Why 89% of companies use AI but only 11% get real value from it

AI is everywhere in business. Its results are not. A gap has opened up between adoption and value creation, and neither budgets nor tools are enough to close it. What separates the few companies that get real benefit from AI from everyone else is not technology. It is a matter of method, governance and management.

Every major study now documents the same finding. According to the 2026 edition of McKinsey's State of AI, published in August, 89% of organizations regularly use AI in at least one business function, up from 88% a year earlier. Adoption is no longer the issue. AI has made its way into office software and into customer service and marketing teams, often without any formal decision ever being made.

But as soon as you move from usage to impact, the picture changes sharply. Only 11% of the companies surveyed by KPMG in its March 2026 Global AI Pulse Survey have moved beyond experimentation to deploy AI at scale across their operations. A study by MIT FutureTech, based not on self-reported answers but on annual reports filed with the SEC, reaches the same figure: only 11% of S&P 500 companies have truly built AI into their core processes. McKinsey's findings are even harsher: nearly two-thirds of respondents see no measurable effect on their operating income, and only 6% attribute at least 5% of their EBIT to AI. So the question is no longer "should we adopt AI?" but "why do the vast majority of those who adopt it get almost nothing out of it?"

The invisible productivity paradox

The first reflex is to conclude that AI doesn't work. That's wrong. Eight in ten respondents to the McKinsey survey say AI has improved their own productivity. Employees write faster, summarize faster, research faster. The gain is real, and it is visible in each individual role.

The problem is that this gain doesn't travel upward. An hour saved by an account manager drafting a proposal doesn't become an hour of value for the company if the approval process stays exactly the same. It dissolves into the organization: a few more meetings, a few more emails, a little more comfort. Individual productivity rises while collective performance stalls. This is the central paradox of this technological wave, and on its own it explains much of the gap between adoption and value creation.

Three reasons AI fails to create value

The first reason is layering. In most organizations, AI has been rolled out in successive layers: a general-purpose assistant open to everyone, a few tools bought by individual business units, and a string of proofs of concept launched with no shared direction. KPMG puts it bluntly: companies that are stuck bolt AI onto their existing workflows, while the 11% that succeed redesign their processes before deploying AI. For them, AI is not a tool. It is an opportunity to rethink how work gets done.

The second reason is the lack of value-based management. Many companies track the number of licenses, usage rates and the number of use cases launched. Very few measure what each of those use cases actually delivers in margin, turnaround time, quality or revenue. McKinsey finds that top-performing companies are far more likely to have a process for measuring impact and an explicit commitment from senior leadership. Without a quantified target, an AI project can neither fail nor succeed. It simply drags on.

The third reason is the most underestimated: AI has been treated as a technology issue when it is first and foremost an organizational one. BCG captures this with its 10-20-70 rule: 10% of the effort goes into algorithms, 20% into technology and data, and 70% into people and processes. Most companies have invested in the reverse order. They bought the tools, then waited for usage to follow.

What companies that create value do differently

The studies converge on a small number of practices that distinguish companies able to turn AI into results. Four stand out clearly.

The first is to start from processes, not tools. Among the top performers identified by McKinsey, 73% have fundamentally redesigned their workflows, compared with 25% of other companies. In practice, this means choosing a high-stakes business process, mapping how it really works, then asking what it would look like if part of the work were handed to AI agents. This is an exercise for the executive team, not the IT department.

The second is to concentrate efforts rather than spread them thin. BCG shows that 70% of AI's potential value is concentrated in a handful of core functions: sales and marketing, manufacturing, supply chain and pricing. Successful companies pick two or three high-impact initiatives and see them through, rather than multiplying pilots that will never leave the lab.

The third is to set a measurable objective for each use case from day one. A margin gain, a processing time cut in half, an improved conversion rate: the indicator matters less than the fact that it exists, is tracked and determines whether the project continues. That is what makes it possible to quickly stop what isn't working and accelerate what is.

The fourth is to take it all the way to the top. In every study, direct involvement from senior leadership stands out as a decisive factor. Not to sign off on a budget, but to make the hard calls: redesigning a process affects organizational structures, roles and individual responsibilities. No project team can settle those questions on its own.

The pitfalls awaiting companies in a hurry to catch up

Recognizing the gap is not enough. Three mistakes come up again and again among companies that decide to speed up.

The first is to confuse speed with haste. Faced with competitors who seem to be moving ahead, it is tempting to launch a massive program with dozens of use cases running in parallel. The result is almost always the same: scattered resources, diluted results and team fatigue that ends up undermining the credibility of the whole effort.

The second is to outsource the thinking entirely. Handing AI strategy to an outside provider without leadership taking ownership of the reasoning produces elegant roadmaps that rarely get executed. A transformation of this kind cannot be delegated. It has to be led.

The third, at the other extreme, is to believe AI can do everything on its own. The faster and more confidently agents deliver output, the greater the need for human oversight. An analysis generated in a few minutes is only valuable if an expert has reviewed it, challenged it and stood behind it. The companies that capture value are not the ones that automate the most. They are the ones that have put the most rigorous controls in place.

Closing the gap: a matter of method

The gap between the 89% and the 11% is therefore neither inevitable nor a question of resources. Companies that get value from AI don't have access to different technology: they use the same models, often the same tools. What sets them apart is the order in which they asked the questions. First the business problem, then the process, and only then the technology.

BCG measures what is at stake: the most advanced companies achieve revenue growth 1.7 times higher than laggards and three-year total shareholder return 3.6 times higher. This is no longer a marginal advantage. It is a difference in trajectory that, if it holds, will reshape the pecking order in many industries.

For a business leader, the right question is not "how much AI have we deployed?" but "which decision, which process, which margin has AI actually changed this year?" As long as the answer stays vague, the company belongs to the 89%. Joining the 11% starts the day that answer becomes precise.

Hymeria's position

Hymeria Consulting was built to help business leaders move from usage to value. Rather than adding one more tool or program, we start from a specific question: a process to rethink, a decision to inform, a use case to assess. Our engagement format is deliberately lean: senior-level scoping, more than 40 specialized AI agents producing the analysis, systematic validation by an expert with over 15 years of experience, and an actionable deliverable in 5 to 10 days, within a fixed budget.

Related service: Operations & Organizational Design, process mapping, internal policies and organizational benchmarking in 5 to 10 days.