Citations are first-class.
Every finding points to a specific paragraph. If we can't cite it, we don't claim it. The discipline coding agents needed to be useful is the same discipline we hold ourselves to.
About
We brought coding-agent techniques to UK planning.
AI coding agents have spent the last few years going from autocomplete to genuinely useful colleagues. They read entire codebases, follow citations across files, plan multi-step changes. The same model class works on planning policy. We did the work to make that practical.
Planning policy is dense, indexed, and citation-driven. A local plan references its own supplementary guidance, which references the national framework, which references statutory instruments. It is - in shape, if not in vocabulary - a codebase. Cross-referenced documents, formal vocabulary, every claim checkable against a specific paragraph.
A coding agent reasons about a repo it's never seen before because the structure is consistent and every claim can be checked against a specific file and line. The same is true of planning policy. Different domain, same shape of problem.
That's the foundation GapSense is built on. Not "AI for planning" in the abstract, but the specific stack of techniques that made coding agents useful - large-context reading, citation-as-output, structured reasoning, tool use - applied to the corpus that decides UK planning applications.
What changes when you apply it
Three principles, then we let the work speak.
The corpus beats the abstract.
Generic AI planning advice is worth less than reading the actual plan that decides the actual application. We do the latter.
Engineering rigour, not vibes.
Cited findings, deterministic structure, fast turnaround. The discipline that makes coding agents useful is the discipline that makes planning ones useful too.
Why now
The capability arrived. Most of UK planning hasn't noticed yet.
Two years ago this product wouldn't have worked. The models couldn't hold a planning corpus end-to-end and reason about it in a single pass. Today they can. The failure mode now is execution, not capability.
We're early enough that most of UK planning hasn't yet seen what becomes possible when an agent has read every paragraph of the policy that decides an application. The teams that adopt this in 2026 set the bar everyone else is held to in 2028.
I've been using coding agents for the past 12 months and the impact on software engineering is immense. I want to build the same to help planning consultants dramatically improve their workload.
Martin Alderson - Founder, GapSense
Operated by Codified London Ltd, registered in England. Press, partnerships, sample reports: hello@gapsensehq.com.
Better applications. Faster decisions.
Tell us about your practice. We'll come back with how GapSense fits.