The State of AI in 2025: A Gap Between Adoption and Impact
Two figures from McKinsey’s 2025 survey present a curious juxtaposition. On one hand, 88% of organizations report using AI in at least one sector of their activity. On the other, only 39% can point to a measurable effect on financial results. This disparity highlights the current state of enterprise AI at the end of 2025.
These statistics are derived from McKinsey’s State of AI report, released on November 5, 2025. The report draws insights from 1,993 respondents across 105 countries, collected during the Nordic summer. The report’s authors note, “88% say they regularly use AI in at least one business function, up from 78% a year ago.” It’s important to remember that these are self-reported numbers, not an audited tally from every company worldwide. They describe usage somewhere within the organization, not necessarily transformative usage.
Adoption Is Not the Same as Impact
It’s easy to misinterpret near-universal adoption as near-universal transformation, but the survey suggests otherwise. The report states, “But at the company level, the majority are still at the experimentation or piloting stage.” Many teams run a model in one area; far fewer have restructured the business around it.
McKinsey’s estimates make the gap concrete. Only about a third of organizations have begun to mainstream AI enterprise-wide, while nearly two-thirds have not. The proportion of people saying AI is fully scaled is only 7%. Regarding profits, around 39% of respondents attribute company-level EBIT impact to AI, with most estimating this figure below 5%. The authors emphasize, “Significant impact of the use of AI on business results remains rare, although our survey results suggest that thinking big can pay off.”
The Pilot Trap
This trend is not isolated to a single survey. An independent report from MIT’s NANDA Project, led by Aditya Challapally in August 2025, reached a similar conclusion. The report found that “the 95% failure rate for enterprise AI solutions represents the clearest manifestation of the GenAI divide.” This figure is disputed and based on one study, with “failure” narrowly defined as the absence of rapid impact on revenue or profit and loss. This aligns with McKinsey’s more muted conclusion.
Why do many efforts stall at the pilot stage? A pilot project can succeed with the enthusiasm of a single team and a modest budget. Scaling requires redesigned processes, retrained staff, leaders willing to take ownership, and a tolerance for disruption that a contained project never tests. The friction likely lies in these areas, not in the model itself.
What the Small Minority Does Differently
About 6% of respondents meet McKinsey’s criteria for “high-performing AI,” meaning AI generates 5% or more of EBIT, plus what the report calls significant value. What sets them apart is not just smarter algorithms but the depth of rebuilding around the tools. The report notes, “Half of the top AI performers intend to use AI to transform their business, and most are rethinking their workflows.”
Workflow redesign is a recurring theme. The most successful companies don’t just automate existing processes; they rethink them from the ground up, integrating AI into workflows and decision-making. This group is also three times more likely to use AI for transformative change rather than reducing efficiencies. While a single year of survey data can’t distinguish cause from correlation, the survey suggests greater ambition leads to greater gains.
McKinsey reports that about 80% of respondents set efficiency as their AI goal, while top performers also seek growth and innovation. Doing existing things slightly less expensively is a lower ambition than doing entirely different things, and greater ambition is associated with greater gain.
The Question That the Investigation Leaves Open
Top-performing companies are dedicating a larger portion of their digital budget to AI, which may indicate an organization already ahead or one aiming for transformation. The report doesn’t resolve this issue. The gap between 88% adoption and 39% measurable impact isn’t a sign of AI failure; it’s a sign that most organizations are still in early stages. Purchasing a tool is the cheapest part, while rebuilding work processes is where a small minority have progressed and most have not. Next year’s survey will reveal if the gap between adoption and impact is narrowing.
For more details, you can read the full report here.
“`
