Senior leaders at some of the world’s larger companies want to be running on AI within two years. Most are nowhere close, and a new study argues the reason has little to do with the technology itself. The Positive Group, a London behavioral-science consultancy, drew on interviews, surveys and focus groups with 35 senior leaders and found a wide gap between AI ambition and AI delivery. Across the group, 86% said they want to reach advanced AI maturity within the next 12 to 24 months. Only 28% said they have embedded AI into core workflows in a way that delivers measurable value at scale. The study places the distance between those two figures on how leaders behave.

The headline figures, 86% and 28%, sit on The Positive Group’s own report page, so they are not buried in a private deck. They are also self-reported by a small group, which matters for how much weight to give them. Advanced AI maturity, in the study’s terms, means AI embedded in core workflows and producing measurable value, the stage past one-off pilots and proofs of concept. By that bar, fewer than three in 10 of the leaders surveyed say they have arrived, even as almost nine in 10 want to be there inside two years.

The sample spans professional services, financial services, aviation, consumer brands and life sciences, with participants from companies including McDonald’s, ServiceNow, KPMG and White & Case. They are based in the U.S., U.K., Singapore, Australia, Hong Kong and Spain, and the fieldwork ran in September and October 2025. It is a small, senior and qualitative sample, so the value is in the pattern it describes more than in any single percentage.

The study breaks the problem into a set of behavioral gaps, each one a distance between what leaders say about AI and what they do with it. The first is role modeling. Just 33% of those surveyed said their leaders visibly demonstrate adaptability with AI, which the authors read as the difference between championing the tools in a meeting and using them in daily work.

The second is communication, where 65% felt their leadership could tell a motivating story about AI, while only 45% said leaders kept their teams regularly updated on real developments. A compelling narrative, in other words, is more common than a steady flow of information. The third gap is training, and it is the most uneven of the three.

Six in 10 leaders say structured training reaches everyone. For the other four, it is partial, ad hoc or absent, which is a thin foundation for the broad adoption the same leaders say they want.

Across every sector studied, the same pattern holds. Where AI stalls, leadership behavior is the common thread.

The study also connects leadership culture to how far a company has scaled AI in practice. Leaders at organizations it classifies as high-maturity, meaning those already scaling AI across the business, rated their management teams markedly higher on behavior than peers still working through pilots. The direction is intuitive and the size is not trivial.

On adaptability, the behavior the study ties most closely to role modeling, leaders at high-maturity companies score about 40% higher than those still running pilots. The same leaders score higher on clarity and trust, which suggests the gap is less about the toolset a company buys and more about how its leaders carry the change.

Underneath the numbers, the study describes four behavioral patterns that recur across the leaders it spoke to. Each one is a way that human dynamics, more than tooling, slow AI down.

The study’s prescription follows from its diagnosis. If behavior is the bottleneck, then the work for leaders is behavioral, and the report groups it into three modes.

The leaders the study quotes describe the same instincts in plainer terms. Tarv Nijjar, who leads product and platform transformation at McDonald’s, said the test is whether you can explain AI in language simple enough for a six-year-old to follow, which is what builds curiosity and trust. Isabel Parker, chief innovation officer at White & Case, said the discipline that matters most is holding to a clear strategy and resisting the pressure to chase every new AI trend.

Will Marien, who heads The Positive Group, frames the next phase of AI as a test of people and judgment ahead of code. He argues the real value shows up when AI is paired with human judgment, meaning and shared understanding. “It is leadership that will determine how powerfully that partnership performs,” he said.

None of this escapes the limits of the study. The figures come from 35 leaders describing their own organizations, and The Positive Group sells leadership development, so it has a stake in the conclusion. Even so, the through-line, that the harder part of AI is human and organizational, lines up with larger research. McKinsey has tied executive sponsorship to materially stronger AI results, with the leading adopters far more likely to have CEO or board-level backing. The underlying research was published in Harvard Business Review. For the leaders racing toward their two-year deadline, the study lands on a single point: the next gains depend on how they lead, as much as on the models they buy.