There is no shortage of evidence that companies are adopting Artificial Intelligence (AI). Every quarter brings another survey showing that more organizations are experimenting with generative AI, putting models into production, giving employees access to copilots, or increasing their spending on the technology. At the same time, there is a growing body of evidence suggesting that the economic impact of all this activity remains much less significant than the level of investment would imply. Both observations can be true, and when the numbers are put side by side, the apparent contradiction becomes easier to understand.

Therefore, a more useful way to think about the current state of enterprise AI is not to ask whether companies are adopting it. They clearly are. The more important question is whether companies are changing their operations quickly enough to turn widespread access to AI into any meaningful business outcomes.

Table of contents

1. 89% of organizations are regularly using AI

McKinsey provides us with a useful starting point. In a 2026 survey, McKinsey found that 89% of respondents said their organizations are regularly using AI in at least one business function. 44% of respondents said that AI is now scaling across their entire enterprise, a figure that has increased from 38% a year earlier, suggesting that organizations are gradually moving beyond individual use cases and isolated experiments. [1] McKinsey & Company

AI is therefore no longer something that most large companies are simply considering. In many organizations, it has become part of day-to-day work. But the difference between 89% using AI and 44% scaling it across the enterprise is important. There is still a very large gap between adoption and industrialization.

And that gap becomes even more apparent in Gartner’s research.

2. Only 22% have successfully scaled AI across multiple business units

2026 research by Gartner found that only 22% of organizations have successfully scaled AI across multiple business units or adopted an AI-first approach. At the same time, 85% of functional leaders surveyed said they planned to increase AI spending in 2026, after allocating an average of 12% of their functional budgets to AI in 2025. [2] Gartner

The Gartner and McKinsey numbers should not be treated as directly comparable measurements. They use different samples and definitions of scaling, and neither survey is intended to provide a single authoritative measure of enterprise AI maturity. What they do, however, provide a useful indication of the shape of the market.

Investment and experimentation are moving considerably faster than the organizational changes required to make AI a genuinely scaled capability. In other words, the question is no longer whether companies are spending money on AI. The question is what that spending is actually producing in return.

3. 64% have moved beyond pilots

Accenture’s research provides another perspective. Its 2026 research found that 64% of respondents said their organizations had moved beyond pilots into production across multiple functions or had begun coordinated enterprise efforts around advanced AI. On the surface, that sounds considerably further along than Gartner’s 22%. However, only 7% of the organizations in Accenture’s research were classified as having the data capabilities it considers sufficiently mature to support scaled advanced AI adoption. [3] Accenture

This distinction matters because moving a model into production is not the same thing as creating the conditions for widespread AI adoption. An organization can have hundreds of successful AI use cases while still having fragmented data, inconsistent processes, and limited ability to connect those use cases together.

In practice, the technology may be moving through the organization considerably faster than the infrastructure and operating model beneath it.

Accenture’s finding that more than 80% of organizations sometimes delay, limit, or alter AI initiatives because of data-related risks makes this even more apparent. The problem is not necessarily a lack of AI capability. Increasingly, it is whether the rest of the enterprise is ready for it. [3] Accenture

4. Only 34% say AI is deeply transforming the business

Deloitte’s research reinforces this point. Its 2026 State of AI in the Enterprise research found that access to AI among workers had increased significantly, while 66% of organizations reported productivity or efficiency improvements from AI. Yet only 34% said they were using AI to deeply transform the business. [4] Deloitte

That difference is arguably more revealing than the headline adoption statistics.

Giving an employee access to an AI assistant can produce a measurable improvement in the amount of work they can complete. It does not necessarily change the process in which that work takes place. So, while the underlying business process remains largely intact, the productivity gain can still be real.

This distinction becomes even more important with AI agents. Deloitte found that only 15% of organizations have scaled orchestrated, cross-functional multi-agent adoption, while only 5% say their business processes are highly prepared for AI agents. [4] Deloitte

The technology is moving towards taking action. The enterprise is still working out how that action should fit into the way the business operates.

5. Only 37% report a positive impact on EBIT

McKinsey’s research adds an uncomfortable financial reality. Although 80% of respondents say AI has improved their individual productivity, only 37% report that AI has contributed positively to their organization’s EBIT. That 37% figure is essentially unchanged from the previous year. [1] McKinsey & Company

The gap between individual productivity and enterprise financial performance is not particularly surprising when viewed in this context.

The benefits of an AI tool accrue immediately to the person using it, whereas translating those benefits into financial results usually requires changes to processes, organizational structures, technology architecture, and sometimes the business model itself.

This is an important distinction for executives trying to understand the return on AI investment. A successful pilot can demonstrate that a model works. It does not necessarily demonstrate that the organization has figured out how to capture the resulting economic value.

That may be one of the reasons the current AI market can feel simultaneously overhyped and completely real. The productivity gains are showing up. The broader financial impact is much harder to see.

