Beyond Vibe Coding: Building AI That Works at Enterprise Scale
Apr 09, 2026
Beyond Vibe Coding: Building AI That Works at Enterprise Scale
Enterprise AI is no longer about experimentation alone. The real challenge now is scaling it in a way that delivers measurable value, holds up operationally, and avoids descending into chaos. In this episode of The Great Indoors, Matthew Roberts is joined by Mark Austin, Vice President of Data Science and AI at AT&T, for a grounded conversation on what it really takes to industrialize AI inside a large enterprise. With generative AI now making up the majority of his organization’s work, Mark explains why governance, ROI discipline, and structured execution have become essential to moving from isolated projects to enterprise-wide impact. Together, they explore the difference between “vibe coding” and what Mark calls AI-fuel coding, why fine-tuned models may offer the rare combination of better, faster, and cheaper, and how AT&T is thinking about agents, orchestration, and AI systems that can operate at scale.
Quote from Placeholder Name
“Believe it or not, 27 billion tokens today. We're going through generative AI. But if you're just starting out, the way that we got there is we set up a nice governance and ROI. So everything comes in is graded with ROI and stack rank, and that's how we get through hundreds of projects. Two years ago, we were two ROI in here. We're five x ROI in a year now. Where we're at in terms of the challenges, it's keeping the ROI growing.”
Speaker
