What McKinsey, Gartner, BCG, and Sequoia say about AI adoption - and what it means for you

Over 2025, AI stopped being an innovation-department experiment and became a line in the annual plan.

But there is a gulf between "adopted" and "profitable," and major studies have already measured it. McKinsey: 88% of organizations use AI regularly, while 39% see a company-wide profit impact. Gartner: more than 40% of agentic AI projects will be shut down by the end of 2027. BCG: only 5% of companies capture durable AI value at scale.

This is not a reason to roll back AI - it is a map of where value is being lost.

Below are the most common business questions about AI adoption, each answered with the cited figure from the primary source (McKinsey, Gartner, BCG, Sequoia) and the mechanism that separates the 5% who got results from the 60% who did not.

All figures are current as of the verification date 28 June 2026, with a link to the report. The AI market changes within weeks, so check the primary source before deciding.

Almost everyone has adopted it: in McKinsey's survey (n≈2,000 respondents, fieldwork in summer 2025), 88% of organizations use AI regularly in at least one business function, up from 78% a year earlier. 62% are experimenting with agents.

So the lag isn't in the fact of adoption itself.

Only 39% of respondents report company-wide profit impact (EBIT) from AI, and most of them put that impact below 5%.

Only 7% have fully scaled AI across the enterprise.

In other words: everyone deploys, a handful extract value.

The right question is not "Are we on time?" but "Is AI being applied to a process where it delivers a measurable effect?"

Because most are early experiments and proof-of-concepts launched during the hype and often applied to the wrong task. Gartner explicitly forecasts that more than 40% of agentic AI projects will be canceled by the end of 2027.

Three reasons: rising costs, unclear business value, and insufficient risk control. Add to that "agent washing": out of thousands of vendors claiming agentic solutions, Gartner estimates that only about 130 offer truly agentic capabilities, with the rest being renamed assistants, chatbots, and RPA.

Takeaway for the buyer: filter not by the word «agent» in the pitch, but by which verifiable business result the project delivers.

How to choose a task and pilot so you do not end up in that 40% is covered in a separate analysis how to avoid failure and what slows AI adoption.

Putting the four reports together yields a single picture.

Almost everyone has adopted it (88%), few see profit (39%), only a small share capture value at scale (5%), and the main lever is process redesign with focus on a few use cases.

This is the KT.Team stance: small strong teams take one process, rebuild it around AI and answer for the business result, not for the fact of deployment.

The practical decision behind these numbers is whether to build AI capability in-house or buy a ready-made solution.

The answer depends not on trends, but on how unique the process is and how often it will change.