In 2025, MIT's Project NANDA reported that 95% of enterprise generative AI pilots delivered no measurable financial return. The figure travelled fast, and its method was debated.
The 2026 evidence points the same way. McKinsey's State of AI 2026 found 80% of firms report better individual productivity. Yet only 37% report any profit effect, the same share as a year earlier.
Only about 6% of firms qualify as AI high performers. So most organisations are getting busier with AI without getting better.
The bottleneck is the organisation
RAND's 2024 study of failed AI projects found the leading cause was misunderstanding the problem AI should solve. Other causes included unready data and chasing technology over outcomes.
Microsoft's 2026 Work Trend Index puts a number on it. Organisational factors explained 67% of the AI impact people reported. Individual factors explained 32%.
Deloitte's 2026 Human Capital Trends adds a warning. Technology-led approaches were 1.6 times more likely to miss expected returns.
The learning gap
The MIT study named the core barrier "the learning gap". Tools that cannot remember context or improve from feedback frustrate people inside real workflows.
The gap is organisational too. Successful firms learned how to embed AI in real processes, choose problems worth solving and let the people doing the work shape the tools.
What the successful few do differently
In the MIT study, firms that bought specialised tools succeeded about twice as often as those building their own. They gave line managers ownership of adoption and targeted process-heavy back-office work.
The 2026 data adds three patterns. Each maps to one of the conditions in The Human Advantage Programme.
They lead the change. Gallup's 2026 report found manager-led adoption was a top driver of frequent use. People need a manager who uses AI openly, explains why it matters and hears the problems. This is Adoption.
They redesign the work. McKinsey's high performers were far more likely to redesign workflows. They also name who directs and checks AI agents, and how long checking takes. This is Adaptation.
They stay ready for the next change. BCG's AI at Work 2026 study found strategic clarity mattered more than tools. Clear priorities let good pilots scale before the next tool arrives. This is Acceleration.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
Why pilots stall
Most stalled initiatives follow a familiar pattern. A tool is bought, training is booked and targets are set. Six months later, usage has faded and the pilot has not spread.
Usually one condition is weak. People hide how they really use AI. Or AI sits on top of an unchanged process, and checking swamps the gains. Or a pilot works but waits months for a decision.
The weakest condition sets the pace for the others. Fixing the strongest one changes little.
What this means for leaders
Before launching another pilot, ask three questions. Will people take on the change and make it part of their work? Who will direct and check the agents, and own the hand-offs? Will the next new tool build on this, or start the change again?
If you cannot answer one of them, fix that first. Otherwise you may be funding the next statistic.
Try this
List your current AI initiatives. For each, score Adoption, Adaptation and Acceleration from one to ten.
Find the lowest score across the list. Then take the matching diagnostic: Adoption, Adaptation or Acceleration.
