“Nobody in our market had a bad year. Two of them had a very good one.”
Six in a hundred
The winner-take-more effect is a small group of companies turning the same tools into a widening lead. They did not spend more. They rebuilt how the work runs, and you added AI to how it already ran.
Your numbers are fine. That is what makes this one hard to see, because the thing that moved is the distance between you and the two companies you lose to.
Same tools, same market, same four years. The gap is the story.
Sound familiar
What you notice first, second and much later.
This one hides for a long time, because none of the early signs look like a competitive problem. They look like an ordinary quarter.
1
First: nothing
Everyone in your market bought the same tools in the same year. Your people report they are faster, and they are right. Eight in ten say AI improved their own productivity.1 The market looks like it always did.
2
Then: a client mentions it
Somebody says the other supplier turns this around in two days. You check, and it is true, and it is one specific thing rather than everything. You file it as a process difference and go back to work.
3
Much later: three tenders in a row
You lose on price or on lead time to the same name three times. The reason sits in a workflow they rebuilt two years ago, and rebuilding yours now takes the same two years while they keep going.
The awkward part of stage two is that it is the only cheap moment. By stage three the fix costs what it would have cost then, plus two years of distance.
The winner-take-more effect is a small group pulling ahead with the same tools everyone bought.
Six companies in a hundred get real financial results from AI, and that share has not moved in a year while everyone kept spending.1
In short
McKinsey asked 1,719 people in 97 countries this spring. 6% attribute at least 5% of profit to AI and call the impact significant. Exactly the same share as a year earlier.1
What separates them is not budget. Nearly three quarters have rebuilt a workflow around AI, up from 55% the year before, against about a quarter of everyone else.1
Among large companies, the share scaling AI agents went from 27% to 40% in one year. Among smaller companies it stayed flat at 22%.1
Microsoft found that the organisation around a person explains 67% of the impact AI has, against 32% for the person's own mindset and behaviour.2
There is a sum further down that tells you how fast the distance grows. It takes two minutes and you can change every assumption in it.
It tends to start with one of these
A client told you what a competitor now turns around in two days
Your AI spend looks like everyone else's and your results do too
You are about to approve another year of the same, and you would like a reason to
You invest in companies and you want to know which of them is quietly compounding
What you take away
A sum that shows how fast the distance to your fastest competitor grows
What the 6% actually do differently, with the numbers behind each one
Why adding AI to a workflow and rebuilding one produce different curves
Where this sits in the five AI Culture Levels, and what moves you there
What we mean by it
Everybody bought the same thing. Six percent did something with it.
McKinsey has run the same global survey for nearly a decade. This year's edition, published on 25 August, asked 1,719 people in 97 countries. The headline is that almost nothing moved: 37% say AI has contributed anything to their profit, the same as last year, and 6% are high performers, also the same as last year.1
Meanwhile the share of companies rolling AI out across the whole organisation went from 38% to 44%, and eight in ten people say it made them personally more productive. So the activity went up and the group getting results did not grow. That is the shape of a market where the same few keep winning.
They did not outspend you. They rebuilt one thing and you decorated one thing, and four years later that is a different company.
Paul Musters
What the 6% do differently is the useful part, and it is duller than most people expect. Nearly three quarters of them have fundamentally redesigned a workflow because of AI, up from 55% the year before. Among everyone else it is about one in four.1 That single behaviour separates the two groups better than budget does.
The measured part
Five things they do more, and one they do the same.
All of these come from the same survey, comparing the 6% with everybody else in it.1
Rebuilt a workflow around AI, rather than adding AI to it
~75%
Everyone else, same question
~25%
Chasing efficiency with AI, the high performers
~80%
Everyone else, same question
~80%
The first pair is where the two groups split. The second pair is where they are identical: both go after efficiency at about the same rate. The difference is that the high performers also go after growth.1
Read that second pair again, because it is the one that changes what you do on Monday. Everyone is chasing efficiency. What almost nobody outside the 6% is doing at the same time is chasing growth or something genuinely new. Efficiency alone gets you a cheaper version of the company you already had.
