There’s been a shift in the tone and urgency of the CEO conversations around AI lately. It’s a confusing and fraught time, and no leader wants to fall behind competitors in the race to transform their businesses. Boards and investors are taking the disruptive power of AI seriously, as evidenced by recent market volatility, and expectations are rising.
PwC’s latest global CEO survey found that 43% of respondents see either increased revenue or lower costs from their AI initiatives. But just as many (42%) say their companies are “stuck” and unable to unlock either revenue or cost benefits.
Why are so many companies seeing disappointing results despite putting so much emphasis on AI? To be blunt, many corporate leaders have yet to achieve clarity on AI. There’s some hype, a lot of reality, and a ton of confusion about AI as a catalyst for business transformation. The real signal—where AI will create value—is often buried under the noise. That noise often comes from leaders overestimating their actual progress in a few common ways.
A closer look at what’s driving the confusion
Overstatements are more than harmless hype. They can skew strategy, misallocate investment, and set expectations that can’t be met.
First, some CEOs assume the AI models are fit for all purposes. They and their teams make investment decisions based on these assumptions and then they’re surprised when real-world performance fails to live up to expectations. In the real world, AI models are still maturing. They’re great at some tasks but need refinements at others. Leaders need to know when and where a model is ready for building industrial-strength capabilities and, perhaps more important, where it’s not.
Second, some boast about the sheer number of AI agents they’ve “deployed.” They assume more is better when, in reality, volume often signals experimentation, not scale. I tell clients that they need to zero-in on value and embrace what we call “bi-modal transformation.”
Mode one focuses on value over volume. Mode two is more experimentation-based and is led by workers, not senior leadership, so a higher volume is acceptable. Mature adoption is mode one, and it looks like focus: once strategy reveals the highest-value AI impact zones, companies should double down on a set of agents that can reliably move the needle and retire the agents that don’t. A company touting a huge agent count may be inadvertently admitting it hasn’t yet identified the shortlist of specialized, proven agents that are capable of consistently achieving the desired results.
Third, some CEOs may think that their company has already “transformed,” often by extrapolating from early wins. They take new capabilities, multiply them by the number of agents in flight, and conclude they’ve become an intelligent enterprise—a more touchless, intelligent capability system that amplifies differentiators, creates some new ones, and drives continuously improving performance.
Transformation isn’t additive. It’s systemic. It requires deliberate integration—connecting data, workflows, controls, and operating model choices so AI is deployed coherently, not scattered across dozens of unrelated programs whose theoretical outcomes overstate practical impact. An intelligent enterprise is a scaled foundation where data and AI work as one. Systems become agile, insight-driven platforms that don’t just cut costs but create competitive advantage.
An intelligent enterprise is a scaled foundation where data and AI work as one. Systems become agile, insight-driven platforms that don’t just cut costs but create competitive advantage.
True transformation also demands hard leadership decisions. In addition to reaching what the industry considers table stakes, CEOs should choose where to place their biggest bets to try to gain advantage. This goes back to the bi-modal approach. Some initiatives should move faster because they receive disproportionate resources, talent, and executive attention. The job is to decide on initiatives that are most likely to differentiate the business and invest accordingly—work that is far harder, and far more valuable, rather than congratulating the organization for having many uncoordinated, bottom-up experiments in motion.
These three misconceptions fuel much of today’s AI hype. Leaders treat the most flattering signals (provider benchmarks, agent counts, and theoretical “capabilities”) as evidence of real transformation. That comforting narrative can crowd out harder truths about what’s actually working in production, leading to diluted focus, misallocated investment, and missed opportunities for durable, AI-driven enterprise improvement.
The next phase of enterprise AI adoption: challenges for CEOs
The self-delusion—we did it, let’s rest on our laurels—is one reason why we’re seeing the letdown at some organizations.
The initial excitement and euphoria wears off as companies receive a reality check from the mixed results produced by all the technological hype. Many companies that bought AI products based on potential are becoming wary. They see employees not using the tech due to fear, lack of trust, or insufficient training or incentives.
These mixed results, and the vibe shift they’re producing, risk derailing the momentum at early adopters. This is Amara’s Law in action: Most people overestimate technology’s impact in the short term and underestimate it in the long term. Disappointment is an understandable and predictable part of the learning process, but it’s not a reason to cut bait.
Procrastination is creating a similar problem at other companies that haven’t seriously started their AI journey. We can see this in our industry benchmarks. Each industry now has an average spend on AI as a percentage of revenue. The laggards are well below the mean, and it shows up in their financial performance relative to their peers.
Overwhelmed with the day-to-day challenges of running the business, these C-suites haven’t encouraged or kept track of AI experiments at their firms. Now, some may fear they’re so far behind that they can’t catch up. Paralyzing psychological forces have placed them in a holding pattern at the very time competitors are testing their ability to scale the most successful AI experiments.
