State of AI in companies: 37% attribute EBIT impact to AI

The state of AI in companies in 2026 fits in two numbers from the McKinsey global survey published in August: nearly nine in ten respondents say their organization regularly uses AI in at least one business function, and 37% attribute some EBIT impact to AI use, about the same share as a year earlier (McKinsey, 2026). Use and budgets grew, and the share that sees AI in the bottom line stood still. The year’s other reports tell the same story with different numbers, and much of the gap comes from the definition: official European statistics, which ask 157,000 companies whether they used any AI technology from a closed list, reach 19.95% in 2025, far from the 88% that circulates in presentations (Eurostat, 2025). Whoever takes these numbers to the board needs to know what each survey measured before putting two of them on the same slide.

At Iasbeck&Co we implement CRM and AI agents in B2B operations, and almost every board meeting about AI that I sit in on starts with a survey number. I read the most cited annual reports and organized what they say about adoption, functions, impact, investment, agents and risk, for the board, the sales, marketing and CS leaders, and HR.

Which are the most respected reports on the state of AI in companies?

I chose five sources that publish every year and describe their method. Three are consulting firms that also sell AI projects, Stanford’s AI Index is an academic report that compiles third-party data, and Eurostat is the only one with an official statistical sample.

Source	Year	Sample and method	Main finding
McKinsey, The state of AI in 2026	2026	Online survey of 1,719 respondents in 97 countries, May 4 to June 8, 2026, weighted by each country’s GDP	44% say AI is scaling across the enterprise; 37% attribute some EBIT impact, the same level as 2025
McKinsey, The state of AI in 2025	2025	1,993 respondents in 105 countries, June 25 to July 29, 2025	88% report regular use in at least one function; 23% are scaling agents and 39% are experimenting
Stanford HAI, AI Index Report 2026	2026	Compilation of third-party data; corporate adoption comes from McKinsey’s 2025 survey and investment from the Quid database	Global corporate AI investment of US$ 581.69 billion in 2025, up 129.9%
Deloitte, State of AI in the Enterprise	Jan. 2026	3,235 leaders from director to board level, in 24 countries and six industries, August to September 2025	20% already grow revenue with AI and 74% hope to; 21% have mature governance for agents
BCG, Applied AI Index 2026	Sept. 2026	1,330 C-suite and senior leaders	7.5% of companies are “future-built” and another 41% are scaling while outperforming laggards
Eurostat, use of AI in enterprises	2025	Official survey by national statistical institutes of 157,000 of the 1.53 million EU enterprises with 10 or more people	19.95% used at least one AI technology; 55.03% among large enterprises

How many companies use AI in 2026?

In the McKinsey series, the share of respondents reporting regular AI use in at least one business function went from 78% in 2024 to 88% in 2025, and the 2026 edition speaks of “nearly nine in ten.” With that indicator close to its ceiling, McKinsey now highlights depth of use: 44% say AI is scaling across the enterprise, up from 38% a year earlier, and 56% use AI in three or more functions, up from 51%. The gap by company size is large: among respondents from organizations with more than US$ 1 billion in annual revenue, 54% report AI at scale, compared with one-third at smaller ones (McKinsey, 2026).

Deloitte measures employee access. The share of workers with access to company-sanctioned AI tools went from under 40% to around 60% in one year, and fewer than 60% of those with access use the tool in their daily work, about the same pattern as the year before. Only 25% of respondents said their organization had moved 40% or more of its AI experiments into production, and 54% expect to reach that level in three to six months (Deloitte, 2026).

Why does one survey say 88% and another say 20%?

Eurostat asks an official sample of European Union companies whether they used at least one technology from a closed list, such as text mining, speech recognition, text or image generation and machine learning. In 2025, 19.95% of companies with 10 or more people said yes, with 17% among small companies and 55.03% among large ones (Eurostat, 2025). McKinsey asks professionals who agreed to answer an online questionnaire whether their organization regularly uses AI in any function, and 36% of those respondents work at companies with more than US$ 1 billion in revenue. The two surveys measure different things, and neither one is wrong.

Stanford’s AI Index states that organizational adoption reached 88%, and that number comes from McKinsey’s 2025 survey, as the report shows in the source line of each chart. The economy chapter warns that the results are self-reported and should be read as directional (Stanford HAI, 2026). When both appear side by side in a presentation, the board is looking at one survey cited twice. Brazil does not appear separately in any of these reports, and the Brazilian numbers are in AI in Brazilian companies.

Survey	Who answers	What counts as “using AI”	Unit and weighting
McKinsey	Professionals across functions and levels, by online questionnaire	Regular use of AI in at least one function of the organization	Respondent, weighted by country GDP
Stanford AI Index	Does not collect; compiles other datasets	Repeats McKinsey’s definition for adoption	Inherits the original source’s method
Deloitte	Leaders from director to board level, half IT and half business	Access to sanctioned tools, daily use, experiments in production, use of agents	Company, through one leader’s view
BCG	C-suite and senior leaders	Maturity classification made by BCG from the answers	Company, by maturity level
Eurostat	Companies sampled by national statistical institutes	At least one AI technology from a closed list	Company with 10 or more people, official sample

In which functions do companies use AI the most?