2×
as likely to have senior leaders who visibly commit to AI, and to have a written way of measuring what AI initiatives deliver
And the gap between big and small opened in one year
This is the number I would put in front of a scale-up board. Among companies with more than a billion in revenue, the share scaling AI agents went from 27% to 40% in twelve months. Among smaller companies it stayed exactly where it was.1
Large companies
More than a billion dollars in annual revenue
Last year27%
This year40%
+13 pointsThe fastest move anywhere in the survey.
Smaller companies
Where most readers of this page work
Last year22%
This year22%
No changeNothing went backwards. The distance grew anyway.
Share of respondents whose organisation is scaling AI agents in one or more business functions, split by company size, in two consecutive editions of the same survey.1
Smaller companies did not go backwards. They stood still while the other group moved, which produces the same result from the outside and feels completely different from the inside. Nothing went wrong in your year.
Do the sum
How fast does the distance grow?
Pick the workflow your customers judge you on. An order, a quote, a claim, a release. Then compare two ways of using AI on it: adding it to the steps you have, or rebuilding the flow. Every number here is yours to change.
days after n years = today × (1 − yearly gain)n, for each of you
You, after 3 years8.6 days
Them, after 3 years6.1 days
They are faster by1.40×
The 5% and the 15% are an illustration of the difference between adding AI and rebuilding, not a measured figure. What is measured: nearly three quarters of the companies getting real results have rebuilt a workflow, against about a quarter of everyone else.1
The number that matters here is the last line, not the first two. Your customer does not experience your eight and a half days. They experience that somebody else does it in six.
Try it with a five-year horizon and the same percentages. That is roughly the length of a strategy cycle, and it is where a difference that looked like a rounding error becomes the reason you are not on the shortlist.
On the ladder
The 6% sit on a level, and you can see which one.
We read companies on five levels, from everyone doing their own thing to AI being part of how the place runs. Here is the whole ladder, with how many companies sit on each step.
01Campfire60%
02Wild West25%
03Blueprint10%
04Engine4%The work gets rebuilt here
05Ecosystem1%
Share of companies per level. Level 4, Engine, is where AI sits inside how the company runs rather than next to it, which lines up closely with the 6% McKinsey identifies.13
At Level 1 and 2, which is about eight in ten companies, AI is something individuals use. Good things happen and they stay where they happened. At Level 3 the method gets written down, so a good workflow can spread. Level 4 is where somebody redesigns the work itself, which is exactly the behaviour McKinsey found in three quarters of the high performers.
Two levels of distance sounds small. It is roughly the difference between the 27% and the 40%, and between a company that compounds what it learns and one that keeps starting over.
What a measurement adds before you redesign anything
An Operating Profile in use. Personality type and AI level in one profile, with the agents that fit it.
Microsoft looked at 29 factors and asked which ones predict whether AI is delivering anything. The environment around a person, meaning culture, manager support and how talent is handled, explains 67% of it. The person's own mindset and behaviour explains 32%.2 More than twice as much sits outside the individual.
Which is why redesigning a workflow fails more often than it should. The new flow assumes people who work a certain way, and nobody checked whether those people are in the building. One profile per person, holding two things: the way somebody works, and how far along they are with AI. You find out who can already run the new version, and who needs something first.
About emaho
emaho measures one Operating Profile per person: personality type and AI level in a single profile. On that we build a personal set of AI agents that fit how that person works, inside the tools they already use. Fifteen minutes to complete, first profile free, built for companies between 20 and 500 people.
Fifteen minutes per person. No credit card, no strings.
What to do
One workflow, three quarters.
Nobody rebuilds a company in a planning cycle, and the 6% did not try. They took one thing apart and put it back together in a shape that assumed AI from the start. Here is what that looks like on a calendar.
