According to the now well publicized MIT-affiliated research and reporting, fewer than 5% of enterprise AI pilots ever deliver measurable business value. The other 95%? They’re still stuck in what we call AI Purgatory: exciting demos, scattered pilots, and vanishing trust.
For the past couple of years, corporate leaders have been sprinting into the AI era with a mixture of urgency and anxiety. Boards are calling. Investors are asking. Competitors are announcing. And CIOs, under immense pressure, have responded the only way they knew how, by launching pilots. Lots of them.
Across company after company we see the same pattern: 30, 50, even hundreds of AI pilots scattered across functions and owned by individual enthusiasts rather than enterprise leaders. Some organizations publicly reference numbers in the hundreds; one global healthcare company announced more than 900. It gives the illusion of momentum. It produces exciting demos. It certainly calms a board, at least temporarily. But it doesn’t create value.
This is AI Purgatory with exciting activity but no real escape path to transformation. The danger isn’t just wasted spend. It’s the erosion of trust: with the board, with the C-suite, with the workforce that’s already skeptical and fearful about the future.
But the best companies, the ones actually seeing measurable impact, are doing something radically different. They’re shifting from Pilot Mania to Portfolio Discipline.
The Pilot Trap: How Good Intentions Create Bad Strategy
Let’s be honest: pilot mania didn’t happen because leaders were undisciplined. It happened because they were scared: Scared of missing the wave. Scared of looking slow to their boards. Scared of making a “big bet” without understanding what’s hype and what’s real.
The safest move was to hedge. So they piloted. But pilots carry hidden costs:
- They fragment attention – Every pilot needs a sponsor, a team, a dataset, an evaluation cycle. The more you run, the more you dilute the talent and focus required to deliver real outcomes.
- They break trust – When employees see a dozen AI tools briefly appear and quietly die, confidence drops. “See? AI doesn’t work here.” Each failed pilot becomes evidence that leadership is experimenting haphazardly rather than investing in well thought out strategies.
- They normalize chaos – Everyone experiments on their own. Governance and data readiness fall behind. Suddenly, you’ve created shadow AI everywhere. Trust erodes not just in technology, but in leadership’s ability to steer it responsibly.
- They create the illusion of progress without creating impact – Demos shine; dashboards stay flat.
What the Best Companies Are Doing Instead
Across dozens of enterprises, from healthcare to finance to retail, we’re seeing a new pattern emerge. Some tech leaders are now taking an approach that looks far more like private equity than like traditional IT. They are building a disciplined portfolio of a few high-impact bets. Not 40. Not 25. Usually three to five.
Eaton provides a clear example of this shift. Rather than running dozens of disconnected pilots, the company deliberately narrowed its focus to a handful of initiatives that mattered to the business and held them to measurable outcomes. As Katrina Redmond, Eaton’s Chief Information Officer, explains, “We’ve shifted from proof-of-concept thinking to prioritizing a handful of high-impact initiatives that deliver measurable value and strengthen business trust in the technology.”
What Eaton has done reflects a broader pattern we now see across the companies that are actually escaping AI purgatory. In practice, those organizations tend to converge on the same four operating disciplines:
1. They are tied to a business challenge the CEO already cares about.
Not “AI strategy.” Business strategy. The breakthrough companies start with a CEO-level mandate: Reduce cycle time by 20%, Free 10% capacity for frontline teams, Improve customer resolution by 15 points. AI shows up as a tool inside a business outcome, not an experiment in search of meaning. This clarity does more than focus investment. It restores trust with the board by anchoring AI to outcomes they already measure and care about.
As Marianne Johnson, Chief Product Officer of Cox Automotive, puts it, “We learned early that twenty pilots do not equal one transformation. At Cox Automotive, we focused our AI investments on solving real pain points for dealers and consumers, reducing cycle times, improving customer experiences, and creating measurable competitive advantages. That discipline is why we now have 20 AI solutions in production delivering measurable value, and we’re just getting started.”
2. They have cross-functional ownership, not isolated champions.
In companies that break out of AI purgatory, the CIO doesn’t own the work alone. The CFO, CHRO, business-unit heads, operations, and data leaders take responsibility for the same outcomes. Each high-impact bet has a single cross-functional leadership group accountable for its success. That alignment eliminates the biggest blockers: stalled decisions, unclear ownership, and competing priorities. Just as importantly, it rebuilds trust across the enterprise that AI is being governed effectively rather than improvised.