Over eight years of research, McKinsey has found IT and marketing and sales to be the functions respondents most often cite for AI use, and in 2025 knowledge management joined that group. Among the most reported use cases are content support for marketing strategy and customer service automation (McKinsey, 2025). Eurostat reaches a similar picture by another route: among European companies that use AI, 34.70% apply it to marketing or sales, the most common purpose (Eurostat, 2025).

For commercial leaders, the most useful data point in the 2026 edition is impact by function. Respondents most often link revenue gains to AI use in marketing and sales, followed by product development and software engineering, while cost reductions show up most in supply chain, service operations and manufacturing (McKinsey, 2026). Deloitte records leaders’ expectation that agentic AI will have the highest impact in customer support (Deloitte, 2026).

How many companies see AI impact on results?

The surveys seem to disagree, and the difference lies almost entirely in the question. At McKinsey, 37% of respondents attribute some positive EBIT impact to AI use in 2026, compared with 39% in 2025, and the group the firm calls high performers, those who attribute 5% or more of EBIT to AI and describe its value as significant, stayed at about 6% in both years. In the same survey, 80% say AI has improved their own productivity, and to me the gap between individual gains and company gains is the most important finding of the year (McKinsey, 2026). The method for turning that gain into a number finance recognizes is in AI ROI.

Deloitte asks which benefits the company has already achieved. Productivity and efficiency appear in 66% of answers and cost reduction in 40%, while revenue growth stands at 20%, against 74% who hope to get there. The share of leaders who say AI has a transformative effect on their company rose from 12% to 25% in one year (Deloitte, 2026). BCG, in turn, classifies companies by maturity and concludes that 7.5% are “future-built” and another 41% are scaling AI while outperforming laggards, together nearly half the market; in its 2025 edition, the same firm had found only 5% generating substantial value (BCG, 2026).

Source	Question about results	Number
McKinsey 2026	Respondent attributes some EBIT impact to AI use	37%
McKinsey 2026	Attributes 5% or more of EBIT and describes significant value	About 6%
Deloitte 2026	Company already grows revenue with AI	20%
Deloitte 2026	Leader describes a transformative effect	25%
BCG 2026	Company classified as “future-built” by BCG	7.5%
BCG 2026	“Future-built” plus companies scaling with superior performance	Nearly half

The 6%, 7.5% and 20% describe a minority that already measures substantial value, while BCG’s “nearly half” and McKinsey’s 37% include small effects and intermediate stages. My reading is that most companies capture productivity gains, and a minority between single digits and one-fifth already sees AI in revenue or profit.

How much are companies investing in AI?

The most cited AI Index figure measures money invested in AI companies, not spending by the companies that use AI. Adding mergers and acquisitions, minority stakes, private investment and public offerings, global corporate AI investment reached US$ 581.69 billion in 2025, up 129.9% from the year before, with US$ 344.66 billion in private investment (Stanford HAI, 2026). It shows where investors’ capital is going and says little about how much an average company should spend.

For the budget of companies that use AI, the consulting firms come closer. At McKinsey, 28% of respondents say their organization spends more than 10% of its total information and communication technology budget on AI, 60% expect to increase investment next year, and one in five says the company already limits use because of operating costs, including tokens (McKinsey, 2026). BCG estimates that AI spending went from about 1.7% of revenue in late 2025 to 3.3%, and that more than 80% of it sits outside the IT budget (BCG, 2026). At Deloitte, 84% of organizations are increasing investment (Deloitte, 2026). Spending outside IT is the figure that worries me most, because each team buys its own tool and nobody adds up the bill.

AI agents: how many companies have moved beyond the pilot?

In 2025, 23% of McKinsey respondents said their organization was scaling an agentic system somewhere and another 39% were experimenting, and in no single function did more than 10% report agents at scale (McKinsey, 2025). In 2026 the share scaling agents was about two in ten, and the average hides company size: among companies with more than US$ 1 billion in revenue, the share went from 27% to 40%, while at smaller ones it stayed at 22%. Agents at scale show up most in IT, knowledge management and software engineering, and in consumer goods and retail they show up most in marketing and sales (McKinsey, 2026).

At Deloitte, 23% of companies use agentic AI at least moderately, and 74% expect to reach that level within two years. Only 21% say they have a mature governance model for autonomous agents (Deloitte, 2026). BCG shows the same gap with another yardstick: 42% of companies expect to give agents real decision-making authority by 2030, and 5% have the full set of controls in place today (BCG, 2026). Designing the first agent is covered in AI agents, and the move into operations in AI from pilot to production.

How are companies managing AI risk?

In McKinsey’s 2025 edition, 51% of respondents from organizations using AI reported at least one negative consequence, and nearly one-third cited problems caused by AI inaccuracy. Companies now manage an average of four AI-related risks, compared with two in 2022 (McKinsey, 2025). In 2026, respondents from high performers more often report working to mitigate unauthorized or unintended actions by AI (McKinsey, 2026).

