Natural Language Search

Problem

Expert users still build rigid queries when the real need is to ask a complex question and trust the path to an answer.

Constraints

Scale across heterogeneous legal corpora, preserve control for power users, and keep results inspectable.

Process

Map current search failure modes, prototype natural-language entry points, and validate with internal analytics and attorney feedback—while directing four concurrent workstreams around one delivery plan.

The turning-point decision

Treat search success as a product outcome—not a feature list—and redesign around time-to-useful-answer.

Final solution

A natural-language search experience with clearer feedback loops for users and for the teams tuning the models.

Measured outcome

A faster path from complex questions to useful results, supported by product analytics and daily model-tuning insights.