McKinsey published its State of AI 2026 in August. It's based on 1,719 respondents from 97 countries, surveyed in May and June. The headline is uncomfortable: almost everyone uses AI, but the share of companies that see it in their profit hasn't moved in a year.
Adoption is no longer the question
89% of organisations now use AI in at least one business function. 56% use it in three or more, up from 51%. 44% say it's scaling across the company, up from 38% a year ago. In 2026, the question is no longer whether companies use AI. It's whether it pays.
The ROI gap
37% of respondents say AI has had at least some impact on EBIT. That's roughly where it was in 2025. Meanwhile, 80% say AI has made them personally more productive.
So individual productivity is up, but it isn't reaching the bottom line. Time saved by one person doesn't turn into money unless the process around that person changes too.
What the 6% do differently
McKinsey calls about 6% of respondents "AI high performers": companies where AI contributes at least 5% of EBIT and significant value. The biggest difference between them and everyone else isn't budget or tools.
Nearly three-quarters of high performers, 72%, have fundamentally redesigned their workflows, up from 55% last year. Among everyone else, it's one in four.
They're also twice as likely to say their senior leaders are visibly committed to AI, 2.7 times as likely to be scaling AI agents, and they aim AI at growth and innovation, not only at cutting costs.
The companies making money from AI redesigned the work first. That's process before automation, with numbers on it.
Smaller firms are behind, but they have an advantage
54% of companies with $1 billion or more in revenue are scaling AI, against one-third of smaller organisations. For AI agents it's 40% against 22%, and smaller firms have barely moved in a year.
But look at what the leaders actually did: they redesigned workflows. That's far easier with 30 people than with 30,000. A small firm can map its whole operation in two weeks and change it in a quarter. A large one needs a programme.
The workforce question
39% of respondents expect AI to reduce headcount in their company next year. Worth knowing: last year, 32% predicted cuts, and only 14% actually saw them.
Meanwhile, 47% of middle managers and individual contributors report negative effects from AI-related change, such as strain. For most firms in 2026, the practical risk isn't mass layoffs. It's tired teams asked to adopt new tools on top of a process that doesn't work.
Costs are starting to bite
One in five respondents says AI running costs now limit how much they use it. 28% spend more than a tenth of their IT budget on AI, and 60% expect to spend more next year. Choosing where AI earns its place is becoming a budget question, not just a technical one.
What this means for recruitment and outsourcing firms
The survey doesn't look at recruitment on its own. But its main lesson carries over well.
- Don't measure AI by how many people use it. Measure it in hours and euros.
- Pick one workflow (screening, timesheets, candidate updates) and redesign it end to end before adding AI.
- Put a senior person in charge of it. Leadership commitment is one of the clearest lines between high performers and the rest.
- Watch running costs from day one.
AI pays when the work around it changes. Everywhere else, it just makes people a little faster at a process that stays the same.