Among the leaders Deloitte surveyed, data privacy and security top the list of concerns at 73%, followed by legal, intellectual property and regulatory compliance (50%), governance capabilities and oversight (46%) and model quality, consistency and explainability (46%) (Deloitte, 2026). At Eurostat, among companies that considered using AI and did not, the most common reason was lack of relevant expertise (70.89%), followed by lack of clarity about legal consequences (52.52%) and data protection concerns (48.83%). The internal rules that answer these concerns are in AI usage policy.

What these numbers do not tell you

Almost every number in this article is self-reported. Whoever answers that AI contributed to EBIT is making an estimate, and the survey does not check the books. McKinsey surveys people, and the published method does not say how many different companies stand behind the 1,719 respondents; Deloitte and BCG survey executives, who tend to see their own AI program optimistically. All three consulting firms sell AI projects, which calls for attention to the sample without discarding the data.

The samples lean toward large and more advanced companies. At McKinsey, 36% of respondents work at organizations with more than US$ 1 billion in revenue, and Deloitte states that it sought a view of organizations at the leading edge of AI. Comparing years also requires care, because the sample changes from one edition to the next; McKinsey reports that it checked the comparison against the 552 respondents who took part in both 2025 and 2026 and found the same pattern.

High performers redesign workflows, measure impact and invest more, and the survey shows association without proving cause. Expectations also miss: among McKinsey respondents in 2025, 32% expected AI-related workforce reductions in the following year, and in 2026 only 14% reported that AI contributed to a smaller headcount, the subject of AI and jobs.

What leadership does with these numbers

This is the plan I propose to a board that will decide its AI budget with these reports on the table.

1. Pick one definition of use and measure your company with it. Ask how many people with access use the tool every day and in which processes, which is Deloitte’s yardstick, and compare with the previous year.
2. Separate individual gains from company gains. List the processes with AI and the indicator that should change in each one. The productivity gain that 80% of McKinsey respondents feel only reaches EBIT when someone decides what to do with the time freed up.
3. Add up AI spending across every team. Ask finance for the total of licenses, credits and consumption, including what marketing, sales and CS buy on their own.
4. Start with marketing, sales and customer service, measuring from day one. The surveys place these functions among those with the highest use and the most reported revenue gains. Record the baseline before turning on any agent.
5. Define agent governance before scaling. Write down what each agent does on its own, what waits for approval and who answers for it. With 21% mature governance at Deloitte and 5% full controls at BCG, whoever does this now gets ahead of the average.
6. Put training in the budget alongside the tool. Lack of expertise appears as the main barrier at Eurostat and Deloitte.

What HR and L&D should take from these reports

The shortage of prepared people is where Deloitte, Eurostat and BCG converge. Leaders surveyed by Deloitte name insufficient worker skills as the biggest barrier to integrating AI into workflows, and the most common response is educating the workforce to raise AI fluency (53%), well ahead of redesigning career paths (33%) (Deloitte, 2026). At BCG, seven in ten companies in the two most advanced groups are already retraining staff (BCG, 2026). For HR, generic fluency training handles the first step, and the gain in results depends on training each team in the process it is going to change, with that process’s approval rules and indicators.

The reports show broad access, growing budgets and few people prepared to carry the individual gain through to results. That is the work of the in-company applied AI training, built on the company’s own processes and data, with cohorts for the board, leaders and sales, marketing and CS teams. It delivers a prompt library by area and an agent architecture, with what each agent reads, what it executes on its own and who approves the rest.

Frequently asked questions

What percentage of companies use AI in 2026?

It depends on the definition. In McKinsey’s global survey, nearly nine in ten respondents say their organization regularly uses AI in at least one function. In the European Union’s official statistics, which ask about specific technologies in a sample of companies with 10 or more people, the figure was 19.95% in 2025.

How many companies get a financial return from AI?

In McKinsey’s 2026 survey, 37% of respondents attribute some EBIT impact to AI use, and about 6% attribute 5% or more. At Deloitte, 20% of companies already grow revenue with AI. The numbers vary because each survey defines results its own way.

In which functions is AI used the most?

IT and marketing and sales have appeared for years as the functions with the most use in McKinsey’s research, with knowledge management joining the group in 2025. At Eurostat, marketing or sales is the most common purpose among European companies that use AI, at 34.70%.

How many companies already use AI agents?

About two in ten McKinsey respondents in 2026 say their organization is scaling agents, rising to 40% among companies with more than US$ 1 billion in revenue. At Deloitte, 23% use agentic AI at least moderately, and 21% have mature governance for agents.

Can I compare McKinsey’s numbers with Eurostat’s or the Stanford AI Index?

With care. McKinsey and Eurostat measure different things, with different audiences and definitions, and the AI Index adoption figure is McKinsey’s own 2025 number. Compare each survey with its own previous edition and use Eurostat as the official statistics reference.

How do I prepare my team to use AI with the results these surveys describe?

Start with the leaders, who decide processes, approvals and indicators, and then train each team in the process it is going to change, using the company’s own data. Track daily use and the edit rate of drafts in the following weeks, beyond satisfaction with the session.