud Summit London.