AI in planning – worth it, or worry?
We read recently that Barnet and Camden London Borough Councils and Dorset Council are trialling two new government-backed AI tools, with the aim of halving the time it takes to process householder planning decisions.
A further 10 local planning authorities are set to join the trial later this year, before a proposed national rollout in 2027 – depending on the evidence gathered.
But it got us thinking: in this new AI age, is it worth it? Is it a worry? And, what needs to happen to make sure AI doesn’t fall over and break everything?
So, in true Law and Order style, we’ll be objective and argue the case for and against.
But first - what is it?
The £8.2m prototype is created by the UK government together with Google DeepMind, Google Cloud, Faculty, local councils, and funded by the Ministry of Housing, Communities and Local Government.
Google DeepMind says the prototype aims to…
- Triage applications.
- Extract and combine data.
- Find relevant policy and constraints.
- Summarise consultation feedback.
- Draft an assessment and report to help planning officers make a decision.
The second tool, Extract, is available to all councils in England. It uses AI to help planning officers digitise decades-old planning documents and maps, sometimes with handwritten notes, and convert it into readily useable data.
Why now? The case for using AI
Around 350,000 planning applications are submitted every year in England, with householder applications such as extensions and loft conversions accounting for nearly 70%.
The current system relies heavily on planning officers checking old documents. For every application, officers must check the local planning rules that apply. Many of these are hidden away in a raft of documents. Anyone who’s looked for documents on a portal will know how arduous navigating these spaces can be.
In addition, the recent government fees for planning applications consultation highlighted a £330m gap between application fee income and service costs in 2024/2025.
These critical time pressures and underfunding are stretching planning teams’ capacity to the limit, leading to costly delays and slower decision-making.
By reducing time spent on straightforward cases, the prototype tested in Barnet, Camden and Dorset could help planning officers focus more on complex applications, such as new housing, major developments and key infrastructure.
This is a vivid example of the argument for using AI across all industries, where on (digital) paper it frees up time spent on everyday tasks to enable teams to focus on strategic and high value work.
In that spirit, the government stresses that the tool is meant to help officers make decisions, not “decide on their behalf”, and that they remain the key decision-maker.
Sounds good, right?
Yes, but… the case against using AI
We found that AI does a pretty good job of highlighting the risks posed by its involvement in the planning process – listing several issues. Our experience and use of AI in shaping our own work points to risks leaders should seriously consider around the AI planning prototype. These include:
Accuracy of information
The prototype tool is meant to analyse applications, find relevant policies, summarise consultation responses, then provide an initial assessment for officers. But what if it gets something wrong as it can do often, like missing a relevant policy, misreading a resident objection, or inventing a planning precedent. The “sorry, my bad!” response won’t cut it, here.
Reliance on AI for critical thinking
It’s welcome that the government says that a human planning officer will always be the key decision maker. But with councils under pressure to halve decision times, there is a significant risk that officers will rely too heavily on AI’s first assessment and rubber stamp applications to save time and move onto the bigger ticket items in their inbox.
Transparency
The prototype is designed to produce reasoned, verifiable recommendations, but that’s not the same as proving that its reasoning is complete, lawful, or contestable. If an officer report is AI-assisted, councils will need a clear audit trail showing what AI suggested, what the officer checked, what was changed, and why… and that’s more work for the officer.
Legal accountability
Planning decisions are governed by the law, local plans and human rights obligations. The government explicitly says that an AI tool used in this space must comply with this. However, the major risk here is practical accountability: if the AI-generated assessment is wrong and challenged, the council remains legally responsible. That means councils will need assurance that the prototype tool does not expose them to any potential legal challenge.
Scope creep
While the initial focus is on householder applications, the prototype tender notice says that the tool is being developed with a view to expanding into other application types, such as change of use, listed-building consents and demolition in conservation areas to name a few. The more complex an application in AI, the bigger the risk to councils.
Public trust
Even a correct AI-assisted decision may be questioned if residents believe a “black box” influenced it. The government says officers remain the decision-makers, but public trust depends on whether councils disclose AI use, explain its role, and let applicants or objectors challenge factual errors in AI-produced summaries.
Are we in an AI planning arms race?
While councils trial the new AI planning tool, others turn to AI platforms, such as Objector AI, to help them object to planning applications. This means that both councils and community members are using AI to influence planning outcomes. As councils need to implement safeguards with the prototype tool to make sure it’s fit for purpose, the same can be said from an objector perspective. AI doesn’t get it right all the time and the prototype must keep pace with AI in objections, as noted by Pinsent Masons.
Information security
Data is the world’s most valuable commodity, right? But we need to make sure it’s held securely and safely. AI can help to speed up planning work, yes. But AI remembers everything, too, which means safeguarding both personal and commercially sensitive information is a major risk for planning teams. This exposure can be a haven for hackers, and in turn pose a legal risk, too.
What do we think (the verdict)?
As we’ve said on several occasions, and in our AI policy, we’ve got to put people at the heart of digital transformation. There’s no replacement for the human element, particularly around critical thinking and safeguarding. In many ways, this trial isn’t really a test of artificial intelligence at all. It’s a test of how effectively humans can use it.
AI isn’t the silver bullet some claim it to be, nor is it the Judgement Day-style existential threat that others fear. Used right it can remove some of the burden that’s weighed down planning departments for years, freeing up officers to focus on the decisions that genuinely need human critical thinking and professional judgement.
But…
The success of these tools ultimately depends on the quality of the safeguards around them. They can’t be treated as an afterthought and must be baked in from the outset. If government want planning authorities to make decisions faster, they need the time and resources to challenge AI outputs, rather than simply accept them.
For me, the jury is out until the risks are addressed thoroughly with clear governance procedures and processes. But one thing is certain: as AI becomes more embedded in planning, getting the balance and processes right between humans and tech will be every bit as important as the technology itself.
Header photo by Ekaterina Bolovtsova on Pexels.
Written by
Share
Latest news & insights
Inspiring news, insights and comments from our Distinctive team.
