Marily Nika on Why AI Isn't the Product—The Experience Is

The Foundation of Growth Hasn't Changed—AI Just Amplifies It

Most growth teams are rushing to slap AI onto their products without asking the right question: does this actually improve the user experience?

Marily Nika, who has held AI titles for over 10 years at Google, has a blunt message for founders trying to capitalize on the AI wave:

AI is not the product. AI tools, AI models, AI agents, these are not the product. The product is the experience.

— Marily Nika

She sees teams confusing the tool with the outcome. AI can personalize onboarding. It can predict churn. It can surface the right content at the right time. But none of that matters if you don't know what value you're trying to deliver and who needs it.

The fundamentals still apply: set a North Star metric, use data analytics tools like Amplitude or Mixpanel to identify drop-off points, and run experiments to validate hypotheses. AI doesn't replace that discipline. It enhances it.

Personalization Is the Unlock—But Only If You Know Your User

Generic growth tactics are dying. Notifications that say "Want to try this out?" don't cut it anymore. What works is hyper-personalized messaging that reflects who the user is and what they care about.

Nika points to Spotify as a benchmark. When Spotify tells you "People who listen to the songs you like are adding these two new songs to their playlist," it's not just a recommendation—it's convenience, discovery, and social proof rolled into one.

When it's like, 'Hey, Marily, we think you would like this,' is by far incomparable to any other strategy. Especially if it even highlights why you would like it based on some information it has about you.

— Marily Nika

AI makes that level of personalization scalable. You can tailor onboarding flows, predict what features a user will engage with next, and automate customer support in ways that feel human. But it starts with data and clarity about what your North Star experience looks like.

Early-Stage Startups Can Move Faster Than Big Companies

One of the most counterintuitive advantages? Startups with limited data can still win with AI.

Nika's advice: don't wait for a massive dataset. Use pre-trained models from GPT-4, Hugging Face, or Google AutoML. Synthesize data using tools like Snorkel AI or Synthesis AI. Or just hardcode initial data to validate your hypothesis.

She worked with a travel startup that wanted to build personalized vacation agendas but had no user data. Her response? Hardcode it. Test the concept. See if people find value. Then scale.

Start with a very narrow use case. Start with a very narrow hypothesis. Solve this tiny problem that's critical, like improving sign-up conversion or whatever, and then see if this is for you, and then the rest will come.

— Marily Nika

Startups also have cultural advantages. They're not paralyzed by compliance reviews or legacy infrastructure. A team of 3 or 4 people using AI to automate research, write code, and analyze user behavior can operate like a team of 20. Big companies are still figuring out whether employees can use Grammarly without violating data policies.

Measuring AI Success Requires a New Mental Model

Growth teams are used to tracking retention, conversion, and engagement. AI adds a new layer of complexity.

Nika breaks success metrics into three buckets: AI proxy metrics (accuracy, false accepts, false rejects), classic product metrics (retention, engagement, growth), and system health metrics (can your infrastructure handle millions of AI-powered interactions?).

The tricky part? AI is probabilistic. You can't just optimize for short-term conversions and assume long-term retention will follow. A change that boosts initial payments might increase churn 9 months later. That's why experimentation culture is non-negotiable.

Run small pilots. Use A/B testing platforms like Optimizely. Hypothesize outcomes—like "personalizing onboarding will increase day 7 retention by 10%"—and validate them before scaling.

Nika also sees a future where AI becomes more proactive. Instead of you querying a dashboard, an agentic AI system will monitor your North Star metric and alert you when something shifts. Tools like Zapier are already moving in this direction, automating workflows based on triggers you define in plain English.

The PM Role Is Evolving—And So Is the Team Structure

Building AI products introduces a new stakeholder: the research scientist.

Traditionally, PMs worked with engineers to ship features. Now there's a trio: the PM, the engineer, and the scientist who builds the model. Each has different incentives. Scientists want to publish research and perfect accuracy. PMs want to ship fast and learn. Engineers want to integrate everything without breaking production.

Nika sees her role as navigating trade-offs all day. When is the model good enough to launch? How do you balance privacy, cost, effort, and impact? It's a dance that requires AI literacy across the entire team.

That's why she founded an AI Product Academy—to teach PMs, founders, and growth leaders how to bridge the gap between technical capabilities and user value. Her prediction? Every product manager will eventually be an AI product manager. The question is whether you'll learn proactively or get left behind.