DeepMind Creates AI to Cut UK’s Planning Time by Half

Google DeepMind has teamed up with the UK government to develop an AI prototype which could cut the time taken by planners to process householder planning applications in half. Currently, it’s being piloted in three councils (Barnet, Camden and Dorset), but it hopes to roll it out across the country by 2027. Developed using Google Cloud and artificial intelligence company Faculty, it seeks to address the administrative burden of processing standard permits — officers can spend hours trawling through policy PDFs to check whether a small application meets guidelines.

The prototype itself won’t make decisions, but it will gather information about a site, identify specific national and local policies with references, synthesize comments made during public consultation, and create an initial assessment report including suggested conditions. Every piece of information generated by the prototype will be checked and amended by a planner, who will also keep track of how it reached its conclusions. Householder applications (such as extensions or loft conversions) account for almost 70% of planning applications in England every year, so even if this process saves just a few hours per application, it could have a huge impact.

1Why this matters beyond Britain

Anyone who’s ever had to wait months for a permit to extend their house knows that planning can be a frustrating process. With the UK aiming to build 1.5 million new homes by 2029, it’s not always a case of there being a shortage of building materials — sometimes, it’s simply that planners are spending too much time wading through paperwork. By focusing on the admin side of things instead of trying to automate decision-making, we can use AI to handle the heavy lifting (i.e. checking and collating policies) while leaving the actual decision making to humans. And since this approach will work equally well for other types of permits, benefit claims etc., we can expect other countries to follow suit.

This isn’t the first time that Google DeepMind has worked on UK planning systems. Previously, it developed Extract alongside Gemini, which digitises historical planning documentation and has been rolled out to all English councils (reportedly saving them an average of 255 hours per year). This latest prototype builds on top of those efforts — after all, you can’t automate anything if you don’t have access to digital versions of your policies! According to Naisha Polaine from Barnet Council, “the tool’s capability to gather pertinent information has the potential to save considerable officer time.”

What makes this really clever is that it only produces suggestions which need to be approved by a human. That’s what we want from AI in the public sector — automation where possible, but keeping humans accountable. Of course, there are potential downsides — if the AI is responsible for summarising feedback from public consultations, it might overlook or downplay concerns raised by members of the public. Similarly, if it references policies, it might make mistakes. But as long as these reports are reviewed carefully before being signed off, and there’s a clear paper trail showing exactly what happened at each stage, then everything should be fine. And remember, this is still just a pilot scheme involving three councils — it’s going to take a lot more work before it’s ready to handle millions of applications.