Why 56 Percent of CEOs See No Return From AI, and What the Other 12 Percent Do Differently

For the past three years, the enterprise conversation about AI has mostly been about adoption: how many pilots are running, how many licences are deployed, how many staff have access to a copilot. That conversation is now colliding with a harder one, about returns. PwC’s 29th Global CEO Survey, drawing on responses from 4,454 CEOs, found that 56 percent report zero measurable revenue or cost benefit from their AI investment, while only 12 percent have achieved both (PwC, 2026 Global CEO Survey). For an initiative that has absorbed a significant share of technology budgets for several years running, that is not a rounding error. It is a credibility problem, and it is landing on the desks of CIOs and CTOs whether or not they set the original expectations.

The gap between deployment and value

The scale of adoption makes the flat returns more striking, not less. Separate research from Writer, surveying executives and employees using AI at work, found that 97 percent of executives had deployed AI agents in the past year, yet only 29 percent reported significant return on that investment (Writer, 2026 enterprise AI adoption survey). Fifty nine percent of companies are now spending at least a million dollars annually on AI technology. Deployment, in other words, has stopped being the hard part. Extracting value from it has not.

This is not primarily a model quality problem. The gap shows up even where the underlying AI capability is genuinely useful at the task level. The issue sits one layer up, in how that capability is embedded, or not embedded, into the way work actually gets done.

Why isolated pilots do not add up

PwC’s analysis points to a specific and fairly diagnosable cause. It identifies “isolated, tactical AI projects” as a primary driver of the value shortfall: a chatbot in one team, a predictive model in another, deployed without strategic coordination or integration into core processes. Individually useful, collectively invisible at the level of revenue or cost that a CEO or board actually tracks.

Writer’s survey reinforces the same pattern from a different angle. Seventy five percent of executives admitted their AI strategy is “more for show” than something that actually guides decisions, and a large share of deployments have not been accompanied by any real change to the workflows or roles they sit inside. Rolling out a capable tool into an unchanged process tends to produce a modest efficiency gain at best, and often just adds a new step for someone to manage.

The organisations getting a return are not the ones with the most advanced models. They are the ones that changed a process, not just added a tool to it.

What the successful minority is doing differently

The 12 percent of CEOs reporting both cost and revenue benefit are not simply further along the same path everyone else is on. The pattern in the data points to a different starting assumption: AI is treated as a reason to redesign a process end to end, rather than a feature bolted onto the existing one. That distinction shows up in three practical ways among organisations getting genuine returns:

• Ownership sits with the business function, not just IT. A pilot run entirely by a technology team, without the process owner accountable for the outcome, rarely survives contact with real operational complexity.
• Success is defined before deployment, in business terms. Cycle time, cost per transaction, error rate, not “adoption” or “usage,” which measure activity rather than value.
• Governance is built in from the start, not retrofitted. Clear rules on what data an AI system can access, who reviews its output, and how errors are caught, so the initiative can scale without creating new risk faster than it creates value.

What this means for enterprise leaders now

For a CIO or CTO under pressure to show results from AI spend already committed, the practical response is not to slow down. It is to stop measuring the wrong thing and start being honest about which initiatives are integrated into a real process and which are still sitting alongside one. That usually means picking a small number of high value workflows, redesigning them properly with the people who own them, and being willing to retire the pilots that were never going to scale regardless of model quality.

Boards are increasingly going to ask for this distinction directly. A technology leader who can show which AI initiatives changed a measurable business outcome, and why the others did not, is in a materially stronger position than one who can only report usage numbers.

Where Jaykan fits

Jaykan works with enterprises and government agencies to move AI initiatives from isolated pilots to properly governed, process integrated capability that shows up in the numbers a board actually tracks. That includes workflow redesign, AI governance frameworks, and the technical architecture to scale a proven use case safely across a large organisation.

If you are carrying AI spend that has not yet produced a clear return, a complimentary, senior led assessment will show you exactly where the value is being lost and what to change first.