The leadership blind spot with AI
A manufacturing CEO showed me his new AI system earlier this year. He’d spent $250,000 on a predictive maintenance platform that promised to cut downtime in half. When I asked how he’d assessed whether his organization was ready to use it, he looked at me like it was a strange question. “The vendor showed us case studies where it worked.”
Six months later the system sat unused. The maintenance team didn’t trust it, the underlying data quality problems were never fixed, and the floor was still running on the same processes it always had. The dashboard looked great in the quarterly meeting. Nobody was using it to make a decision.
I’ve watched some version of this play out across a lot of industries. Executives buy AI before asking whether their organization can actually absorb it, and then they’re surprised when the value never shows up. A 2025 MIT study found that 95 percent of corporate generative AI pilots are delivering no measurable return . The gap between what AI promises and what it delivers is rarely about the technology. It’s almost always about readiness, and readiness is the part nobody’s looking at.
I’ve sat through hundreds of vendor presentations, and they all run the same way: impressive case studies, big ROI numbers, a slick demo where everything just works. The trouble is that vendors are selling you the last chapter of a book you haven’t started writing. The results they show come from organizations that already did the unglamorous work, the data cleanup, the process changes, the training, the governance, before the AI ever went in. The sales deck skips all of that, because none of it photographs well.
So executives end up choosing tools based on a demo, leaning on the same business intuition that’s served them for years. That intuition has real blind spots here. You can’t see your own data quality or integration debt until it derails a project halfway through. The vendor’s case studies came from companies that had already solved the foundational problems, so they tell you nothing about the work you still have ahead of you. The demo ran on clean data in a controlled setting that doesn’t resemble your operation. And first-time implementations almost always take longer than anyone budgets, because the readiness gaps only surface once you’re in. You don’t need to become technical to make good AI decisions. You need a way to see the things your intuition can’t.
What readiness actually means
After enough of these, I started running leadership teams through a simple readiness check before they look at a single vendor. It comes down to three questions.
The first is about data. How clean, complete, and connected is the data this AI would run on? Can it move between your systems without someone exporting spreadsheets, and do you actually know who owns it and whether you’re allowed to use it the way you’re planning to? Most organizations rate their data higher than it deserves. Be honest here, because the model can only ever be as good as what you feed it.
The second is about process. Is there a specific decision the AI is meant to inform, and does someone own that decision? Can your current workflow actually bend to act on what the system tells you, or will the insight land on a desk where nothing changes? An AI recommendation nobody’s set up to act on is just an expensive notification.
The third is about people. Does your team have the skills to run and troubleshoot this once the vendor is gone? How much change is the organization already absorbing right now? And is there enough trust in data-driven decisions that people will actually use the thing, or will they quietly route around it the way that maintenance team did? If you’re weak in any of these, that isn’t a reason to abandon AI. It’s a list of what to fix first.
A healthcare provider I worked with wanted an AI system to optimize patient flow. Their instinct was to start by comparing vendors on features and claimed ROI. Instead we ran the readiness check. Their data was a real problem, patient information spread across three systems that didn’t talk to each other, with big historical gaps. Their processes were workable but rigid. Their people were the bright spot, technically capable and open to it.
So they changed the order of operations. Before touching a vendor, they connected their systems, ran a focused data-quality effort on the specific inputs the model would need, and reworked the patient-flow process so it could actually use a recommendation. About six months later they implemented the AI, and it delivered close to what the vendors had promised, because by then the foundation was there to deliver on.
That’s the whole move, and it’s less exciting than it sounds. Start from the business problem, not the tool. Assess your readiness honestly, even when the answer is uncomfortable. Be willing to spend a few months building foundations before you buy anything, then start small and learn from a limited rollout. Measure business outcomes rather than technical milestones. It feels slower. In practice it’s faster, because almost every organization I’ve watched invest in readiness first got to real value quicker than the ones who bought the platform and tried to sort out the foundations later.
As AI works its way deeper into every industry, the executives who come out ahead won’t be the ones who bought the most advanced tools. They’ll be the ones who knew what their organization could actually use, and built what was missing before they spent the money. That kind of judgment never shows up in a demo. It’s the part of the job that’s still yours.