Why Your AI Programme Is Not Delivering and What to Do About It
I have had some version of the same conversation four times in the past month. A company has rolled out AI across the business, the board wants to know what it is delivering, and the results are still hard to see. Inevitably, someone starts asking whether the technology itself is the problem.
It is not.
A National Bureau of Economic Research survey of more than 6,000 senior executives across the US, UK, Germany, and Australia, published in Harvard Business Review this month, found that roughly 90% reported no measurable improvement in productivity attributable to AI across the last three years. David De Cremer of Northeastern University, who analysed the data, identifies the pattern clearly: leaders are deploying AI against the most urgent problems they can see, rather than the problems where AI can create genuine strategic value. The result is a proliferation of pilots that solve visible, near-term friction while leaving the underlying operating model untouched.
McKinsey’s latest global research, published on 8 July, provides a structural framework that I have found genuinely useful. They surveyed 750 employees and leaders and mapped organisations across three horizons of AI maturity.
Horizon One: isolated pilots and point solutions.
Horizon Two: AI scaling across functions.
Horizon Three: the operating model itself has been redesigned around human-AI collaboration, with roles, workflows, and decision rights rebuilt from first principles.
Only 11% of organisations have reached Horizon Three. The majority are still at Horizon One, regardless of how much they have spent or how many tools they have deployed.
The McKinsey finding every C-suite leader should pay attention to is that employees are often more ready for AI than the organisations they work for. The real drag comes from slow decision-making and outdated structures and roles that were never designed for this way of working.
Deloitte’s research on rewiring the enterprise operating model for AI scale, published earlier this year and widely referenced this fortnight, puts numbers on the gap. Three quarters of surveyed leaders say their operating model MUST change within the next 12 to 18 months to drive greater value. Only about a quarter say they are already changing it continuously. The distance between those two figures is where AI value disappears.
Three questions worth putting to your leadership team this week
How many of your AI initiatives have actually changed how decisions get made, how workflows, or how roles are defined? If the honest answer is few or none, you are at Horizon One regardless of what the AI budget looks like.
Who owns the operating model redesign? If it sits with IT or with a standalone AI team, the problem has been misdiagnosed. Redesigning how an organisation works is a leadership and talent challenge as much as a technology one. The CHRO and COO need to be central to this conversation, not downstream of it.
What does your board actually know about where you are on the transformation curve? Most boards are receiving updates on AI adoption. Very few are having the harder conversation about whether the organisation has fundamentally changed around it.
The talent impact
This shift is creating roles that most organisations have yet to start hiring for.
The demand I am seeing in the market is for leaders who can sit at the intersection of AI capability and organisational design: people who understand what the technology can do and who have the change management experience to rebuild structures around it.
The market is placing a premium on Chief AI Officers who can lead transformation, not simply explain the technology. They need the judgement to decide where AI belongs, where it does not, and how the organisation must change to make it useful.
The Safest Job Is at the Company You Are Avoiding
Every week, I speak to senior professionals who are anxious about AI and their job security. A significant number are gravitating towards organisations that are moving slowly on AI, reasoning that caution equals safety. New research published this month suggests they have the logic exactly backwards.
A study from Ramp Economics Lab and Revelio Labs, tracking the AI spending and workforce records of nearly 22,000 US companies between January 2021 and February 2026, found that firms spending more on AI were actually less likely to lay off workers than those spending less. Heavy AI adopters grew headcount by 10% and entry-level hiring by 7%, while low adopters cut both.
The companies embracing AI most aggressively were also the ones growing, hiring, and building the structures that will define the next decade.
The reason is not complicated. AI is not simply automating jobs out of existence at the companies deploying it well. It is expanding what those organisations can do, which creates more demand for human judgement, oversight, and strategic direction, not less. The work that remains after AI handles the routine is harder, more consequential, and better paid.
For anyone thinking about their next career move, this matters. The companies making real progress with AI are also working out how jobs, teams and decision-making need to change around it.
Being inside one of those businesses gives you something difficult to learn from the outside: a practical understanding of how AI actually fits into the work, where people still add the most value, and what good collaboration between the two looks like.
That experience will count for a LOT. Three years from now, the people who understand this first-hand are likely to be far more valuable than those who waited for the market to become clearer.
This piece first appeared in Talent Pools Nexus, our weekly read for leaders in research, insights and consulting. Subscribe on LinkedIn.