This quarter · choose and look
Pick the flow your customers judge you on. One, not three, and not a programme. Then spend two weeks finding out who could actually run a redesigned version of it, because this is where most redesigns die: the process is fine and the people were never part of the plan.
Next quarter · rebuild the middle
Take out the steps that only exist because a person used to do the previous one by hand. Set a growth goal next to the efficiency goal while you are at it. Both groups in the survey chase efficiency at about the same rate, so efficiency is the entry fee. What the high performers add is a target that only makes sense if the work is different.
The quarter after · make it survive
Give it a name at board level and a number that gets reported. High performers are twice as likely to have senior leaders who visibly commit, and twice as likely to have a written way of measuring what the work delivers.1 Those two travel together, because a redesign without a number gets quietly abandoned in month five.
Three quarters is not fast, and that is the honest part. The companies ahead of you started theirs a while ago, which is exactly why the calendar matters more than the ambition.
Which workflow would you rebuild?
Send me the one your customers judge you on and what it takes today. I will tell you what I would look at first and where these redesigns usually stall. I answer these personally, usually the same week.
The first cost is invisible on your own numbers. You improve every year, your margin holds, and the distance to the fastest company in your category grows anyway. Run the sum above with five years instead of three and you will see how quietly that happens.
Then there is the money you spend without moving. Sixty percent of companies expect to increase their AI investment this year, and the share getting real financial results has not changed in twelve months.1 Spending more on the same shape of work is how you stay in the 94%.
The one that actually decides it is time. Rebuilding a workflow takes quarters, not weeks. Every year you wait, the company that started earlier is a year further into a curve that keeps bending, and you begin from where you are now rather than where they were.
Questions
Questions about the widening gap
What is an AI high performer?
In McKinsey's global survey it is an organisation that attributes at least 5% of its profit to AI and describes the impact as significant. In the August 2026 edition, based on 1,719 respondents in 97 countries, 6% of companies met that bar. That share was identical to the year before, even though the number of companies scaling AI across the enterprise rose from 38% to 44%.
Why do a few companies get results from AI while most do not?
Because of what they did with it rather than how much they spent. Nearly three quarters of the high performers in McKinsey's 2026 survey have fundamentally redesigned a workflow around AI, up from 55% a year earlier, against roughly a quarter of everyone else. Adding AI to a process that keeps its old shape produces a faster version of the same result.
What is the difference between adding AI to a workflow and redesigning one?
Adding AI speeds up steps that already exist, so the handovers, approvals and waiting time stay where they were. Redesigning asks which steps still need to exist at all now that a machine can do part of the work. The first gives you a few percent a year. The second is what separates the 6% from everybody else in the research.
How do you find out how far ahead or behind your company actually is?
Not by comparing tool spend, because everybody buys the same things. emaho reads organisations on five AI Culture Levels, from everyone doing their own thing to AI being part of how the place runs. That places you on the same ladder as the companies pulling ahead, and the scan takes three minutes and six questions.
Is the gap between AI leaders and everyone else actually widening?
On one measure it clearly is. Among companies with more than a billion dollars in revenue, the share scaling AI agents went from 27% to 40% in twelve months, while among smaller companies it stayed flat at 22%. The smaller group did not go backwards. The distance grew because the other side moved.
How much do AI leaders spend compared to everyone else?
More, though not enough to explain the difference. High performers in McKinsey's 2026 survey are more than twice as likely to put over 15% of their IT budget into AI, and more than half expect to raise investment by 10% or more. But roughly 80% of both groups chase efficiency with AI. The split is in what else they chase.
Should we chase efficiency or growth with AI?
Both, and the second one is where the difference sits. In McKinsey's 2026 survey, about 80% of high performers and 80% of everyone else pursue efficiency. What most high performers add is a growth or innovation goal alongside it. Efficiency alone gets you a cheaper version of the company you already had.
Where can I read about the other AI challenges that decide who pulls ahead?
How long does it take to redesign a workflow around AI?
