Planning and AI: opportunities and limitations

The last edition of Planning and Development InSite addressed the role of AI in planning, with an insightful piece by John Mason giving an overview of AI in planning.  

In just two months since that piece was written, AI has continued to develop and extend its influence.  

AI is already used extensively in the property industry – in everything from masterplanning and construction to lettings and property management. The UK PropTech Association and British Property Federation recently reported that over 78% of industry leaders are seeking to incorporate AI into their business in the next three to five years.  

So just how usable is AI in planning? For planning consultants and developers, is AI an ‘existential risk’ (to quote tech leaders including Elon Musk) or a ‘force for good’ (as stated by 1,300 experts in response)? To what extend can it support us in our day-to-day activities?  

To understand AI’s capabilities, I delegated some routine planning tasks to ChatGPT.  

Task 1: An application covering letter  

Instruction to ChatGPT: “Write a covering letter for a reserved matters application pursuant to outline planning permission Ref: P/XXXX/20”.  

• AI was very good in preparing a well-structured letter, but the accuracy of the letter could not be relied upon. For example, the letter set out the five matters which could be reserved as part of an outline planning permission but included ‘Ecology and Environmental Impact’. This is actually not a reserved matter as defined in article 2 of the Town and Country Planning (Development Management Procedure) (England) Order 2015 and therefore should not have been included.  

I had hoped that AI would be able to find more information on the original planning application from the internet (for example, the date of permission, a description of the development and the relevant local planning authority), on the basis that the information is publicly accessible. It didn’t – but this may be due to me using the free version of ChatGPT.  

Overall, helpful to see how AI suggested setting out a covering letter for this type of application but the level of information required was not detailed or accurate enough. For this type of task, our proforma approach carried out by the team remains preferable.  

Score: 2/10  

 

Task 2: A speech for a planning committee  

Instruction to ChatGPT: “Draft a speech to address planning committee as the agent of an application for the construction of 11 homes on a brownfield site in Merstham, in the Borough of Reigate and Banstead. The applicant is an affordable housing provider and the application will provide 100% affordable housing”.  

• ChatGPT’s text was a helpful tool to start gathering ideas for a speech, especially the introduction and conclusion. I wouldn’t use the full speech that ChatGPT gave me, but I think I could take parts of it for adaption and expansion.  

AI is potentially useful for this task because of its ability to summarise the key points for a 2-3 minute speech - something that can challenge those immersed in the project. ChatGPT proved very adept at framing the benefits including affordable housing provision and provision of a biodiversity net gain.  

On the other hand, to prepare for a planning committee meeting, a planning consultant needs to know the full detail in advance of members’ questions. To rely on using AI to write the speech would risk missing out on this important stage of preparation.  

Furthermore, ChatGPT lacks the knowledge about consultation that had been undertaken with neighbours, local community, the local authority.  

Overall, I think this is a helpful tool for getting started on preparing a speech for a planning committee and could help save time. However further input is definitely needed and the knowledge behind the application is required anyway for on the evening questions.  

Score: 5/10  

 

Task 3: A flyer ahead of a public consultation event  

Instruction to ChatGPT: “Prepare text for a flyer for a public consultation event ahead of a planning submission for 10 new dwellings”  

• The information provided was very accurate  

I would definitely look to use ChatGPT for this function again. Apart from adding in specific details on the date of the event and a small summary of the scheme proposed, I don’t think I would need to add much to the output.  

Score: 8/10  

 

Task 4: Consultation summary

Instruction to ChatGPT: “Summarise objections to a planning application using the following comments…”.  

• Because I inputted the raw data into the website, I was reassured of the accuracy of this output.  

Again, a very useful and accurate function which would certainly save time in the future. It would also be useful for case officers, sparing them time in writing reports to committee.  

Score: 9/10  

 

Conclusion

Based on my experience of the free version of ChatGPT, my view is that it is most reliable when the raw data is inputted by the user. I wouldn’t trust AI for specific planning policy or legislation as in my experience this has been inaccurate, but it could be a useful tool for gathering a structure for a report, speech or flyer, saving significant time.  

One of the limitations to this research was that I was only able to use the free version; if I had used the paid-for version, the function would also use information available on the internet. Whilst past learnings are important, planning requires us to be more forward thinking: only people can provide the creative new ideas that are required in planning.  

It's also important to bear in mind that planning isn’t straightforward, and one size does not fit all. Decisions are a careful balance subjectivity and objectivity. What might have been acceptable in one location or instance may not be acceptable in another as public sentiment, local issues, local policies and planning history vary in each instance.  

So, in conclusion, while there is a role for artificial intelligence in completing some of the more straightforward tasks (albeit with human oversight) relating to planning, when it comes to decision-making, human intelligence must prevail.