At Gathering #5, Julien Steel, Head of Product at Henchman, opened with a deliberate narrowing of scope. The talk would not cover differences between LLMs, prompt engineering techniques, or vector database architectures. Instead, Julien focused on the product questions that determine whether an AI feature gets adopted or abandoned: how to categorize AI capabilities, how to set user expectations, and how to evaluate performance in production.
Henchman, a legal tech company founded in 2020, has grown to 35 people with 300% revenue growth, over 10 million in venture backing, and hundreds of clients across 5 continents including Top 100 law firms. Their product puts a legal knowledge base at lawyers' fingertips: connect a database, process contracts, access clauses and definitions, and enrich everything with AI recommendations. The numbers behind the product tell the story of scale: an average of 825,000 contracts processed, 13.5 million clauses extracted, 1.6 million terms defined, and 1.5 million documents on average per deployment.
