Why great AI products are all about the data

Shaun Clowes is a product leader who has served as chief product officer at Confluent and MuleSoft, previously worked at Metromile and Atlassian (where he built an early B2B growth team), and co-created Reforge courses on retention and data for PMs. The conversation covers why many PMs underperform, how to use outside-in thinking and LLMs for customer research, why proprietary data and context drive AI product value, how B2B SaaS moats hold up against easy cloning, and career and decision-making advice.

01Key learnings

• Start every document from the customer, market, and competitor perspective rather than internal execution, since outside-the-building thinking is what creates real value.
• Right-size qualitative research: interview roughly 7 to 14 people, because fewer gives too little signal and more stops yielding new insights.
• Use LLMs to hunt for where your strategy doesn't fit customer feedback, not to confirm what you already believe, and to infer competitor strategy from public documents.
• Build a continuous feedback river of interviews, NPS, competitor info, and usage data so insights flow in constantly rather than in occasional bursts.
• For AI products, invest in data management: models are replaceable, so value comes from bringing timely, high-quality, well-structured context to the model.
• In B2B SaaS, defensibility lives in configured business rules and years of embedded workflow evolution, not the UI or data model, so those rules also constrain agents.
• When a data result contradicts your intuition, first assume it is wrong; check upstream, downstream, and the broader business metric before presenting it.
• Balance PLG with sales-led motions; a company with both many customers and substantial revenue is very hard to displace, and incentives determine which behaviors actually happen.

02Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.