"Artificial intelligence could greatly improve efficiency and decrease cost: generating and analysing housing or employment projections, reviewing and categorising site submissions, managing consultation and even auto-generating reports and analyses"
- John Mason - Carter Jonas
Imagine a future where master plans and house-type packs can be reviewed and tweaked at the click of a button. Where you can talk to a chatbot about whether your fence requires planning permission.
Where consultation on local plans or individual planning applications is targeted, tailored to your interests, and timely. Where local authority processes are streamlined and efficient, decisions are made on time, and officers have plenty of availability to discuss your proposals.
Sounds too good to be true? Perhaps it is. But artificial intelligence has the potential to revolutionise any industry it touches, and the planning profession is no exception.
Over 3.5 million planning applications are submitted every year in the UK. Over one-third of planning applications contain basic errors, and 250,000 hours are spent by council officers running validation checks.
The Alan Turing Institute has already run pilots on how application validation could be automated through the use of machine learning, training computers to “read” plans and documentation to check for errors. It found great potential to automate parts of the process, but the systems they trained were unable to differentiate between (for instance) existing and proposed plans, something a human could tell at a glance.
It is almost cliché to bemoan the length of time it takes to produce a new Local Plan – up to five years in some cases, by which point they could be hopelessly out of date. From the effects of the pandemic to the adoption of new technology to addressing the climate crisis, plan makers must be able to adapt quickly to fast-moving changes, but are currently unable to do so due to labyrinthine bureaucracy and the sheer amount of evidence that must be processed.
Artificial intelligence could greatly improve efficiency and decrease cost: generating and analysing housing or employment projections, reviewing and categorising site submissions, managing consultation and even auto-generating reports and analyses.
The DLHUC’s PropTech engagement fund is being used by 13 local authorities across the country to pilot the use of AI to manage public consultation on Local Plans. Authorities have adopted technology in a variety of ways: for instance, Greater Cambridge analysed social media feedback that wasn’t being captured on the consultation portal, whilst Southampton used 3D models to show how new proposals would look.
AI is able to review consultation responses and automatically categorise them, pick out key themes and identify trends. This promises significant improvements in the ability to run Local Plan consultations, which can attract tens of thousands of comments which currently take an enormous amount of time to process.
At the other end of the scale, could artificial intelligence be used to review minor planning applications?
Householder applications, Certificates of Lawfulness or conditions discharge take a significant amount of officer time to process but are for the most part relatively simple, requiring objective decisions on whether the submission accords with specific legislation.
As with the validation process, this could in theory be done by a computer program, with a planning professional required to review the final recommendation. Similarly, simple pre-application enquiries for small-scale development could be automated with a chatbot: you could interact with your planning department in the same way you would with your bank or mobile phone provider.
Are we then heading to a future where a development proposal can be managed, submitted, and determined by a computer?
I don’t think so. Artificial intelligence has no intrinsic agency (it must be told what to do) and no accountability (its output must be evaluated by an accountable human).
Have the consultation responses been summarised correctly? Do the auto-generated parts of a report make sense? Where have the data used in models or reports come from? Are there inaccuracies? Is it replicating unintended biases?
Planning in the UK is not a “tick-box” exercise and I do not believe should it be. Planning relies on the exercise of judgement and the weighing up of the planning balance. Considerations of design or the impact of a proposal on heritage assets are subjective.
Applicants and officers need room for discussion on where trade-offs or improvements can be made, and where departures from planning policies can be justified. And of course, decisions must have some kind of democratic oversight to ensure public good is balanced against private interest.
And we must be mindful of potential downsides. Automation could enable the targeting of more specific groups on a Local Plan or application consultation, aiding public engagement. On the other hand, the tailoring of consultation to what a computer perceives your interests to be could lead to an artificial narrowing of options or reinforcement of filter bubbles.
Also, whilst developers working across different authorities could benefit from the standardisation of validation requirements, automation could do away with the creativity and “colour” of individually prepared Local Plans or officer reports. Standardisation of masterplans to ensure they can be read by computers could also lead to further homogenisation of new developments at the expense of innovation and the ability to tailor a layout to site-specific circumstances.
Artificial intelligence is coming whether we like it or not. Local authorities, companies and individuals need to be able to be able to adopt, use and understand these tools, requiring time and investment. It’s no secret that many planning departments and consultancies are struggling with resourcing at the moment, which could lead to a future of winners and losers.
Nevertheless, artificial intelligence has enormous potential to speed up the development process for applicants and authorities. AI will transform the way we undertake data-driven and administrative tasks, easing workloads and allowing us to spend more time “planning”.
Negotiating good planning outcomes will continue to require human actors to exercise nuance, common sense, creativity and critical judgement, all things that cannot – and should not – be automated.