As the Chief Technology and Operations Officer at Travelers, Mojgan Lefebvre, put it, “Technology doesn’t transform companies, people do. We structured our AI deployment with cross-functional ownership from day one. When engineering, data science, product, and business leaders are all accountable for the same outcomes, you eliminate the biggest blockers to real transformation.”
This model forces the hard conversations early: who owns risk, who owns workflow redesign, who owns workforce impact, and who owns results. Instead of AI initiatives getting stuck between functions, accountability is shared and progress accelerates.
3. They run “prove-and-scale” waves, not demos.
The winners don’t celebrate pilots. They celebrate value proven at small scale and velocity of expansion. A typical wave looks like:
- 30 days: prove value in a controlled slice (one region, one product, one workflow)
- 60 days: scale to broader teams
- 90 days: integrate into enterprise workflows
Cisco provides a clear example of how this discipline works in practice. Cisco’s People, Policy, and Purpose (3P) organization piloted its Working with AI program in just four weeks with five cross-functional teams. Within the following six weeks, the program was scaled across the broader 3P organization, after two months it has been extended to the Product Development organization, and ultimately will be rolled out across the enterprise level. “We reviewed 24 workflows so far, with on average 30% of their activities augmented by AI,” shared Gianpaolo Barozzi, 3P CTO.
The metric isn’t “How many pilots?”. It’s “How many workflows changed?”, “How many hours returned to the business?”, “How quickly can we expand?”. These metrics matter because they demonstrate, visibly and credibly, that AI can be trusted to deliver.
4. They create a “portfolio discipline” and kill 80% of ideas.
The most effective leaders adopt a ruthless mantra: If it’s not tied to strategy, it doesn’t get funded. This is a hard shift for CIOs who have historically been rewarded for experimentation. But in the AI era, discipline is the new courage. And the discipline pays off. Companies that narrow focus see: faster scaling, higher adoption, greater trust, more predictable ROI, and a far more confident board.
At Johnson & Johnson, portfolio discipline starts with the business problem, not the technology. Leaders deliberately narrow AI investment to a small set of use cases tied to strategy because a handful of initiatives consistently deliver the vast majority of impact. According to Jim Swanson, Johnson & Johnson’s EVP & CIO, “Instead of spreading resources across dozens of ideas, we prioritize the few AI initiatives that can deliver step-change outcomes. At Johnson & Johnson, the top 10–15% of initiatives generate roughly 80% of the impact—and that clarity is what allows us to scale faster, build trust, and deliver results that matter.”
A similar lesson has emerged at Liberty Mutual Insurance. As Global CIO Monica Caldas notes: “That shift to disciplined, business-led execution is already playing out at Liberty Mutual Insurance. We learned quickly that success with AI isn’t about volume — it’s about intentionality. By focusing on a disciplined portfolio of business-led, enterprise scale priorities, we’ve accelerated both value creation and trust across the organization. When organizations align around a small number of high impact outcomes, AI becomes a force multiplier.”
Case in Point: Escaping AI Purgatory in 6 Months
Across organizations we coach, we’ve seen a repeatable pattern across those that are now leading the way in AI transformation:
- Scrap 80% of pilots.
- Choose 3–5 “bet-the-business” use cases.
- Form a cross-functional leadership squad.
- Run 90-day prove-and-scale sprints.
- Track both leading metrics (usage, cycle time, speed to pilot) and lagging ones (value, cost reduction, satisfaction).
- Deliver visible business outcomes.
Momentum becomes self-reinforcing. The culture shifts from “AI is confusing” to “AI is how we work now.”
The Bottom Line
We are exiting the “Look, we’re doing AI!” stage of this era. The next 12 months will be defined by a simple divide: Companies with 100 pilots and no results. And companies with a few high-impact bets and transformational outcomes.
In the end, escaping pilot purgatory is less about technology and more about discipline. Clear priorities. Focused leadership attention. A willingness to stop doing things that don’t matter. When organizations make that shift, the work stops feeling chaotic and starts feeling intentional again. Not because the technology changed, but because the leaders did.
Sometimes progress begins not with a breakthrough, but with the courage to simplify.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