Quarters rather than weeks, which is exactly why the gap compounds. The high performers who report fundamentally redesigned workflows went from 55% to nearly 75% over a year, so even for them this is a multi-year build. Every year of waiting starts you from where you are now rather than where they were.
Why does AI investment keep rising while results stay flat?
Because spending is easier to decide than redesigning. Sixty percent of companies in McKinsey's 2026 survey expect to increase AI investment, while the share reporting real financial impact has not changed in a year. More money on work that keeps its old shape produces more of the same result.
What separates AI leaders besides redesigning workflows?
Three things, all from the same survey. They are twice as likely to say senior leaders visibly commit to AI, twice as likely to have a defined process for measuring what AI initiatives deliver, and three times as likely to be scaling AI agents across most business functions. They are also 3.3 times as likely to intend a fundamental transformation within three years.
Close by
Which side of the gap you land on
Each of these is a reason a company keeps buying tools and stays where it is.
Two thirds of whether AI delivers anything sits in the organisation around the person. Before you rebuild a workflow, find out who in your company is already working the way the new one assumes.
Look at the organisation, not the toolingTwo thirds of whether AI delivers anything sits in how the work around the person is arranged.
Find out who could run the redesigned versionThe profiles tell you who could carry a rebuilt workflow tomorrow and who would need a year.
Rebuild one workflow with those people in itStarting with the workflow and hoping the people catch up is the expensive order to do this in.
First profile free · no credit card · built for companies of 20 to 500 · you decide what your team gets to see
Not ready to put your team in anything yet? Start with the level of the company instead. The Culture Level scan is six questions, three minutes, and asks nothing of you.
Fifteen years of leadership and team development in Dutch scale-ups. That practice now sits in software: one Operating Profile per person, with agents that actually fit. He writes these pages from what he runs into with clients, not from a research summary.
McKinsey, The state of AI in 2026: On the road to ROI, published 25 August 2026. Online survey run 4 May to 8 June 2026, 1,719 respondents in 97 countries, weighted by each country's share of global GDP; 36% work for organisations with more than one billion dollars in annual revenue. Source of: 6% are AI high performers, attributing 5% or more of EBIT to AI and calling the impact significant, unchanged from 2025; 37% attribute at least some EBIT impact to AI, also essentially unchanged; 44% report AI scaling across the enterprise, up from 38%; among large organisations the share scaling AI agents rose from 27% to 40% while smaller organisations stayed flat at 22%; nearly three quarters of high performers have fundamentally redesigned workflows, up from 55% the year before, against about a quarter of other respondents; roughly 80% of both groups pursue efficiency, while most high performers also pursue growth or innovation; high performers are twice as likely to report senior leadership commitment and defined processes for measuring impact, more than twice as likely to spend over 15% of their IT budget on AI, three times as likely to be scaling agents in most functions, 2.7 times as likely for other agentic AI, and 3.3 times as likely to intend fundamental transformation within three years; 80% say AI improved their own productivity; 60% expect to increase AI investment.
Microsoft, 2026 Work Trend Index Annual Report, published 5 May 2026. Survey by Edelman Data x Intelligence among 20,000 knowledge workers who use AI at work, across ten markets including the Netherlands, fielded 18 February to 7 April 2026, alongside anonymised Microsoft 365 telemetry. Source of: organisational factors such as culture, manager support and talent practices account for 67% of reported AI impact against 32% for individual mindset and behaviour, measured across 29 factors. Microsoft notes these are self-reported statistical associations rather than causal effects.
emaho AI Culture Levels. Share of organisations per level, calibrated against BCG 2025 and McKinsey 2025. Level 4, Engine, holds about 4% of organisations.
Numbers are quoted as published. The 5% and 15% in the sum are an illustration of the difference between adding AI to a workflow and rebuilding it, not a measured figure, and both are adjustable. The three stages in the recognition block are composites drawn from client situations rather than transcripts.
