Speeding up planning decisions with AI

Introduction

Planning departments across England face a growing backlog of householder applications—simple extensions, loft conversions, and conservatories that nonetheless consume a disproportionate share of officers’ time. A new AI‑driven service, PlanAssist, promises to change that dynamic. Built on Google DeepMind’s Gemini platform, the tool automates routine checks and drafts assessment reports, allowing human planners to focus on judgment‑heavy decisions. Early pilots in Barnet, Camden and Dorset have shown a 58 % reduction in assessment time, and the system is now entering a 12‑month beta that will expand to 15 councils.

How PlanAssist Works

1. Document validation – The system ingests the submitted forms and architectural drawings, confirming that all required fields are present and correctly formatted.
2. Policy cross‑check – Using the latest local and national planning regulations, it flags heritage constraints (e.g., tree‑preservation orders) and any policy breaches.
3. Historical context – It searches past objections, resident comments, and prior decisions to surface relevant precedents.
4. Draft report generation – A fully cited assessment is produced, ready for a planner to review, edit, and sign off.

The workflow is deliberately designed to keep the final decision in human hands. Planners receive a concise, evidence‑based draft that they can accept as‑is or modify, preserving professional accountability while eliminating repetitive clerical work.

Pilot Results and Early Impact

• Time savings: Average assessment time per householder application dropped by 58 %.
• Report accuracy: Over 70 % of AI‑generated reports required no edits before approval.
• Workload shift: Routine applications, which make up roughly 70 % of a planner’s docket, are now largely handled by the system, freeing officers for complex cases.

These gains translate into tangible benefits for residents—shorter waiting periods for home improvements—and for the broader economy, supporting the UK government’s target of 1.5 million new homes by 2029.

From Pilot to Scale

PlanAssist is entering a 12‑month beta phase that will roll out to twelve additional councils in three cohorts, bringing the total to fifteen participating authorities. The cohort design includes both metropolitan and rural districts, ensuring the tool is tested across diverse planning contexts. If the beta meets performance and safety benchmarks, the government intends to make PlanAssist available to all English councils by 2027, with every AI‑generated recommendation still subject to a qualified officer’s final review.

AI in Service of Communities

The PlanAssist case study illustrates a broader principle: AI can accelerate public‑service delivery when it augments—not replaces—human expertise. By handling data‑intensive, rule‑based tasks, the system reduces backlogs, improves consistency, and allows planners to apply their professional judgment where it matters most. As more councils adopt the tool, the cumulative effect could be a faster, more transparent planning process that supports both individual homeowners and national housing goals.

Looking Ahead

Continued collaboration between Google DeepMind, the UK government, Google Cloud, and academic partners such as Faculty will be essential to refine the model, address bias concerns, and ensure robust security. Future iterations may expand beyond householder applications to larger development proposals, integrating richer spatial data and community‑engagement inputs.

Bottom line: PlanAssist demonstrates how responsibly deployed AI can streamline a core public service, delivering measurable efficiency gains while preserving the essential role of human decision‑makers.