SaaStr 839: Why Most SaaS Companies Will Fail at AI (And How to Avoid It) with Intercom's CPO

What It Covers

Intercom's Chief Product Officer details the company's transformation from SaaS to AI-first, including launching Fin AI agent that resolves over 1 million customer queries weekly at 65% resolution rate. He covers cultural changes, engineering practices, go-to-market shifts, and specific mistakes SaaS companies make when attempting AI transformation.

Key Questions Answered

• •Cultural transformation requirements: Successful AI transformation demands changing everything - org structure, product roadmap, build processes, metrics, sales approach, and pricing models. Companies must go too far to know where boundaries are, including eliminating entire teams and processes. Intercom's CPO took over two-thirds of marketing, deleted the marketing calendar, and rebuilt from scratch because iterating from existing state would not achieve necessary change.
• •Product architecture shift: AI-first products require three layers - AI/RAG system, application layer, and custom model layer. Companies must break workflows into discrete steps, point AI at each component, and understand that reliability compounds across steps. Even with high individual step performance, complete workflow success rates drop to 90% or lower, requiring systematic experimentation and custom model training for competitive advantage.
• •Engineering velocity doubling: Intercom mandated every designer ship code to production, moving from zero designers shipping code to 100% within 18 months. Design became cheap instead of expensive, enabling PMs and engineers to prototype rapidly. The company set hard metrics for doubling engineering productivity using AI coding tools, making adoption non-negotiable - team members either embraced the change or left the company.
• •Go-to-market complexity: AI product buyers changed from single decision-makers to three-person committees - the functional leader, a C-level executive responsible for AI transformation, and an AI-fluent technical evaluator. These buyers operate in different universes, attend different events, and require distinct marketing approaches. Product differentiation shifted from UI features to infrastructure quality, RAG system performance, and scientific evaluation rigor at scale.
• •Self-harm decisions necessity: Companies must make revenue-damaging decisions to win long-term, including accepting 10% revenue hits and disrupting existing seat-based business models. Intercom launched Fin knowing it would cannibalize existing product usage. Avoiding customer feedback from those resisting AI proves critical - many customers initially rejecting AI later became Fin users after their own transformation pressures increased.