In our view, early adopters do have an advantage, but they risk losing their lead if AI fatigue over disappointment that ‘singularity hasn’t yet arrived!' sets in. And AI laggards probably aren’t as far behind as they think. In some cases, they may even be able to learn implementation lessons from other organizations without the associated costs.
C-suite leaders who move quickly and decisively still have time to take advantage of the opportunity before their competitors do. The investor community is looking for signals from management that their strategy adjustments are sound, funded, and programmatically executed.
AI adoption questions... and how to get answers
There are plenty of client success stories to share. A global hospitality brand reduced time-to-hire by almost 75% while doubling applications through AI-enabled workforce transformation. An EV manufacturer built an AI-native finance function that shrinks forecasting from weeks to minutes. One of the world’s largest retail companies reduced software development cycle time by 60%.
CEOs are excited to hear about proven results. They’re asking how they can apply what’s working—how to move the needle.
How do we pick where to make our big AI bets?
How do we organize to improve the chances of success around outcomes?
What are the most meaningful metrics to use to measure progress in our AI journey?
How can we get employee buy-in for sustained AI use instead of just ad hoc experimentation?
How can we get from isolated AI experiments to an intelligent enterprise that is more touchless and efficient?
We have reached a critical point where we finally have enough data and feedback to glean lessons from all the early experiments and success stories. Similar to previous transformative advances like the internet and railroads, AI adoption started with a “wild west” phase where mostly uncoordinated pockets of boosters tested the abilities and limits of its business applications. Even when these AI champions had C-suite backing and the budgets that came with it, they were working without a playbook.
Similar to previous transformative advances like the internet and railroads, AI adoption started with a “wild west” phase where mostly uncoordinated pockets of boosters tested the abilities and limits of its business applications.
We’ve learned helpful rules of thumb from those experiments that we can apply in the next phase of tech adoption and business change. AI clearly excels at some tasks, including reading, derivative writing (taking notes, writing summaries), coding, researching and reviewing. But it can struggle with math, complex planning, innovation, original writing, deception, drift, and task length.
These challenges are usually manageable when overseen by people with the right skills. Now it’s up to leadership to figure out how to turn those tactical wins into strategic P&L victories.
This is where your C-suite can make a major difference.
Setting a strategy that guides the AI journey and gives your teams confidence in picking their spots.
Getting the best engineering talent, combined with their deepest domain experts, guided by actively engaged leadership.
Leaning in and owning the AI-enabled change agenda by helping teams get over the hump and into value realization.
Determining which AI products or business partnerships make the most strategic sense for those investments.
Designing processes to scale implementation throughout the enterprise.
Identifying and training the talent needed to execute and adopt the new technology and ways of working.
On the (bench)mark, get set, go!
There are no standard answers to most of the questions that CEOs have about AI. There are too many variables—industry, company size, company culture—for a cookie-cutter approach. Instead, I tell CEOs that to move forward they should use benchmarks to inform their strategy. Benchmarking will help leaders establish proper allocation of their specific organization’s budgets, people, and data.
CEOs should set high standards in the benchmarking process and push back against one-size-fits-all metrics.
CEOs should set high standards in the benchmarking process and push back against one-size-fits-all metrics. Remember the companies trying to measure AI progress by the number of agents that they have deployed? That’s akin to an F1 team trying to win a race by fielding more cars than the competition. But F1 teams don’t achieve victory by having the most cars. They win by having the most capable car, a driver who knows how to win with it, and a team that makes it all come together. Benchmarks signal what the new formula for winning is. CEOs need to decide what that means for their business.
Leaders approach this challenge by assessing the performance of the area they’re looking to transform through AI, including factors like cost, customer service metrics, and workload. They set a goal against that baseline so that their teams feel a sense of urgency and adopt a pace reflecting that urgency. Instead of uncoordinated experiments, teams can now pursue a disciplined march toward value realization.
The bottom line
In today’s fast-moving AI landscape, this is the moment when AI shifts from technological curiosity to enterprise accountability. Disruption requires hard choices. Where are you differentiated and where do you have an opportunity to lead? Where are you going to let your competitors figure things out and tolerate a lag while they do? And where might you exit certain parts of your business because AI will create such a gap that the investment isn’t worth making.
Strategy and hard choices can get you to the right answers. Ultimately, our job as leaders is to set the strategy and guide enterprise decisions in spite of uncertainty and constraints. Doing nothing is still a choice—even if it feels like the default—and the price of inaction is increasing rapidly. The companies that win in the AI era are not those running the most experiments. They are the ones making the clearest strategic choices.
Contact us