6. The top 20% of companies capture 74% of AI’s economic value

There are signs, however, that some companies are beginning to make that transition. PwC’s 2026 AI Performance Study found that almost three-quarters of the economic value from AI is being captured by just 20% of organizations. Those leading companies are also considerably more likely to use AI to pursue growth opportunities, reinvent their business models and redesign workflows rather than simply adding AI tools to existing processes. [5] PwC

The significance of this finding is less about the precise 74% figure and more about what it says about the distribution of value.

AI adoption does not appear to be producing uniform gains across organizations. Some companies are finding ways to turn AI into meaningful improvements in revenue and efficiency, while many others remain focused on deploying tools and proving individual use cases.

There’s a fairly important distinction here. The companies getting the most from AI do not necessarily appear to be the ones with the most AI projects. They are more likely to be the companies that are changing the way the business works because of AI.

That might sound like a subtle difference. It probably isn’t.

7. Only 15% have scaled multi-agent AI

The same pattern is emerging around AI agents. Deloitte’s research found that 15% of organizations have scaled orchestrated, cross-functional multi-agent adoption. Only 5% say their business processes are highly prepared for AI agents, while 72% say they lack unified, accessible data and 70% do not feel they can adequately trust and govern agents. [4] Deloitte

This is important because agents potentially change the economics of enterprise AI.

A conventional generative AI application largely assists a person with a task. An agent can typically execute a sequence of tasks, interact autonomously with enterprise systems, and make decisions inside of defined boundaries. That makes the opportunity considerably larger, but it also exposes weaknesses that were less important when AI was primarily being used as an employee productivity tool.

Data quality, permissions, process design, governance, and integration suddenly become constraints on what the technology can actually deliver.

The organizations that have been able to give employees a chatbot do not necessarily have the foundations required to let software execute an end-to-end business process.

8. Only 7% have the data foundations for scaled advanced AI

This may be the number that explains all the others.

Accenture found that just 7% of surveyed companies qualify as “data reinventors”, meaning they have progressed far enough in building AI-ready data capabilities to support scaled adoption of advanced AI. The other 93% are trying to push AI forward while still dealing with fragmented data, poor data quality, siloed systems, and a lack of business context. [3] Accenture

That creates an uncomfortable mismatch. The AI capability is advancing extremely quickly, while the enterprise underneath it is not necessarily keeping pace.

The same issue shows up in IBM’s research, although from a slightly different angle. Only 11% of technology executives surveyed said their organizations were completely prepared for the scale of AI agent deployment. IBM also found that organizations which built control directly into their AI systems deployed 16 times more AI agents than those relying on manual governance. Those organizations also reported 18% higher operating margins, although IBM’s analysis does not establish that governance itself caused those higher margins. [6] IBM Newsroom

The broader point is that governance, data, and architecture are becoming less of a supporting function around AI and more of a prerequisite for actually scaling it.

What these numbers are really telling us

Together, these numbers paint a picture that is more nuanced than either of the two dominant narratives surrounding enterprise AI.

Narrative number one is that adoption is happening everywhere, and that the transformation is already underway. Narrative number two is that companies are spending enormous amounts of money on AI without seeing meaningful returns. But neither of these two narratives really captures the complete picture on their own.

What appears to be happening instead is that organizations are moving through several different stages at different speeds. While it is true that AI experimentation has become widespread, employee adoption is accelerating, and production deployments are becoming increasingly common. Scaling those deployments across the enterprise, however, is considerably harder, and redesigning the business around AI remains even harder still.

Therefore, the technology is advancing faster than the required organizational changes necessary to actually capture its full value. That distinction is likely to become increasingly important over the next few years. The competitive question is moving away from whether an organization has access to the latest models or whether its employees are using generative AI. Those capabilities are rapidly becoming commonplace. The more difficult question is whether the organization can change its processes quickly enough to take full advantage of them.

The numbers suggest that this is where the real divide is beginning to emerge. The companies that are furthest ahead are not necessarily those with the largest number of AI pilots or the highest level of employee access. They are increasingly the companies that are able to connect AI to their data, redesign existing workflows around it, establish the governance required to dare to automate decisions, and measure the resulting impact on the economics of the business. There is also something to be said about the fact that not everything will require a powerful AI model. Some tasks should still be routed to the right resource at the right cost, where the orchestration of new vs. established technologies becomes a competitive advantage.

That is why the headline number for AI adoption can be both very high and somewhat misleading. Enterprise AI is already widespread. What remains relatively uncommon is the organizational transformation required to make widespread AI adoption economically significant.

The next phase of the market is therefore unlikely to be defined simply by how many companies adopt AI. It will be defined by how many companies manage to turn adoption into a different way of operating.

Sources

[1] McKinsey & Company, “The State of AI: Global Survey 2026”

[2] Gartner, “Survey Finds Only 22% of Organizations Have Successfully Scaled AI Across Multiple Business Units”

[3] Accenture, “AI-ready data: New rules of data for the advanced AI era”

[4] Deloitte, “AI Agents are Only the Beginning: Deloitte Survey Examines the AI Readiness Gap”

[5] PwC, “Three-quarters of AI’s economic gains are being captured by just 20% of companies”

[6] IBM Institute for Business Value, “New IBM Study Finds CIOs and CTOs Face Growing AI Control Gap”