On 17 June, the Ministry of Housing, Communities and Local Government unveiled two AI tools designed to speed up England's planning system. The timing is not subtle. The government has staked its credibility on building 1.5 million homes this Parliament, and the planning system has long been blamed as the chokepoint that keeps that number out of reach.
The first tool is a prototype called Augmented Planning Decisions, now in early testing with Barnet, Camden and Dorset councils. It triages householder applications, summarises the key information and produces an initial assessment for a planning officer to review. The ambition is to cut decision time on straightforward extensions and loft conversions from eight weeks to four.
The second, Extract, is already live everywhere. Every council in England can now use it to convert decades of paper planning records, some with handwritten annotations, into searchable digital data within minutes, a task that previously consumed hours of manual transcription per file.
Both tools were built with Google DeepMind, Google Cloud and the AI consultancy Faculty, working alongside local planning authorities.
Marc Waner, chief executive of Faculty and chief technology officer at Accenture, framed the prototype as a way of redirecting scarce planning capacity: "By using AI to support planning officers with clear recommendations, with humans retaining final sign off, we will help cut approval times on simple renovations in half. This will give councils more time and resource to focus on the bigger infrastructure projects that will improve communities and drive economic growth."
Why householder applications are the test case, not the target
It is worth being precise about what these tools actually touch. Householder applications, the loft conversions, extensions and similar small-scale projects that homeowners submit, account for close to 70% of all planning applications in England each year. They are also, by definition, not the housing supply problem. A single extension does not move the needle on 1.5 million new homes.
What the government is betting on is a capacity argument, not a volume one. If ‘Augmented Planning Decisions’ genuinely halve the time officers spend on routine cases, that time gets redeployed to the major housing and infrastructure applications that do count towards the target.
Minister for Data and Modern Digital Government Ian Murray put the logic plainly: "When someone wants to add a bedroom or convert their loft, they shouldn't be waiting months for a straightforward decision. And planning officers shouldn't be spending hours digging through decades of paper records when making the decisions that really matter."
That is a reasonable theory of change. It is also one that depends entirely on execution that has not yet been tested at scale. The prototype is live in three boroughs. National rollout, if the pilot succeeds, is targeted for 2027. Every assessment will still require sign-off from a qualified planning officer, which is the right safeguard but also a reminder that the tool augments judgement without replacing the queue.
A planning system already mid-overhaul
This announcement does not land in a vacuum. England's planning apparatus has been undergoing sustained legislative change since the Planning and Infrastructure Act received Royal Assent in December 2025, and the Ministry has been consulting on a revised National Planning Policy Framework that places more weight on housing delivery over the caution that ministers argue has characterised the existing system.
A new National Scheme of Delegation, laid in Parliament earlier this month, is designed to speed up decisions on smaller applications such as larger extensions and loft conversions by removing them from committee scrutiny altogether.
Seen against that backdrop, the AI tools look less like a standalone tech launch and more like one lever among several being pulled simultaneously: legislative reform, delegated decision-making and now automated triage. Industry voices remain sceptical about whether any of this addresses the deeper structural issues.
Steve Leitch, a housebuilding figure quoted by The Intermediary, argued that uncertainty in the system has become a bigger problem than delay itself, particularly for smaller developers: "It starts with sites, schemes and the practical reality of trying to move projects through a system which, for many, increasingly feels unpredictable and at times simply broken."
Developers, he added, are becoming more selective about which sites they pursue and slower to commit as labour costs and regulatory complexity rise.
That tension matters for how this story should be read. Speeding up the easy 70% of cases is a genuinely useful operational fix. It does nothing, on its own, to resolve the unpredictability around major scheme approvals that developers say is now their bigger constraint.
What this means for project pipelines
For project delivery professionals, the practical implications sit in two places.
First, expect councils piloting the prototype to start clearing householder backlogs faster, which should reduce the noise around minor works competing for the same planning officer attention as larger schemes.
Second, watch how quickly Extract gets used to digitise historic site records relevant to brownfield and infill projects. A planning officer who can pull up decades of prior decisions on a site in minutes, without requesting paper files from an archive, changes the due diligence timeline for anyone assessing site constraints.
There is also a governance signal worth tracking. Every output from Augmented Planning Decisions is reviewed by a human before a decision is issued. That human-in-the-loop design, paired with the recent embarrassment of AI-generated hallucinations surfacing in professional reports elsewhere this month, suggests the government is conscious of the reputational risk in deploying AI on statutory decisions. Expect that safeguard to be cited repeatedly as the pilot expands, and expect scrutiny if it is ever loosened.
The bigger picture for AEC
We have tracked AI moving from back-office productivity tool to front-line decision support across the built environment this year, and this is one of the clearest public-sector examples yet.
Lila Ibrahim, Google DeepMind's Chief AI Readiness Officer, framed the opportunity in similarly structural terms, pointing to the chance for the UK to use AI to unblock a system that has constrained delivery for years, going beyond simply digitising its existing inefficiencies.
Whether that ambition survives contact with three London boroughs and a 2027 rollout target is the question worth revisiting in twelve months. For now, the signal for AEC leaders is straightforward: planning capacity is being treated as a solvable engineering problem, no longer an immovable constraint, and that shift in framing is arguably more significant than the tools themselves.
Takeaway
The 70% of applications being sped up are not the ones blocking the 1.5 million homes target. The capacity freed up for major schemes is the real metric to watch, not the householder turnaround time itself.
Firms with live applications in Barnet, Camden or Dorset should expect faster householder turnaround now, and should flag any project sitting on data from those councils as a candidate for early benefit from Extract's digitisation of historic records.
The human sign-off requirement is doing political as well as practical work. If KPMG's hallucinated AI report taught the market anything this month, it is that the optics of unchecked AI output in professional or statutory contexts carry real reputational risk, and government will be cautious about being the next cautionary tale.
Developer sentiment data suggests unpredictability, not raw speed, is the bigger constraint on site selection right now. Tools that speed up decisions without improving consistency may not move investment decisions as much as ministers expect.
Track the 2027 national rollout milestone closely. A successful pilot that fails to scale on schedule would be a more telling story than the launch itself.
Ready to build smarter?
Project Flux tracks where AI is actually changing how projects move through approval, financing and delivery, beyond simply where it is being announced. Subscribe for the weekly newsletter that separates genuine operational shifts from press release noise.
Links and Stuff
All content reflects our personal views and is not intended as professional advice or to represent any organisation.
/
1