Despite widespread adoption, organizations are reaching an AI plateau because they treat it as a typical technology implementation rather than one requiring fundamental change.
Implementing AI in various forms has been all the rage in enterprises in the last few years, but evidence that all the investment is paying off is scant.
In fact, it appears that organizations are hitting an AI plateau, where AI investments have made some gains, but productivity is leveling off.
AI deployments are on the rise. According to the McKinsey State of AI 2025 report, 88% of surveyed organizations now use AI for at least one operational function. This is up from 78% in the previous year's survey. However, this investment is not paying off, at least for now.
Just under 40% of organizations report that AI investments have generated any impact on profitability. In most cases, this amounts to less than 5% of total EBIT [earnings before interest and taxes], and just 6% of organizations for which AI investment has had more than a 5% impact on EBIT.
Agentic AI use is growing, according to the report, with 62% of organizations experimenting with AI agents. While 23% of these say they are scaling AI agents in at least one function, there's no single function where the "scaled or fully scaled" share exceeds 10%.
Hidden barriers to AI transformation
All this suggests that organizations use AI widely, but these deployments are not paying off yet.
Organizations are hitting an AI plateau because they're deploying AI technically, but they're not doing it in a way that's visible to the effects of the changes, said Chris Willis, chief design officer and futurist at Domo.
Organizations believe they can buy and deploy the tools and assume that the successful transformation will just happen. However, there are many hidden barriers to transformation, including organizational inertia and a culture that doesn't foster innovation, he said.
One of the biggest factors in organizations hitting the AI plateau is the mistaken assumption that everyone is a natural innovator or will have the freedom to innovate in ways they haven't been asked to before, Willis said.
"AI leaders say, 'We're just going to give everybody all these tools, and then they're going to figure it out,' but that's like magical thinking," he said. "Did they hire innovators? [Was the guy] in accounting hired because they like to work outside and think outside the box?"
These employees may or may not be creative or innovative thinkers, but organizations are task-based, functional hierarchies, and people are hired and incentivized to work within them, Willis said.
Organizations are hitting an AI plateau because they are deploying AI out of FOMO and the pressure to solve problems, without clearly defining the problems, he said.
If you're trying to transform, you have to transform more than one thing, you have to redesign how decisions, workflows and accountability are handled.
Chris WillisChief Design Officer and Futurist, Domo.
"If you're trying to transform, you have to transform more than one thing. You have to redesign how decisions, workflows and accountability are handled," Willis said.
AI leadership gap
Organizations are hitting an AI plateau and are struggling to get value from their investments, said Mike Kazmier, head of AI at Banyan Software.
There are two reasons why AI programs are flailing, he said. One is that they lack the leadership to marshal them through the deployment and make it a business priority. The other is that they are skipping at least one of the six pillars of true transformational change.
"Too many companies think that just buying the technology gets you the instant reward without addressing the operating model changes, the people's skill improvement and the data that's required, because data is the fuel of all AI," Kazmier said.
To avoid hitting an AI plateau, organizations need both a solid strategy at the top and a robust governance system at the bottom that manages risk and ensures value, he said.
AI proliferation across the enterprise is accelerating at a speed that leaders are finding difficult to manage, said Fern Halper, founder of the AI Foundations Group.
There's pressure on executives to move on AI quickly, Halper said. They must acknowledge this reality while simultaneously dealing with their organizational, data and governance readiness.
The pace of AI change has led some executives to wonder if they can keep up, she said. For example, the former CEO of Coca-Cola, James Quincey, stepped down in March amid the company's transformation to AI and digital growth.
However, organizations are caught between the AI transformation hype and the reality, where many are stuck in experimentation and pilot mode, largely because they have not thought enough about the foundations, Halper said.
"Organizations have stalled out because they were trying to use easy-to-use tools, but they hadn't put their data foundation, governance or skills foundation in place, and they can't get any further because they've plateaued out," she said.
Not just another tech deployment
Dan Leiva, founder at CXAmplify, agreed that one reason for organizations hitting an AI plateau is that they roll it out like any other technology project, treating the rollout as the end of the project.
"It's not owned after the rollout, which is the most important part because there's drift, there's changes in the upstream data, there's no watching it after it goes," Leiva said. "The big difference with agentic AI is that it doesn't stop when it rolls out. You've got to watch it after it rolls out."
Organizations are now assessing how to deploy AI successfully and avoid AI plateaus.
Willis said that one way to move forward is to come back to the fundamentals of technology deployments. Organizations should pick a few use cases that are well-understood, have value and already have unified governed data.
"You can then embed AI into the workflow rather than creating a separate siloed tool, and you can keep humans doing what humans do very well -- which is bringing judgment to the process," he said. "Then you can measure the impact and see if you can repeat that outcome."
The problem with AI is that many organizations are jumping straight to the prototype stage without fully considering if they are solving the right problem or should be solving that problem at all, Willis said.
"For example, for automation, people aren't asking if they should automate this, so they're creating new problems, or they're using new and expensive technology to solve a thing that they've solved already," he said.
The AI transformation and the question of productivity are not new stories, Willis said.
A generation ago, organizations began to furnish all employees with PCs, promising to greatly improve productivity -- even if most people at the time didn't know what to do with them.
There was a huge amount of investment, hype and FOMO from the organizational side, and a huge amount of anxiety on the employee side, Willis said. But it took years for organizations to realize productivity gains.
"Today, we can't imagine doing work without PCs. But it takes time because we have to understand the technology and that there are other parts around the technology," he said. "There are missing technologies, missing cultures, missing role definitions and missing education. You need all this and more before you can really leverage what's happening."
Jim O'Donnell is a news director for TechTarget, where he covers IT strategy and enterprise ESG.
