CDO Vision LA 2026 | Transforming Industries with AI: Enterprise Success Stories

AI conversations have evolved. Earlier, leaders were asked if they were investing—now, they’re being asked what those investments are actually producing. At CDO Vision LA 2026, AI and Data leaders who live inside that pressure every day sat down to talk honestly about what's working, what isn't, and what it actually takes to move AI from a promising initiative into something that runs at enterprise scale.

This set the tone for the panel Transforming Industries with AI: Enterprise Success Stories at CDO Vision LA 2026, moderated by Hoang Richter, Region Lead USA, Digital & AI Innovation at BSH Home Appliances Group, featuring Mumtaz Vauhkonen, Senior Director, Global AI/ML and Data at Skyworks Solutions; Michael Mayes, Head of Transformation, Insurance & Retirement at Apollo Global Management; and Nathan Wan, VP, AI at Ensemble Health Partners, who shared how AI is being embedded into core business functions to drive real outcomes.

The conversation covered three areas where most organizations are still getting it wrong: how they measure success, how they approach scale, and how they handle the people caught in the middle of it all.

Stop Tracking the Wrong Numbers

Michael was direct about a problem he sees constantly. Companies celebrate metrics like "models run in 2 hours instead of 24" or "queries resolved automatically." Those numbers feel good. But they don't tell you if the business improved. "I don't care how fast the models go. It's what can I do with that speed. Am I able to make a decision faster, deploy capital more quickly, unlock real value?"

The metrics that matter are business outcomes. Faster time to market. Lower cost per transaction. Revenue impact. Hoang added a consumer-facing dimension: every dollar embedded in a product has to translate into something a customer will pay for. If it doesn't, scaling only makes the problem bigger.

Michael also pushed leaders to look honestly at how AI budgets are split. How much goes toward keeping existing systems running? How much toward new capabilities? How much toward genuine bets on what comes next? In a field moving this fast, spending most of your AI budget on the first bucket is a risk in itself.

Scale in Slices, Not in One Shot

There's an old engineering rule: the last 10% of a project takes 90% of the effort. In AI, the panel argued, it's harder than that. Customers don't remember the 99% of interactions that go right. They remember the one that didn't.

Mumtaz described the approach that works: build one component, ship it, stress-test it at scale, then move to the next. Productionize as you go, rather than waiting until the whole system is built to find out what breaks. Michael echoed this from a data perspective. Scope the problem down, fix what you need for that slice first, get it into production, then move to the next. "Everyone says there's a huge data problem," he said. "Break it down."

The panel agreed on one more thing: don't lock yourself into one model, one vendor, or one architecture. Mumtaz noted that giving teams the freedom to try different tools, then course-correcting when needed, is more effective than mandating a single platform from the top. Flexibility costs more upfront. But it's cheaper than rebuilding when the next shift arrives.

The People Problem

Technology is the easier half. The panel spent more time on the harder part: getting people to actually change how they work.

Michael introduced one of the most practical ideas of the session. Before rewriting human job descriptions, write job descriptions for your AI agents. What does the agent own? What are the expected outputs? Until those questions have clear answers, you're not really deploying AI at enterprise scale. And until human roles are updated to reflect what people do once agents handle the rest, the workforce picture stays unresolved.

Mumtaz addressed the reskill versus hire question directly. You need both. But you need a small core of people who understand AI deeply enough to translate it for every function in the business. That work is ongoing. Staying current and keeping the broader organization up to speed, he said, takes 15 to 20% of his working week. That's not overhead. That's what adoption actually costs.

Nathan added the dimension most leaders underestimate. In many industries, AI doesn't just change processes. It directly threatens the roles of the people doing that work. Assuming enthusiasm from leadership will trickle down naturally is a mistake. Being clear about what AI is for, what it isn't replacing, and what the new role looks like is what determines whether you get adoption or pushback.

On governance, the panel reframed it. In regulated industries especially, governance isn't what slows you down. It's what lets you move fast within clear limits. When leadership agrees on what's off the table, teams stop escalating every edge case and start making decisions.

The organizations pulling ahead right now are not the ones with the most detailed AI roadmaps. They're the ones that ship, measure, learn, and adjust faster than the market moves around them. They build flexibly because they know they'll have to adapt. They invest in their people because adoption is the multiplier. And they hold themselves to business outcomes because that's the only number that holds up in a board meeting.

The next few years won't be defined by who had the boldest vision for AI. They'll be defined by who built the systems, teams, and habits to actually deliver on it.