Creating Intelligent Products

This article addresses the challenge of how product managers should approach building "intelligent products" that blend deterministic and probabilistic approaches using AI technologies. Marty Cagan argues that many product teams, especially in B2B and regulated industries, mistakenly view probabilistic AI solutions as irrelevant to their mission-critical businesses.

Cagan's approach emphasizes that probability is central to intelligence, not an alternative to it. Drawing from his experience dating back to 1980s expert systems, he explains how early AI attempts failed because they relied purely on rule-based deterministic systems. He advocates for embracing probabilistic behavior as integral to intelligent products, citing examples like Waymo's autonomous driving (which handles life-critical decisions while continuously learning from 1,500+ vehicles), Google Translate, Spotify's Discover Weekly, and Netflix recommendations.

The key outcomes highlighted include Waymo's successful deployment across four major US cities after a decade of development, representing over 50 million miles of real-world driving experience and 20 million miles of daily simulated driving that enables fleet-wide learning.

For product managers, the critical takeaway is to move beyond viewing products as either purely deterministic or probabilistic. Instead, they should embrace hybrid approaches where AI's probabilistic nature becomes a core feature rather than a bolt-on addition. This is particularly relevant for domains where human experts already rely on "educated guesses based on probabilities" – which Cagan argues applies to virtually every type of expertise from medical diagnosis to sales forecasting.