I’ve spent the last two decades building and scaling operating-intensive businesses, including founding Freshly, which was acquired by Nestlé in a transaction valued at approximately $1.5 billion, and now leading Petfolk, a fast-growing veterinary clinic platform backed by more than $150 million in capital. Across those experiences, one lesson has become increasingly clear: when a new technology meaningfully changes what is possible, organizations must rethink how they operate in order to capture its full value.
Executives and boards across industries are voicing the same concern. After billions invested in artificial intelligence initiatives, many organizations report little measurable return. That frustration is real and well documented. According to a PwC Global CEO Survey, which Fortune has previously covered, 56% of companies say AI has yet to deliver either cost savings or increased revenue, and only about 12% report gains on both fronts.
Speaking with Fortune at Davos, PwC Global Chairman Mohamed Kande argued that the shortfall is not about AI’s capability, but execution, noting that many companies “forgot the basics,” including clean data, disciplined processes, and governance.
The takeaway many leaders are drawing is that AI is failing to live up to expectations.
That conclusion is wrong.
The problem is not the technology. It is how leaders are framing the opportunity and how they are measuring success.
Most companies are deploying AI through an efficiency lens. They ask where it can reduce labor, automate workflows, or deliver quick payback inside existing organizational structures. They then evaluate those efforts using traditional return-on-investment metrics designed for software tools or for headcount reductions.
That approach misunderstands what AI actually changes.
AI is not simply a better way to do the same work. It is a new economic input that collapses the marginal cost of high-quality analytical and intellectual labor. That shift has consequences that most organizations are only beginning to understand.
Synthetic Human Intelligence Hours as a New Unit of Work
Every major business transformation of the last century followed the same underlying pattern. A foundational input became dramatically cheaper, and usage expanded exponentially. During the Industrial Revolution, the falling cost of energy converted mechanical power into what were effectively cheap mechanical human hours, enabling machines to multiply physical labor at unprecedented scale. More recently, cloud computing collapsed the cost of computing, storage became effectively infinite, and digital distribution went global overnight.
AI now represents the next turn of that same economic wheel. It is driving the marginal cost of high-quality thinking toward zero.
To describe this shift clearly, it helps to name it. I call it Synthetic Human Intelligence Hours, or SHIH.
Synthetic Human Intelligence Hours are high-quality analytical and intellectual work generated by AI at near-zero marginal cost and deployable at scale. They are not artificial people. They are synthetic intelligence capacity. A new unit of productive effort.
Once you view AI through this lens, the confusion around adoption begins to make sense. Organizations are trying to force a technology that creates Synthetic Human Intelligence Hours into systems designed for scarce human attention.
That mismatch shows up clearly in the data. An MIT research report based on its 2025 State of AI in Business study, which Fortune has also covered, found that only about 5% of integrated AI pilots were delivering measurable value, while roughly 95 percent showed no tangible financial impact. The researchers describe this gap as the “GenAI Divide.”
The report goes further, explaining that most failures stem not from the models themselves, but from poor integration into real workflows, overreliance on generic tools, and a tendency for companies to treat AI as a standalone experiment rather than embedding it into core operations. The findings are based on interviews, employee surveys, and analysis of real enterprise deployments.
That statistic is often framed as evidence that AI does not work. A more accurate interpretation is that leaders are measuring the wrong thing. They are evaluating a capacity-expanding input using efficiency-based metrics.
That is a leadership error, not a technology failure.
What This Looks Like Inside a Real Business
At Petfolk, we operate across 36 veterinary clinics today and are scaling toward hundreds as part of a $150+ million backed effort to fundamentally disrupt veterinary medicine. Our regional managers are accountable for nearly every dimension of store-level performance across their regions: revenue, labor, scheduling, inventory, procurement, medical quality, compliance, patient outcomes, pricing, customer experience, team development, retention, training, and culture.
Each regional manager is effectively responsible for thousands of micro-decisions per week, informed by hundreds of reports, dashboards, audits, reviews, and operational signals. All of it ultimately rolls up to the performance of individual clinics.
Today, a strong regional manager might spend forty to fifty hours a week reviewing reports, identifying issues, and supporting clinic leaders. Even with excellent analysts, the work is constrained by time. Tradeoffs are inevitable. You sample data instead of examining everything. You go deep in some areas and shallow in others.
Our goal over the next year is to fundamentally break that constraint.
We are building AI agents to generate Synthetic Human Intelligence Hours alongside our regional managers. The ambition is simple and radical. We want to turn a 40- to 50-hour human workweek into the equivalent of a 500-hour analytical workweek without asking the human to work any more.
The regional manager still works forty hours. The remaining 460 hours are SHIH.
Those agents will review every invoice, every schedule, and every inventory decision. They will analyze every NPS score, eNPS score, Google review, performance metric, and more. They will compare results not just week over week, but across time horizons, cohorts, and locations. They will work through our entire learning and development library to generate bespoke development plans for individual team members.
All of that intelligence is synthesized and delivered to the regional manager. The human decides what matters. The human prioritizes. The human communicates and leads.
Functionally, the role changes. A regional manager is no longer operating with the analytical bandwidth of one person. They are operating with what would previously have required an entire team of analysts.
We would never have attempted this in the past. Not because it was not valuable, but because it was economically impossible. The cost of human analysis made it unscalable.
AI changes that equation.
Why ROI Misses the Point Early
One reason many leaders become disappointed with AI is that changes like these do not show up cleanly or immediately in financial results.
Turning on Synthetic Human Intelligence Hours does not instantly reduce costs. It does not automatically increase revenue the week it is deployed. In the early stages, the gains are subtle. Decisions get slightly better. Patterns are caught earlier. Teams improve incrementally. Waste is reduced quietly rather than dramatically.
This is not a flaw. It is the nature of compounding systems.
Returns from intelligence capacity expansion accumulate over time. Just like any compounding effect, they look small at first and almost invisible in isolation. But over long horizons, they dominate outcomes.
Organizations that evaluate AI exclusively on short-term efficiency metrics will miss this entirely. Organizations that understand SHIH as a compounding advantage will design for durability rather than immediate optics.
That disconnect helps explain why PwC also found CEO confidence in revenue growth at a five-year low. Weak AI returns are feeding broader strategic uncertainty, not because the tools lack power, but because organizations have not yet redesigned around them.
The benefits show up not as a single line item, but as better decisions repeated thousands of times.
The Question That Matters Now
As the marginal cost of thinking collapses, the scope of what organizations can afford to analyze expands dramatically. The competitive divide will not be between companies that automate faster and those that do not.
It will be between companies that continue to think in terms of efficiency and those that redesign around capacity and compounding advantage.
AI will not replace humans. It will redefine what small, focused teams are capable of accomplishing.
The question leaders should be asking now is not where they can cut costs.
It is: If high-quality thinking were almost free, how many Synthetic Human Intelligence Hours would you deploy, and what problems would you finally take on?
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.
