AI pilots multiplied into 300 use cases per company, and Brett Greenstein calls it a drag

AI adoption has become mandatory in practice, even if it is still optional on paper. Executives and employees have heard the same refrain: AI is part of your job now, and embracing it is the move. But at Fortune Brainstorm Tech in Aspen, the group’s core problem was blunt. The technology is not the bottleneck. The operating system around it is.

Brett Greenstein, chief AI officer at consultancy firm West Monroe, captured the mismatch in a way that should make any exec pause. He said an executive typically knows “there are three things that will move the needle for their business - not 300 things.” Yet when we ask teams how many AI use cases they have, the answer is often “300,” and Greenstein’s point was that they are not all equally important. That’s the spark behind the “drowning in AI” theme: too many concurrent pilots can create an internal drag, where effort and attention get scattered across initiatives that either never finish or do not connect to strategic outcomes.

To understand why this happens, look at incentives inside most large organizations. When AI pilots start multiplying, each one tends to attract its own champion, its own team, and its own set of KPIs. That structure is rational. People want to measure progress, and every project needs a scoreboard. But that same structure can slow the company’s ability to move quickly, because resources get tied up in execution cycles for too many bets. Greenstein’s framing is not anti-AI. It is pro-finding the few initiatives that truly “move the needle,” and then managing everything else like a disciplined portfolio.

Sean Bruich, SVP and CTO of pharmaceutical giant Amgen, brought the same idea into sharper focus by describing the organizational challenge as much more than technical. He said some companies may need to focus less on the technical dimensions of AI and more on the business and political challenges of implementing it. In his view, the explosion of pilots becomes a drag precisely because each pilot comes with a champion, a team, and a data engineering squad, plus KPIs that make it easy to keep going without being sure it is delivering business value. He also warned that some firms are too slow to kill pilots that are not producing outcomes. In other words, the cost is not just the time spent starting pilots. It is the momentum spent refusing to stop them.

Dan Gill added an operator’s instinct for what “value” really looks like when you are trying to ship. As chief product officer at Carvana, a logistics-heavy firm that arranges vehicle purchases between millions of buyers and sellers and runs a massive financing operation, Gill argued that finishing matters more than partially advancing. He said, “Getting one thing all the way done is much more valuable than five things progressed to 20% each.” His reasoning ties directly to why AI seems to accelerate early activity. In his words, “prototyping is cheap, and documentation is cheap, and code generation is cheap.” That affordability makes it easy to do things quickly. But the tougher, more business-critical work is getting one thing fully done and iterating tightly, not spreading attention across many nearly-finished efforts.

The panelists were also clear that this problem is not solved by asking engineers to prioritize harder, because the real lever sits at the top. Nizar Trigui, chief technology officer at logistics giant GXO, described the needed shift as business-led transformation. He said the work is no longer “just a technology-led transformation,” and emphasized that it must be led from the top: “from the board,” “from the CEO,” and across the entire executive team. This matters because boards and senior leaders control the allocation of budgets, the adoption of metrics, and the willingness to change course. If AI initiatives can grow endlessly at the team level without executive-level prioritization, the company ends up with lots of motion and uncertain outcomes.

This is easier said than done, the group noted, because there is tension between adopting AI efficiently and the metrics that have historically signaled success. Traditional indicators like headcount size, spreadsheets, and KPIs can unintentionally reward more pilots rather than the few that deliver measurable business results. Even so, the consensus was that leaders do not have the luxury of treating AI as an experiment forever. Gill framed it as irreversible: “I think we’ve crossed the Rubicon … The ships are burnt.” He added, “There is no going back, but the future is bright.” Taken together, the warning is about execution discipline, not AI enthusiasm. For peers, the strategic stakes are straightforward: if your company has 300 use cases and no clear hierarchy of what moves revenue, cost, service, or risk, you may be burning internal resources while competitors with tighter focus convert AI into business value faster.