During Kestrel’s design-partner year, before the product was released, architects kept asking a version of the same question. In Marian Pulford’s telling: “I know that this error is correct, because I’m an architect who’s been in the field for 30 years. How would my youngest member of staff, who got out of grad school a year ago, know to verify this?”
That question sits at the center of “Where We Don’t Put AI,” Marian’s conversation with Alex Lee on the third episode of What’s Left?, Lee’s podcast on leaders thriving in an age of AI. A senior architect can spot a wrong answer on sight. The people who can’t are the ones a compliance tool actually has to be built for, and taking that seriously changes where AI belongs in the product.
The answer Kestrel landed on is the part of the episode worth hearing in full. The detection of errors is deterministic. “We can run that same check a thousand times and get the exact same answer,” Marian says. “People can open up a log, audit the response that they’ve gotten from us, and say: I can really believe what this software is telling me.” The standard behind that is concrete. An architect who wants to open a 10,000-line error report and verify every finding by hand can do exactly that.
AI is in the product where it earns its place. In Marian’s words, “it’s 2026, and no competent technologist would build software without AI in it.” It does the reading, processing the enormous volume of legal text the checks run against, and it does the explaining, unpacking what a violation means, why it matters for safety or accessibility, and how to fix it without breaking something else in the building. And for the roughly 20 percent of code that is genuinely interpretive rather than black and white, the software never presents an interpretation with the same confidence as a determinate finding.
The conversation ranges wider than the product. There is the industry’s missing middle: a generation that left architecture in the Great Recession, senior specialists now retiring, and firms asking, in Marian’s words, “Bob’s going to retire. He’s the guy who always knows these answers. What are we going to do?” (Architosh made the same argument about Kestrel’s launch.) There is the permitting saga at Denver’s RiNo Art District that started the company, where, as her architect co-founder put it, “everyone loses money and sleep and hair.” And it closes on taste: “You can slopify the entire world if you’re not careful.”
The written version of the argument is Marian’s essay “Where AI doesn’t belong.” The episode is the version with the stories in it.
Listen to the full episode on Apple Podcasts, Spotify, or YouTube.
Or open up the log yourself, on one of your own projects: Schedule a demo →