Google DeepMind is partnering with the UK government to build an AI planning prototype that aims to cut the time it takes to process householder planning applications in half. According to Google DeepMind, the tool is built with Gemini and is already being trialed with local planning authorities in Barnet, Camden and Dorset, with a national rollout to all councils planned from 2027.
This is a direct attempt to clear one of the biggest bottlenecks standing between Britain and its target of 1.5 million new homes by 2029.
What was announced
- Google DeepMind is co-developing an AI prototype alongside the UK government, Google Cloud, Faculty, and councils in Barnet, Dorset and Camden.
- The goal: help planning officers cut application decision times by 50%.
- It follows Extract, an earlier Gemini-built tool from the government’s Incubator for AI (i.AI) that turns old planning documents into clean digital data.
- The prototype falls under Google DeepMind’s National Partnerships for AI program.
Why it matters
Householder applications, things like loft conversions and extensions, make up nearly 70% of all planning applications each year. For a typical case, officers spend hours cross-referencing policy documents, historical files and PDFs by hand. That manual grind is the bottleneck.
What stands out here is the targeting. Rather than chasing flashy use cases, Google DeepMind is going after the high-volume, repetitive work that eats officers’ time. Clear the routine cases faster, and planners get more room to focus on the complex applications that actually need human judgment.
How the tool works
According to Google DeepMind, the prototype acts as a skilled assistant for planning officers, handling the heavy lifting across four jobs:
- Consolidating data: Pre-processing backlogs, flagging data gaps, and pulling key site information onto one screen.
- Identifying local policies: Surfacing relevant national and local policies, pre-assessing compliance, and providing exact citations to verify.
- Summarizing feedback: Reviewing consultation letters to spot key objections or precedents.
- Drafting assessments: Producing a first draft of the final report, including rationale and proposed conditions.
The human stays in charge
This is the part worth paying attention to. The planning officer remains the final decision-maker. They review every line the tool generates, edit the reasoning, and keep full authority to approve or reject.
Google DeepMind also says the prototype records its work at every step, building a clear chain of thought and an audit trail for each decision. That accountability layer matters in public-sector AI, where a wrong call isn’t just an inconvenience but a decision a citizen can challenge. Drafting assessments and citing policy is exactly where an AI tool could hallucinate, so the verify-everything design and the citation requirement are sensible guardrails.
The bigger picture
Governments worldwide are testing how AI can deliver public services faster, but most pilots stay stuck in proof-of-concept. What’s different here is the path to scale: a real workflow, named councils running trials, and a stated 2027 national rollout. The earlier Extract tool gives this effort a track record to build on rather than starting cold.
It also signals where the public-sector AI market is heading. Google Cloud and Gemini are now embedded in core government infrastructure decisions, not just experiments. For practitioners building AI tools for regulated or high-accountability environments, the design pattern on display (AI does the heavy lifting while a human verifies and signs off) is becoming the template that gets these systems approved.
What to watch next
- Trial results from Barnet, Camden and Dorset. The 50% figure is the goal, not a proven outcome yet. Real decision-time data will tell the story.
- The 2027 national rollout. Whether a prototype tuned to three councils generalizes to hundreds is the hard part.
- Accuracy and trust. How officers handle AI-drafted assessments, and whether the audit trail holds up under appeal, will shape adoption.
If it delivers, this could become a reference case for AI in public administration well beyond planning. Full details are available at the original Google DeepMind source.