Descriptions:
Chris Lovejoy — a Cambridge-trained medical doctor who later became an AI product builder and early employee at healthcare AI startup Anterior — presents a framework for how vertical AI companies can systematically embed domain expertise into their products and organizations. The talk was delivered at AI Engineer Singapore and builds on his previous presentation at the San Francisco AI Engineer conference, which reached roughly 100,000 viewers.
Lovejoy argues that winning in vertical AI is fundamentally an organizational problem, not a model selection problem, and that roughly 50% of generative AI projects were abandoned last year (per Gartner) largely because organizations failed to properly integrate domain knowledge. He introduces three roles domain experts can play: oracle (directly reviewing and iterating on outputs), evaluator (defining objective metrics and running systematic quality assessment), and architect (designing the pipelines and systems that encode expert judgment at scale).
Two detailed case studies anchor the framework. Granola — the AI meeting notes company that recently crossed a $1 billion valuation — is shown as a case where a journalist-background first employee operates as a near-permanent oracle, directly authoring and iterating prompts because meeting note quality resists objective measurement. Tandem, a UK medical AI scribe product, illustrates how a domain expert can begin as an oracle but must transition to an evaluator role as scale makes manual review impractical. The talk is practical guidance for any team building AI products in healthcare, legal, finance, or other specialized domains.
📺 Source: AI Engineer · Published May 16, 2026
🏷️ Format: Opinion Editorial
