On June 25, a small group of senior engineering leaders gathered at Merchant & Trade in Uptown Charlotte, representing hospitality, retail, and a few other industries that are running AI at real scale. We started on familiar ground, pipeline governance, build vs. buy. Then someone asked the question that ran the table for the next two hours: if AI cost and consumption governance becomes part of SDLC governance, who actually owns it?
The ownership question: nobody has an org chart for
Platform engineering? FinOps? Something that doesn't exist yet at most companies? Eventually, the consensus was that product owners should hold this long-term. But almost everyone agreed we're not there yet, and in the meantime, FinOps teams are driving most of the adoption and action.
What made this land was a comparison to cloud infrastructure a decade ago. Nobody worried about cloud spend until the bills got big enough to force the conversation. AI cost is tracking the same curve, just faster. A few leaders raised the security angle, too: cost anomalies are starting to function as a lagging indicator of compromise. A spike in compute or storage usage isn't just a FinOps problem anymore; it might mean a bad actor is exfiltrating data or mining crypto on your account.
Build vs. buy, live at the table
One of the better moments of the night wasn't planned. Two leaders with opposite philosophies, one running a build-your-own shop that's actually worked, the other firmly in the buy-best-of-breed camp, debated it directly. Neither side backed down, and nobody seemed to think either one was wrong, a reminder that "what's right" still depends on what your team is actually good at.
The part that's actually hard
Visibility alone isn't solving anything. One leader shared a story that stuck with me: a team's cloud spend looked flat for 12 months. Then an anomaly alert fired. Their first reaction was "we don't run AI services in our account." Forty-eight hours of triage later, they found a former colleague had spun up an AI service nearly two years earlier and left it running.
That's happening more than people want to admit, and it's part of why most of the table agreed humans in the loop aren't optional, even as AI automates the toilsome work. There was also shared whiplash over how fast billing models changed. Plenty of leaders got caught off guard when major AI providers moved from seat-based licensing to consumption-based pricing, scrambling their forecasts of finance and engineering budgets.
Underneath it all is one question executives keep asking, with no clear answer yet: What am I actually getting for my AI spend?
Where this leaves us
Nobody claimed to have solved this. What came through clearly is that AI cost governance is no longer a side conversation; it's becoming inseparable from SDLC governance itself, and the organizations that figure out ownership now will be far ahead of those still treating it as a budget line to revisit later.
We're hosting more of these throughout the year. Hope to see you at the next one.