Y CombinatorSpenser Skates: How Amplitude Rewired Itself Around AI
Through a bottoms-up experiment phase that triggered hard reorgs; Amplitude rebuilt its analytics roadmap around AI-native teams and dropped SaaS assumptions.
At a glance
WHAT IT’S REALLY ABOUT
Inside Amplitude’s Painful, All-In Reinvention As An AI-Native Company
- Amplitude CEO Spencer Skates describes how the company shifted from deep skepticism about AI to making it a central pillar of its product and org strategy.
- The transition required top-down conviction, bottoms‑up tooling adoption (like an internal “AI week”), multiple org restructurings, and replacing or augmenting leaders who were strong in classic SaaS but not AI-native.
- Skates contrasts traditional SaaS product building—customer request–driven, deterministic, reliable—with AI product building, which is technology-first, probabilistic, and demands users who tolerate failure and iteration.
- He also reflects on founder psychology, how to learn totally new skills (like enterprise sales), and the difficult evolution from hands-on founder to large-company public CEO in the AI era.
IDEAS WORTH REMEMBERING
5 ideasAI product strategy must start with understanding model capabilities, not customer feature requests.
Because AI capabilities are “jagged,” customers can’t clearly specify what’s possible; teams need a technology-first view, then map those capabilities back to real workflows and products.
Driving AI adoption in an existing company requires both symbolic and practical interventions.
Amplitude ran an AI week, mandated hands-on use of tools like Cursor, showcased live coding demos, and explicitly used a “burn the boats” metaphor to signal that AI wasn’t optional.
Leaders and org structures must change when a SaaS company goes AI-native.
Amplitude did two major reorganizations in a year, moved out strong but non–AI-native leaders, acquired AI-first teams, and created dedicated AI product teams rather than treating AI as side projects.
AI-era users must be willing to work with imperfect systems that frequently fail.
Unlike traditional SaaS where one failure kills trust, AI workflows often involve repeated prompts, model switching, and editing; products must support this iterative, error-tolerant usage pattern.
“Feature, not company” dynamics and commoditization will crush shallow AI tools.
Skates views many AI visibility products as easily replicable and ripe to be given away for free by incumbents; sustainable businesses must sit downstream of such features with deeper value propositions.
WORDS WORTH SAVING
5 quotesThere is going to be a reinvention of analytics in the next few years, and we want to be the ones to go lead it.
— Spencer Skates
In SaaS, you go to your customers, ask them what they want, prioritize that list, and start building it. With AI, you have to start from what the models can actually do.
— Spencer Skates
There is a point that you get to a year, maybe two years in, where from a rational standpoint, you probably should quit. But for whatever reason, the successful ones don’t.
— Spencer Skates
As a founder, your job is always to run to the most difficult problem in the business and lead from the front.
— Spencer Skates
You have to be very clear in your own head about what you’re trying to learn, and then be open to where it comes from.
— Spencer Skates
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome