Artificial Intelligence in Planning

Artificial intelligence (AI) tools are being considered as a way to speed up decision-making, decrease workload and automate administrative tasks carried out by planners, as part of delivering the Plan for Change to deliver 1.5 million homes over the next parliament. Two government departments, the Ministry of Housing, Communities and Local Government (MHCLG) and the Department for Science, Innovation and Technology (DSIT), are working closely to develop a generative AI tool which converts old planning documents into digital data in 40 seconds, freeing up time for planners, who would otherwise spend 1-2 hours on this administrative task, in order to focus on determining applications with quality data.

MHCLG is also exploring other ways in which AI can be used in the planning system to enhance the way data is accessed and extracted from planning documents. A 6-week project focused on using AI to process and interrogate large volumes of data from local plans to extract insights and answer sample questions. The accuracy of the responses ranged from 43% to 59% to 76% depending on the question, highlighting that the quality of the input data impacts the accuracy of the tool. Because AI is known to sometimes produce false or biased results, the importance of citation features linking to source material remains important. The project also developed a way to classify local plans and create a standardised table of contents, which has led to separate work to consider how this can be applied to organising consultation representations. 

Another interesting project was carried out by the Alan Turing Institute (ATI), who trained machine learning models on planning applications to detect objects such as floorplans. For context, ATI identified 12 common mistakes that lead to 80% of applications being rejected, which included: incomplete sections, incorrect fees or descriptions, incorrect or missing drawings, floorplans, elevations, scales, north arrows or site plans, all of which have to be manually identified by planners. Object detection was performed using a deep learning framework to predict and classify drawings within applications. This model showed potential for speeding up the evaluation process by quickly detecting common mistakes before the application is reviewed by a planner. However, at this stage, the development model sometimes made false predictions and cannot yet recognise whether a plan is an existing floorplan or a proposed. 

There is growing enthusiasm for testing what AI tools are capable of being used for to streamline processes under the planning regime but it is important to note that decision making and regulatory bodies will also have guidelines about the use of AI. For example, the Planning Inspectorate (PINS) published guidance in October 2024 which governs the responsible use of AI and requires disclosure of the systems, source of information and what material the AI has been used on.

It is also worth noting that the Government recently launched the formal qualifying process of its new initiative relating to Artificial Intelligence Growth Zones (AIGZ). The intention is for AIGZs to have expedited regulatory pathways to enable quicker development of AI technology infrastructure including hyperscale data centres and computing facilities. This is in addition to the revisions made to paragraph 86(c) of the NPPF in December 2024 which requires planning polices to pay particular regard to facilitating development to meet modern economic needs including identifying suitable locations for data centres and digital infrastructure.

If you have any queries about AI in planning, please do not hesitate us.