A Grounded AI Strategy for Analytics Products

Every founder I talk to wants AI in their analytics product. The question I push back with is: AI to do what, exactly? Because in MarTech the easy win — a chatbot bolted onto a dashboard — usually solves a problem nobody had. The strategy I advocate uses the structure already in the product. The Insights-Before-Numbers pattern is, at its heart, a layer that interprets raw numbers into meaning. That's precisely where AI belongs: generating and prioritizing the interpreted insights, spotting the anomaly in a campaign worth surfacing, drafting the 'why' behind a shift. The raw numbers stay as the verifiable ground truth underneath. Critically, AI shouldn't end at narration. Each AI-generated insight carries the same action button as any other — so the model isn't just talking, it's routing the user to a decision. That keeps AI accountable: an insight that produces no action is one to question. For a founder, this is a defensible strategy because it's measurable and trustworthy. You're not promising magic; you're using AI to do the interpretation work at scale, with the numbers always one click away to check it. Augment the judgment, don't hide it.