The AI value gap is an adoption gap
Almost every company now uses AI. Very few see it in their earnings. The missing piece isn’t a better model. It is people using AI inside real work.
If you sit on a board today, you have almost certainly approved spending on AI. You have probably seen pilots, demos and a slide promising productivity gains. What you may not have seen is the gain itself, in the numbers that matter.
You are not alone.
Widespread use, rare returns
McKinsey’s latest global survey, The State of AI in 2025, asked 1,993 participants in 105 countries how their organisations use AI. The findings are striking:
88% say their organisation regularly uses AI in at least one business function, up from 78% a year earlier.
About two-thirds have not yet begun scaling AI across the enterprise.
Only 39% report any impact from AI on enterprise-wide EBIT, and most of those say it accounts for less than 5% of EBIT.
About 6% qualify as AI high performers: organisations that attribute 5% or more of EBIT to AI and report significant value from it.
In other words, using AI has become normal. Earning from it has not. Most companies aren’t failing to try. They are stuck between the pilot and the payoff.
What the 6% do differently
The most telling finding is not about technology. McKinsey reports that high performers are nearly three times as likely as others to have fundamentally redesigned individual workflows around AI.
That should change how leaders think about the problem. The gap between buying AI and benefiting from it isn’t mainly a model problem or a data science problem. It is an adoption problem: whether thousands of people change how they do everyday work, inside the systems they already use.
That is precisely the problem digital adoption exists to solve.
Three places AI value leaks away
Watch where AI initiatives stall inside an enterprise, and the same three failure points come up.
People don’t know how to use it in their work. An AI feature launched in a CRM, an ERP or a separate chat tool is often ignored. Not because it is useless, but because nobody showed people when and how it fits the task in front of them. A general announcement and a webinar don’t change daily habits.
People don’t trust the answer. Staff in banks, insurers and other regulated businesses are right to be cautious. An answer without a source, or one drawn from content they shouldn’t see, is a risk. When trust is low, people check everything twice or stop using the tool.
The answer doesn’t turn into action. Even a good answer often leaves the employee to do the work by hand: copy the data, open another system, raise the request. The insight is there. The value isn’t, because nothing actually happened.
How a digital adoption platform closes each gap
A digital adoption platform sits exactly where AI meets daily work: inside the applications people use. That puts it in the right place to address all three.
Guide people through new, AI-assisted ways of working, step by step, on the real screen, shown to the right roles at the right moment.
Answer their questions from the company’s own approved knowledge, with every answer citing its source, and only from content each person is allowed to see.
Act on requests across connected systems, within existing permissions and approvals, so routine work is done, not just described.
Measure what changed: which teams completed which tasks, where people struggled and how adoption is moving over time, so leaders can link the investment to outcomes.
This is what workflow redesign looks like on the ground. Not a strategy document, but every employee being shown, helped and supported in the new way of working until it becomes the normal way.
The question for the board
The next AI budget conversation shouldn’t start with “Which model?” or “Which vendor?” It should start with “How will our people actually adopt this, and how will we know?”
The companies in McKinsey’s 6% have answered that question. The rest of us can learn from them.
Next step: Take one AI investment you have already made, and map where it is used, trusted and acted on, and where it isn’t. We can help.
Source
- McKinsey, The State of AI in 2025 (survey fielded 25 June to 29 July 2025), as reported by IDM (opens in a new tab) and IT Brief UK (opens in a new tab)
Rather talk it through?
Thirty minutes on your own screens, with your own questions. No slideware.