Pendomonium 2026: Built by you, powered by Pendo

What the video covers

This is the opening keynote from Pendomonium 2026, Pendo’s annual product management conference, held in Raleigh, North Carolina on March 24–25 with 2,000 attendees. Todd Olson, Pendo’s CEO and co-founder, leads the session and is joined by OpenAI Chief Economist Dr. Aaron Chatterji and Christian Idiodi from SVPG. The keynote announces eleven platform launches and frames them around a central argument: that companies are deploying AI agents faster than they can understand what those agents are doing, and that product observability is the next operational gap to close.

Who it’s for

Product managers at companies that have begun shipping AI features and need to think about how to measure them. Also useful for product leads who are building the case for dedicated analytics infrastructure as AI surfaces in more user-facing touchpoints. The SVPG contribution brings a perspective on product fundamentals that keeps the session from becoming a pure vendor announcement.

Key takeaways

1. Adoption metrics no longer capture the relevant gap. Chatterji introduced the phrase “intelligence divide” to describe the compounding advantage accruing to teams that use AI most effectively — not just teams that use it at all. For product managers, this reframes the internal benchmark from whether users click the AI button to whether those interactions produce outcomes that reduce real work.

2. Agents create a new instrumentation problem. Olson opens with a specific observation: AI agents are being deployed at a pace that outstrips the ability to monitor them. Pendo’s response is Novus, a product agent that lives in a codebase, watches how users actually experience a product, and takes action based on what it finds — surfacing issues, recommending improvements, and verifying impact without requiring manual log review. This positions agent-level observability as infrastructure, not a feature.

3. Shipping velocity and quality compound in opposite directions depending on what you measure. A demonstration from Builder.io described a 3x increase in code shipping velocity achieved by optimizing for meaningful feedback and functional quality rather than output count. Teams that measure AI impact at the task-completion level — not the lines-written or features-shipped level — consistently outperform those that treat speed as the primary metric.

4. Intelligence in software is becoming a category expectation, not a differentiator. Idiodi’s framing from SVPG is that proactive use of technology — understanding what is technically feasible and building toward it — is what distinguishes teams that set direction from those that react to it. As AI capabilities become table stakes, the product judgment about what to build with them becomes the actual moat.

Worth watching if…

Your team has shipped AI features and is now trying to figure out what to measure and how to improve them, or if you are evaluating where to invest in product analytics infrastructure as agentic workflows become more common in your product.