Why You Can’t Build AI Without Progressive Delivery

Former GitHub CEO Thomas Dohmke once told James Governor, “You can’t do AI-based development without progressive delivery.” It’s a striking claim, but Governor — co-founder of the analyst firm RedMonk and co-author of the new book “Progressive Delivery” — sees the connection as obvious. In this episode of The New Stack Makers, Governor sat down with TNS Founder and Publisher Alex Williams at AWS re:Invent to explore why the same disciplines that transformed software releases over the past decade are not only still relevant but also genuinely critical for anyone building with AI. Guardrails Built for Uncertainty The parallel becomes clear when you consider what AI systems actually do in production. Models hallucinate. Behavior varies widely between large language models. Outputs are non-deterministic in ways traditional software never was. “You’re rolling something out, you’re not exactly sure how it’s going to behave, how the system will behave. What if you change the model? What impact is that going to have?” Governor asked. “We need to measure that. We need to see its actual behavior. If you know, then think about doing a rollback. All of those things — that’s progressive delivery.” The practices that emerged from feature flags, canary releases and observability aren’t just nice-to-haves for AI applications, he added. They’re the foundation. AI concepts like evals, versioning and controlled rollouts have roots in the traditional software delivery life cycle. DevOps Forgot Someone But Governor’s book isn’t just about AI. It challenges a fundamental assumption baked into how we talk about software delivery. “In DevOps you’ve got Dev and you’ve got Ops, right? But where’s the third loop?” he asked. “Where is the user, and the feedback from the user about their experience of your digital product and service?” The book builds on a framework of four A’s (abundance, autonomy, alignment and automation) to argue that progressive delivery closes that gap. It’s about getting the right product to the right user at the right time, and actually measuring whether you succeeded. To underscore the point, Governor quoted Charity Majors, co-founder and CTO of Honeycomb.io: “It doesn’t matter if you have five nines if the user isn’t happy.” He pointed to Amazon as a case study in navigating these tensions. The company’s legendary two-pizza teams and service ownership model gave engineers enormous autonomy but also created alignment challenges when different teams made different tooling choices. Now Amazon is pursuing more standardization, trying to balance developer freedom with organizational coherence. Listen to the full conversation to hear how the book came together with co-authors Kim Harrison, Heidi Waterhouse and Adam Zinman, why IT Revolution’s Gene Kim decided to publish it, and what Governor learned from talking to companies living these progressive delivery challenges firsthand. The post Why You Can’t Build AI Without Progressive Delivery appeared first on The New Stack.