In January 2026, PwC’s 29th Annual Global CEO Survey polled more than 4,400 CEOs across 95 countries about AI. Fifty-six percent reported zero financial benefit from AI in 2025. Neither higher revenues nor lower costs.
Anyone surprised? I’m not.
In my last post, I made the case that AI adoption falls short because of organizational gaps, not technology gaps. Thinking back to computer-assisted ordering (CAO), ERP and ecommerce, I see similar patterns decades apart. I also recall that progress in CAO was measured in years and decades. But today, that kind of delay won’t work.
A lot of the urgency comes from the fact that in grocery, 1.7% net is the norm. Two pandemic years briefly changed that. But now we’re back. At 1.7%, there’s no room for operational slippage and therefore no time to be late. If a competitor gains ground, the lead widens and compounds. Going back to that same PwC survey, it did find that 56% of CEOs reported zero AI benefit. However it also found that 30% reported revenue gains. So that’s one in three stacking up early wins, and they’re the ones pulling away.
Microsoft’s 2026 Work Trend Index, published this month, confirms why. Organizational factors — culture, management, talent practices — drive more than twice the AI impact of individual effort alone. The gap between organizations that get this right and those that don’t is already widening. This business doesn’t provide much room to catch up.
I’ve spent a career closing gaps like this deliberately, and the first thing I want to understand, in any company, is where the work actually happens. Not where it’s supposed to happen. But rather, where it gets done — the floor, the distribution center, the route, the storefront. Often, this isn’t precisely mapped on the org chart.
Take Ready Pac Foods. When I arrived, the COO met me on Day One with this bombshell, “We might not survive the winter.” He wasn’t off base because the company had a fragile infrastructure, no clear productivity plan, five years of frozen wages and a PE owner watching the clock. For a CEO focused on transformation, the quick answer might have been to shut something down or prune SG&A costs in order to show the board something fast.
But my approach was different and involved pay increases for frontline workers, communications in multiple languages and strategic leadership hiring. I also had to break the news to the PE board that the strategy that got the company into this mess would not be the one that got us out. Technology wouldn’t be the solution either. We needed a new course.
It worked. Ready Pac grew by more than 60%, and Bonduelle acquired the company a few years later. The sequence worked because we understood the people and the work before we reached for anything else. No amount of technology was ever going to save that company. The people had to come first. From there, we could start updating the tech stack of the day to get more efficient and productive.
The goal, always and forever, is to understand which processes are working and which ones aren’t … and develop new processes to strengthen execution. That’s not a technology question but rather leadership and team. Until you can answer it, there’s no way to know where something like AI fits. Tech isn’t a fix-all for broken processes. AI implementation built on top of unresolved process problems won’t produce results. It’s a sure bet for enterprise-grade slop.
That’s the pattern playing out in AI right now. Back-of-house deployments — warehouse automation, cold storage computer vision — are scaling where process discipline already exists. Front-of-house is where the costly write-downs have landed.
I keep coming back to the third part of my eight-word grand summation from years ago — know yourself, know the world, act on instinct. The third part means full send, 0 to 90 with not a second to lose. We don’t get to act on instinct slowly. The whole point of doing the first two parts well is it gets you ready to act with the confidence of a prepared mind.
With any new tech and especially AI, the objective is to make it visible and accountable before you ask people to trust it. And when you’ve done that, move fast and keep learning.
The companies that figure out early AI wins won’t announce it. They’ll just start outperforming. Remember, 30% of PwC’s survey group reported positive results. It shows that the window is open. It just won’t stay that way forever.