When work gets faster, but results don’t.
In previous issues, we have gradually shifted our focus from technology to the organization itself: from how AI changes thinking to how it changes collaboration and the way teams are structured. The next step seems natural: if individual employees have already become significantly more productive, why is this so often not reflected in company results?
The latest McKinsey Global Survey on the State of AI 2026 provides a striking starting point for this paradox. 80% of respondents say AI has increased their individual productivity. Yet only 37% report a positive impact of AI on their organization’s EBIT. And that latter figure has barely changed over the past year, even as AI adoption continues to accelerate.
This does not mean that AI is failing to deliver economic value. Companies are already reporting cost reductions in some functions and revenue growth in others. But a local productivity gain and an organization-wide result are not the same thing.
And this raises a question:
When does AI transformation actually create productivity — and when does the organization simply start to stall?
We often see this when working with clients on process optimization. Imagine an analyst who uses AI to reduce the time needed to prepare a report from two days to a few hours. Their productivity has increased several-fold. But if the manager still reviews the report only a week later, legal approval takes another five days, and key decisions are made at a monthly board meeting, the analyst’s increased speed barely changes the economics of the overall process.
Or consider a sales team that can now prepare three times as many proposals. If each proposal still gets stuck in price approval, AI has simply moved the bottleneck further down the chain.
The employee became more productive. The process did not necessarily.
This is one of the central illusions of the current stage of AI transformation. As additional output becomes easier to generate, it becomes tempting to mistake more output for higher productivity. But time saved only becomes valuable when the organization can actually use it: to accelerate decisions, increase business volume, improve quality, launch a new product, or genuinely reduce costs.
This is where the companies McKinsey identifies as AI high performers become particularly interesting. They represent only around 6% of respondents — organizations that report significant AI value and attribute at least 5% of EBIT to AI. Their share has barely changed compared with 2025.
But what distinguishes them is not simply how much AI they use.
Nearly three-quarters of high performers say they have fundamentally redesigned their workflows because of AI, with some rebuilding them from scratch. Among other companies, only about a quarter report doing so.
That is an important distinction. An organization can add AI to an existing workflow and gain time. Or it can ask why the workflow itself should remain unchanged if one of its key operations can now be performed in a fundamentally different way.
In the first case, AI becomes a new tool. In the second, it becomes a reason to change how work is done.
High performers are also significantly more likely to use AI not only for efficiency, but for growth and innovation. They are more likely to rethink workforce planning, systematically measure AI’s impact, determine where human oversight is required, and involve leadership in the transformation. McKinsey describes this as coherence of approach: results come not from an individual tool, but from coordinated changes to workflows, management, measurement, and scaling.
A significant shift emerges.
AI is creating new productive capacity faster than most organizations can redesign themselves around it.
Production accelerates. The organization’s ability to absorb what it produces does not necessarily keep pace.
This, in our view, is where the boundary between AI adoption and AI value creation lies.
The value of AI is determined not only by how much an individual can produce, but also by how much change the organization can absorb.
This also changes the way we think about optimization. For leaders, it is no longer enough to measure only the time saved or the amount of additional work produced. What matters is what happens to the entire process after AI is introduced: where the bottleneck moves, how freed-up capacity is used, which roles and control mechanisms change, and whether the productivity gain shows up in costs, margins, speed, or quality.
Sometimes AI genuinely allows an unnecessary layer to be removed. Sometimes freed-up capacity is better directed toward more complex work. And sometimes the economic value appears only after the process itself is redesigned, rather than a single operation within it.
For leaders, the question has therefore changed. It is no longer enough to ask: “How much time can AI save our employees?” or “How much additional work can they now perform?” The more important question is: “If our employees have already become more productive through AI, what needs to change in the organization itself for that productivity to become visible in its results — and can the organization redesign itself quickly enough?”
Perhaps this is where the real distinction now lies between companies that use AI and companies that truly capture its value.
🔗 McKinsey & Company. The State of AI in 2026: On the Road to ROI. August 25, 2026.
