AI according to McKinsey in 2026: adoption is established, value still needs to be built

Situation, complication, recommendation

Situation. McKinsey has measured AI adoption for nearly ten years. In 2026, it is almost universal and AI agents are beginning to scale in large companies ^[1].

Complication. Individual gains do not translate into financial results: the share of respondents reporting an impact on EBIT is not increasing (37 %), and the share of “high-performing” companies remains at 6 % ^[1].

Recommendation. Approach AI as an organisational transformation — a few high-stakes domains, processes redesigned end to end, trained leaders and impact measured in EBITDA — rather than as a deployment of tools.

1. Almost universal adoption, with agents arriving

The state of AI in 2026: On the road to ROI survey, published on 25 August 2026 by McKinsey and QuantumBlack, AI by McKinsey (1 719 participants across 97 countries, May–June 2026) shows adoption deepening ^[1]:

• nearly nine in ten respondents report regular use of AI in at least one function;
• 44 % say AI is scaling across the entire company (38 % a year earlier), and 56 % use it in at least three functions (51 % in 2025);
• in organisations with more than one billion dollars in revenue, 40 % are scaling AI agents (27 % in 2025), compared with 22 % — unchanged — in smaller organisations;
• 32 % have decided not to buy at least one piece of software because it could be developed internally using coding agents;
• one in five respondents says that AI operating costs, including “token” costs, are already limiting its use.

2. The paradox: individual productivity does not show up in EBIT

80 % of respondents say AI has improved their productivity and 50 % say it helps them make better decisions. But only 37 % attribute a contribution to EBIT to it, as in 2025, and the share of high-performing companies — those attributing at least 5 % of their EBIT to AI and considering its impact significant — remains around 6 % ^[1].

The article “From adoption to impact: Three horizons of AI transformation” (McKinsey Quarterly, July 2026), based on a survey of 750 employees and executives, sheds light on this gap ^[2]. 70 % of respondents feel personally ready to work with AI, but only 27 % of executives consider their organisation ready for the necessary changes. Organisational readiness accounts for 48 % of the gap between executives creating value with AI and the others, compared with 25 % for individual readiness. This mismatch was already apparent in early 2025 in the United States, according to McKinsey’s “Superagency in the workplace” report: executives estimated that 4 % of employees used generative AI for at least 30 % of their daily work, whereas 13 % of employees reported doing so themselves ^[3].

3. Three maturity horizons — value is concentrated in the third

The “three horizons” framework, proposed by McKinsey ^[2], distinguishes: tooling (individual access to general-purpose tools), automation (cross-functional processes automated at scale) and reinvention (roles, processes and the operating model redesigned around AI). Only 11 % of surveyed executives place their organisation in the third horizon — but nearly half of them report enterprise-wide value creation.

4. What the successful 6 % do

According to these two surveys ^[1][2], organisations deriving significant value from AI share five characteristics:

1. They aim for growth, not just efficiency. They are 3.3 times more likely than the others to want to fundamentally transform their business within three years.
2. They redesign processes. Nearly three-quarters of high-performing companies report having done so extensively (55 % a year earlier), compared with a quarter of the others. Even in the first horizon, process redesign is associated with a 5.3 times greater likelihood of reporting value creation (32 % versus 6 %).
3. Their leaders get involved. They are twice as likely to report visible leadership engagement and impact measurement processes; in the second horizon, leadership teams that are highly comfortable with AI are associated with a 3.9 times greater likelihood of value creation (35 % versus 9 %).
4. They invest more. They are more than twice as likely to allocate more than 15 % of their digital budget to AI.
5. They manage risks. They do more to limit technical vulnerabilities and unauthorised actions by AI systems. According to McKinsey’s “State of AI trust in 2026” survey (around 500 organisations, December 2025–January 2026), nearly two-thirds of respondents cite security and risks as the main barrier to scaling agents, and average responsible AI maturity has risen only from 2.0 to 2.3 over a year ^[4].

5. Employment: forecasts revised downwards, expectations upwards

14 % of respondents report an AI-related reduction in headcount over the past year, less than half of the 32 % who anticipated it in 2025; two-thirds report little or no change. For the coming year, 39 % anticipate a reduction and 43 % no significant change, while only 13 % are worried about their own careers ^[1]. Trust is the key factor across all horizons: employees who doubt their organisation’s support are 1.5 times more likely to be anxious about AI-related changes ^[2].

6. Our takeaways

For an executive or investment fund, four implications:

1. Focus: choose two or three high-stakes domains where costs or revenues are concentrated, and redesign their processes end to end, rather than multiplying pilots.
2. Measure: establish a baseline before deployment and track impact in EBITDA and cash, not just hours saved.
3. Train the leadership first: the executive team’s command of AI is one of the strongest predictors of value.
4. Govern agents: identities, permissions and traceability must precede scaling.

Our AI R&D

Limitations. These surveys rely on online self-reporting; The state of AI weights responses according to countries’ share of global GDP, and the “three horizons” panel is not representative of the market. They establish associations, not causation.