Google DeepMind is teaming with the UK government, Google Cloud and Faculty to prototype an AI planning assistant built on Gemini. The system targets a 50% cut in decision time for homeowner applications, a crucial step toward the nation’s goal of delivering 1.5 million new homes by 2029, by automating document cross‑referencing and risk scoring. Early trials focus on boroughs facing backlogs.

The prototype extends the earlier Extract tool that converts legacy PDFs into searchable data. In Barnet, Camden and Dorset it pre‑processes backlogs, flags relevant national and local policies, summarises consultation letters and drafts initial assessment reports. The system also highlights data gaps, prompting officers to fill missing information before final review. Since householder filings make up nearly 70% of cases, automating these steps frees planners for complex projects.

Planners retain final authority; the AI logs every inference to provide a transparent audit trail. Trials suggest the assistant could shave hundreds of hours from each council’s workload, directly supporting the housing pipeline. Early feedback shows faster turnaround on loft conversions and extensions, the most common case types. The tool is slated for national release in 2027, offering a reusable model for other governments seeking AI‑enhanced public services.