Artificial intelligence entered the planning debate in 2025 with competing expectations about speed, access and reliability. One concrete government project was Extract: a tool intended to turn historic planning documents into structured information that councils and software systems could use.
Announced in June, Extract addressed a practical bottleneck. Important details about sites, boundaries and policy areas were often stored in paper maps, scanned documents or older formats. Finding and converting that information required manual work before it could support modern digital services.
The government said trials in Hillingdon, Nuneaton and Bedworth, and Exeter had digitised planning records in around three minutes each, compared with one to two hours manually. Those were trial results for processing records, not a promise that complete planning applications could be decided in minutes.
MHCLG’s technical account described separate tasks: reading text, tracing boundaries and placing those boundaries accurately on a modern map. Its team tested outputs against known answers, underlining that different parts of the process needed different accuracy checks.
The distinction matters. Converting a map into usable data is different from deciding whether a development should receive permission. A faster route to information can support officers without removing the need to assess evidence, policy and the circumstances of a proposal.
At the June announcement, the government expected wider council availability by spring 2026. That was the timetable presented at the time, rather than evidence that every council was already using the system in November 2025.
For applicants and residents, the potential benefit was easier access to consistent records. For councils, it was the prospect of reducing repetitive processing and directing more staff time towards substantive work.
The project offers a more precise way to discuss AI in planning: identify the task, check the output and separate demonstrated results from future ambitions. Better data can help a planning service, but it is not the same thing as an automated planning judgment.