England is expanding AI planning to 15 councils—but its 58% time claim has no public denominator
PlanAssist will assist officers for 12 months after a three-council trial. Officials report faster tasks, yet published sources do not disclose the sample, task definition or error rate—and Parliament says no control group was used.
The short version
- What happened
- PlanAssist is moving from Camden, Barnet and Dorset to 12 more councils, making 15 councils in a 12-month test. Officers will use it alongside normal processes and remain responsible for decisions.
- Who it affects
- Homeowners and small businesses could receive routine planning decisions sooner, while planners could spend less time assembling documents. The available evidence does not yet show whether time savings persist or whether errors, rework, appeals or unequal outcomes change.
- What remains unproven
- Official announcement rather than independent evaluation
Keep in mind
Official announcement rather than independent evaluation
No public sample or denominator
No control group in the first phase
No task definition
No council-level results
No accuracy, override, rework or appeal data
No uncertainty interval
Confounding from staffing and process reforms
Government and commercial delivery interests
No final national-rollout decision
Multi-source analysis · 3 linked sources. Primary analysed source date: 30 October 2026. Earliest linked supporting source: 16 June 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 5 min The evidence, methods, implications and limitations
A three-council prototype is becoming a 15-council service test
England is expanding PlanAssist, an AI-supported planning tool, from Camden, Barnet and Dorset to 12 additional councils for a 12-month test. The system brings an application, local policy and site constraints into one workspace and produces an initial assessment for a trained officer. It is intended for routine householder cases—extensions, loft conversions and similar applications—which the government says make up nearly 70% of applications. Officers remain responsible for checking the material and making decisions; councils will use the tool alongside their normal process rather than as mandated software.
The expansion is material because the original June announcement described an alpha trial backed by an £8.2 million government contract with Google Cloud, Google DeepMind and Faculty. The latest release says early testing cut time spent on selected tasks by 58% and helped reduce routine householder processing from eight weeks to four. It also states an intention to make the tool available to every English council by the end of 2027. Those are consequential public-service claims, but they come from the department and commercial partners building the system—not an independent evaluation.[1][3]
The published evidence cannot isolate what caused the reported improvement
Neither government release identifies how many applications, officers or tasks produced the 58% figure. There is no definition of a key task, council-by-council result, baseline distribution, confidence interval, accuracy measure, appeal rate or rework count. The eight-to-four-week claim is also not accompanied by dates, a denominator or a comparable group of applications. A written ministerial answer on 20 October adds a decisive qualification: the first phase involved three local planning authorities and used no control group. It described the results as promising and said no final national-rollout decision had been made.
That design means the figures cannot show how much change came from PlanAssist. Planning time can also shift with application mix, staffing, seasonal demand, policy, document quality and local workflow. A new national delegation scheme and 420 government-backed planner recruits are being announced alongside the technology. If those changes overlap, a before-and-after comparison could attribute their combined effect to the AI tool. The official release also does not report whether faster review changes missed constraints, incorrect recommendations, officer corrections, applicant resubmissions or later appeals.[1][2][3]
The distinction that matters
What changed—and what the evidence can establish
The programme is scaling, but the public evidence remains an uncontrolled early test.
Test scale
Three councils
15 councils for 12 months
Reported speed
Selected planning tasks
58% less time claimed
Evaluation design
First testing phase
No control group used
Not reported
Accuracy, rework and appeals
No public denominator or error rate
Faster decisions would help applicants, but the next test needs measurable safeguards
For homeowners and small businesses, a shorter wait could reduce uncertainty and the cost of holding up building work. For planners, assembling policies and constraints automatically could return time for complex cases. The corresponding risk is that an incomplete or misleading synthesis creates extra checking work—or looks authoritative enough that an omission travels into a decision. Human sign-off is important, but it does not by itself measure whether officers detect system errors or whether workload pressure changes how closely they review recommendations.
The 15-council phase is an opportunity to publish a stronger evaluation. The key research question is: compared with ordinary processing of similar applications, does PlanAssist reduce end-to-end determination time without increasing corrections, omissions, rework, appeals or unequal outcomes? A useful protocol would predefine eligible cases and outcomes, report denominators and council-level variation, preserve an appropriate comparator, and record officer overrides and applicant experience. Until those results appear, the 58% figure should be read as an official early-testing claim—not proof that the system halves planning delays at national scale.[1][2]
Continue the story
Read next
Continue exploring human oversight: compare the finding and its evidence limits.
Can AI assist judges without deciding cases?
England and Wales' judiciary says approved AI may support tasks such as summarisation, transcription, redaction and document analysis, but every output used in judicial work must be reviewed and approved by the responsible judge. The position sets governance boundaries; it does not publish pilot results or authorise autonomous decisions.
Read the next reportExplore the full source and evidence record
What this means for people
- Homeowners and small businesses could receive routine planning decisions sooner, while planners could spend less time assembling documents. The available evidence does not yet show whether time savings persist or whether errors, rework, appeals or unequal outcomes change.
Global context
This is an English local-planning programme operating under England's planning law, staffing model and council data. Its early figures should not be generalised to public-service AI in other countries or to higher-stakes automated decisions without local evaluation.
What the evidence does not yet show
- Official announcement rather than independent evaluation
- No public sample or denominator
- No control group in the first phase
- No task definition
- No council-level results
- No accuracy, override, rework or appeal data
- No uncertainty interval
- Confounding from staffing and process reforms
- Government and commercial delivery interests
- No final national-rollout decision
What to watch next
- Published evaluation protocol
- Application and officer denominators
- Comparable non-PlanAssist cases
- Council-level processing times
- Accuracy and omitted-constraint rates
- Officer overrides and workload
- Applicant resubmissions and appeals
- Equality and accessibility outcomes
- Vendor cost and procurement transparency
- National rollout decision
Living evidence record
Impact record IAI-1BL7GHU
Evidence stage
Announced
Confidence
Supported
Reporting basis
Multi-source analysis
Independent or research support
Not yet
Record status
Monitoring
Last checked
31 October 2026
Source trail
3 direct sources across 2 source types.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Evidence trail
Sources used for this report
Links checked 31 October 2026
Research question, topics and regions
Whether PlanAssist reduces end-to-end planning time and error without weakening accountable professional review
Topics: Local government · Planning · Public services · Human oversight · Evaluation · Google DeepMind · Faculty
Regions: England · United Kingdom · Europe
This report is labelled multi-source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.