Everyone bought the capability. Almost nobody can prove the value.
THE VALUE GAP · No 1Everyone bought the capability. Almost nobody can prove the value.
Enterprise AI is now table stakes. The organisations that pull ahead won’t be the ones with the most of it — they’ll be the ones who can stand in front of a board and say, with a number, what it returned.
———
“There is nothing so useless as doing efficiently that which should not be done at all.”
— Peter Drucker
There is a moment most transformation programs never quite recover from, and it does not happen at go-live. It happens months later — in an ExCo meeting, a budget meeting, a corridor conversation you half-overheard — when someone finally says the quiet part out loud: “I’m not sure we’re really getting the value out of it.” Nobody answers with full confidence. The room moves on. The question does not.
That question is the whole of it. The platform is live. The tickets are closing. There is a dashboard somewhere with reassuring numbers on it. And still, no one in the room can say what it was actually worth. I have watched this pattern from every seat at the table for more than thirty years — as the customer, as the integrator, as the adviser called in once a program had already lost its shape. It was true of enterprise platforms long before AI arrived, because the value gap was never really about the technology of the day. It opened the first time an organisation bought a capable platform — a ServiceNow, an ERP, a service-management tool — and expected returns without lifting its own IT maturity, without driving the automation the platform made possible, and without improving the experience of the people on either side of it. AI is the sharpest example we have yet seen. The foundations it stands on are the same ones that have decided platform value for twenty years.
The numbers make the case better than any argument I could. McKinsey’s late-2025 State of AI found roughly 80% of organisations have deployed generative AI, yet more than 80% report no material impact on earnings.¹ PwC’s 29th Global CEO Survey is blunter still: only 12% of chief executives report both cost and revenue gains from AI.² Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027, undone by unclear business value and weak controls.³ Even the vendors’ own scorecards concede it — ServiceNow’s 2026 Enterprise AI Maturity Index puts the organisations getting a positive return at around one in three. It is a vendor claim, but a revealing one.⁴
Read those together and a single pattern falls out. The capability is everywhere. The value is almost nowhere — or at least, almost nowhere anyone can prove.
———
60–70% of organisations are getting only a fraction of the value their platform was meant to deliver. It is the rule, not the exception.
— Bridge The Value Gap
———
Here is the uncomfortable part, and it is the argument of the book this series accompanies: the gap is almost never a technology problem. The platform works. The models work. What is missing sits upstream and downstream of the build. The value was never named before the project began. It was never owned by anyone once it shipped. And it was never measured against the outcome it was meant to move. “We went live” quietly became the finish line, when it was only ever the starting one.
Look at where the money actually has to go. Deploying the technology is the cheapest and easiest part of the whole exercise — the model is bought, switched on, and running within weeks. The expensive part, and the part almost nobody funds, is the reengineering of the process and the organisational change around it. Gartner, which now runs a body of work under the blunt title Value Is Trapped, finds that the organisations getting a return on AI invest up to four times more in the foundations — data, governance, AI-ready people and change management — than those that do not, and estimates that every hundred days of AI implementation demands something closer to two hundred days of change management.⁵ The software is the down-payment. The transformation is the mortgage.
And here is the part that should stop a board cold. If you deploy the technology and skip the process work and the change management, the question is no longer merely where is the value — it is whether you have created any value at all. A faster version of a broken process is not value; it is the same output, produced more expensively, with a new licence attached. Abhijit Dubey, chief executive of NTT DATA, puts it plainly: AI transformation is ultimately a people transformation.⁶ Leave the people and the process untouched and you have not transformed anything — you have automated the status quo and called it progress.
I grew up in England, and one of the first pieces of design I ever noticed — before I knew what design was — was three words painted on the edge of a London Underground platform. Mind the Gap. The warning does not close the gap between the carriage and the platform. It does something more useful. It makes the gap visible, names it, and puts the responsibility for crossing it squarely in front of you. That is, in the end, all this work really is: making the value gap visible, so you step over it rather than into it.
Capability became a commodity. Measurement became the moat.
So what do you actually do? Three things, and none of them are technical.
Name the value before you build. Not “efficiency”, not “innovation” — a business outcome, in dollars or in a metric your finance team already trusts. If you cannot name it, you cannot close a gap to it, because you have not said where the far side is.
Measure it independently. Not the vendor’s activity console, which is engineered to show usage — prompts served, tickets deflected, adoption curves climbing reassuringly. Measure the outcome you named, against where you started. Activity is not value, however green the dashboard.
And own it. Every diagnostic I have ever run ends on the same question: who in your organisation owns the outcome of this platform? Not the management of it — the outcome. If you can name that person, protect them. If you cannot, that is your first decision, and everything else waits behind it. Ownership is not a RACI box; it is a named human who loses sleep over whether the value arrives, with the authority to change the work when it does not.
Why does this keep happening? Because activity is easy to show and value is hard to prove, and under quarterly pressure the easy thing wins. An adoption curve climbs on its own; a realised-value number has to be defined, defended and owned by someone willing to be held to it. So the program quietly optimises for the metric it can produce rather than the outcome it was funded for — and everyone reports green while the gap widens underneath.
None of this is anti-AI. It is the opposite. AI raises the stakes rather than removing them; it amplifies whatever foundations you already have, accelerating the good ones and exposing the thin ones, simultaneously and without apology. If you believe in the capability — and you should — then you should want to prove it pays. Because in a market where everyone has the capability, proof is the only thing left that is scarce.
The organisations that pull ahead from here will not be the ones with the most AI. They will be the ones who can answer the ExCo question with a number, and mean it.
———
THE VALUE GAP — The full argument, and the BRIDGE framework for closing the gap, is in Bridge The Value Gap, out now — get the book.
———
References
1. McKinsey, The State of AI, November 2025.
2. PwC, 29th Annual Global CEO Survey, January 2026.
3. Gartner, forecast on agentic AI project cancellations, June 2025.
4. ServiceNow, Enterprise AI Maturity Index, 2026 (vendor claim).
5. Gartner, “Value Is Trapped: Unlocking AI ROI Through Organizational Change”, 2026 — successful AI programs invest up to ~4x more in foundational areas (data, governance, AI-ready people, change management); change-management effort estimates.
6. Abhijit Dubey, CEO, NTT DATA — “AI transformation is ultimately a people transformation”, 2025.