The AI adoption gap: what the 2026 research shows

IBM’s 2026 CEO study: the key findings

IBM’s Institute for Business Value, working with Oxford Economics, surveyed 2,000 CEOs and equivalent senior leaders across 33 geographies and 21 industries between February and April 2026. The findings show a disconnect between leaders’ confidence in employee readiness and reported use:

• 25% of the workforce uses AI regularly, according to the surveyed leaders.
• 86% of CEOs believe their employees have the skills to collaborate with AI.
• 83% say AI success depends more on people’s adoption than on the technology.

Leaders also expect substantial training needs: 53% of employees will need upskilling for their current roles, and 29% will need reskilling for different roles between 2026 and 2028. Read IBM’s findings and methodology.

The 86% figure measures CEO opinion about employee skills; the 25% figure measures leaders’ estimates of workforce use. They are different measures, rather than a direct comparison of employee access and adoption.

Why employees may not use AI

MindStudio’s analysis identifies several practical barriers: tools that do not fit specific jobs, a lack of role-specific guidance, concerns about trusting outputs, and insufficient support after rollout. Its central argument is that organizations need to connect AI capabilities to the work employees actually perform.

Employee research adds detail. In Gallup’s February 2026 survey of 23,717 U.S. employees, non-users in organizations that make AI available reported these barriers:

• 46% preferred their existing way of working.
• 43% cited concerns about data privacy, security, or compliance.
• 39% did not believe AI could help with their work.

These figures concern non-users with AI available at work, not all employees. Source: Gallup’s research on adopters and holdouts.

Workflow fit and manager support are associated with higher use

Gallup also found substantial differences in frequent use, defined as a few times a week or more:

• 88% versus 55%: employees who strongly agreed AI fit their existing systems and processes, compared with those who did not strongly agree.
• 78% versus 44%: employees who strongly agreed their manager actively supported AI use, compared with those who did not strongly agree.

These associations do not establish causation, but they identify conditions linked to more consistent use. See Gallup’s comparisons.

What low adoption means for AI investment

For a company paying per seat, unused licenses are one direct cost. The broader question is whether the investment improves completed work enough to justify software, implementation, training, and review costs.

License counts and training attendance cannot answer that question on their own. An employee may attend a workshop without using the tool afterward. Another may use AI daily but spend enough time correcting its output to offset the initial time saved.

A useful evaluation therefore separates three measures:

• Access: how many employees have an approved tool available.
• Adoption: how many use it repeatedly for relevant work.
• Value: whether completed tasks improve in speed, quality, or cost.

Time saved should include preparation and checking. It should also be distinguished from cash savings: freeing an hour of employee capacity does not automatically reduce company spending by an hour’s salary.

What organizations can do to improve adoption

For teams planning an AI rollout, our recommendation is to make the implementation specific enough to evaluate:

• Define the task. Identify its inputs, required output, frequency, and current completion time before choosing how AI will help.
• Provide guidance for that role. Show employees how to use approved tools on their own tasks, including how to check the result.
• Set clear operating rules. Specify permitted data, required review, and who handles mistakes or exceptions.
• Assign ongoing support. Give employees a named person to contact when an attempted use fails or needs adjustment.
• Measure repeat use and results. Track whether employees return to the workflow and whether its output meets the agreed standard.

RideAlong addresses the discovery and guidance parts of this process. It observes activity, identifies recurring workflows, and guides people through using AI on the task in front of them. Teams can use those suggestions to decide where to focus their adoption efforts.