KeyPedia Unlocked: Steve Prewitt

KeyPedia Unlocked: Steve Prewitt George Kwiecinski, co-founder and CEO of KeyPedia, sits down with Steve Prewitt, who leads the Applied AI practice at Genpact and heads its Anthropic partnership. Steve's path runs through Pfizer, Charles River, VNSNY, Healthfirst (as Chief Analytics Officer), Roivant's digital innovation program, and Exponential Data, the startup he helped scale to a Genpact acquisition in two years. He now mentors founders through PharmStars, the pharma tech accelerator where he and George met. They get into where AI is actually being bought in pharma and life sciences (hint: not AI co-scientists), why regulatory interaction is ripe for automation, what separates a real vertical AI company from an LLM wrapper, the death of seat-based pricing, and why 2026 is a crazy good time to be a founder. KeyPedia builds the regulatory intelligence layer for post-market quality and compliance, turning audits, inspections, warning letters, recalls, and enforcement signals into real-time, defensible insights for pharma quality and regulatory teams. Learn more: https://www.globalkeysolutions.net Chapters 0:00 Intros: how George and Steve met through PharmStars 2:33 23 years in New York tech 3:24 Pfizer, CROs, and bending a career into life sciences 6:09 VNSNY ride-alongs and Healthfirst analytics 8:10 Service companies vs direct industry: it's all B2B 13:00 Exponential Data and the Genpact acquisition 17:07 Mentoring founders and the Roivant playbook 20:44 Big companies and the innovator's dilemma (the Google case) 25:07 Leaders vs laggards: why biotech has more to lose 26:46 What pharma will actually buy: cycle time, variability, risk 29:21 Frontier models vs vertical AI startups 31:29 Beating the "it's just an LLM wrapper" objection 33:13 Selling AI to non-technical buyers 36:03 Automating the work nobody wants to do 38:22 Why regulatory interaction can be automated, MLR review 40:04 SaaS in 2026: systems of record and bolt-ons at risk 43:07 Forward-deployed engineers and the end of PowerPoint consulting 47:48 Data moats and the consortium model 50:29 Tokens, usage pricing, and what enterprises will pay for 55:27 Why it's a crazy good time to be a founder 57:04 Book picks: Project Hail Mary, Tomorrow and Tomorrow and Tomorrow #pharma #AI #regulatory #lifesciences #compliance #startup