A legal marketplace where the hardest step is knowing what to ask

The problem

Legal marketplaces fail on the demand side, not the supply side. Lawyers are willing to list. The difficulty is that a person with a legal problem usually cannot name it, and a directory that asks them to pick a practice area is asking for the answer, not the question.

LeyFind's brief was [what leyfind asked us to solve]. Underneath it sat the structural problem every marketplace of this shape has: matching quality is bounded by the quality of the intake.

The approach

We put an assistant in front of the directory rather than beside it. Its job is narrow and stated plainly to the user: take a description of a situation in ordinary language and produce a structured brief a lawyer can read in thirty seconds and price.

That framing set the engineering constraints:

• The assistant classifies; it does not advise. Anything that reads as legal advice is a liability for the platform and for the professionals on it, so the model's output is a structured intake record, not an answer.
• Every match is explainable. A user can see why a professional was surfaced, because an unexplained ranking on a two-sided platform is indistinguishable from paid placement.
• The supply side is a first-class product. The professional's inbox, not the consumer's search page, is where a marketplace is retained or lost.

A marketplace is only as good as its intake. The AI here is not a feature bolted to a directory — it is the mechanism that makes the directory usable at all.

The outcome

Delivered in [n] weeks: consumer intake, the assistant, professional profiles, the matching surface, and the professional-side workflow.

[what changed for leyfind after launch].