AI products have a retention problem. The initial excitement of a demo fades quickly once users encounter hallucinations, unpredictable outputs, or simply a tool that does not fit into their existing workflow. At Gathering #5, Julien Steel, Head of Product at Henchman, addressed this head-on with a talk grounded in the hard lessons of building an AI-powered contract drafting tool for legal professionals.
Henchman operates in the legal tech space, where accuracy is not a feature request but a non-negotiable. Lawyers will not tolerate a tool that gets things wrong even occasionally. This constraint forced the team to think differently about how AI products earn and keep trust, and the resulting insights apply far beyond legal tech.
The core argument was simple but counterintuitive: the model is not the product. The experience around the model is. And that experience must be designed for trust, not just capability.
Original Presentation Slides
Download the slides from this talk as presented at the gathering.
