The AI Confidence Trap: Workers Say They’re Ready, but ROI Is Stuck at 57%

Why AI Confidence on the Job Is Outpacing Business Results

Companies are spending more than ever on AI, but the productivity payoff is still missing for a large share of enterprises. Gartner forecasts AI spending will reach $2.59 trillion this year, up 47% from 2025. Yet a Domino Data Lab study found that the share of enterprises whose ROI does not outpace their AI investment has held at 57% since 2025.

WalkMe’s AI at Work Pulse survey helps explain the disconnect. While 90% of employees say they feel confident using AI, about half report spending more time trying to get AI to complete a task than the task would have taken manually. More than half of employees say their manager expects more output in the same amount of time, and 33% admit they have pretended to be more skilled with AI than they actually are.

The pressure goes beyond individual managers. Dice found AI skills are now listed in 73% of tech job postings, KPMG says nearly half of companies are willing to pay an 11% to 15% salary premium for AI skills, and a PwC survey of 1,000 financial services executives found that 86% believe AI skills training is more important than an MBA for many new hires.

The confidence problem reaches senior leadership too. More than half of employees feel senior leaders are championing an AI strategy they do not fully understand. More than a quarter of managers have pretended to know how to use AI in a meeting or presentation, while 39% of senior leaders admit approving or purchasing an AI tool they do not know how to use. Employees, meanwhile, want guidance built into the tools themselves: one-third said AI worked well because guidance was available on screen while working.

Inside the Confidence Trap: Leaders, Managers and the ROI That Isn’t Showing Up

WalkMe’s Confidence Trap: Adoption Is Up, Outcomes Are Not

The survey points to a gap between self-reported confidence and actual business results. A 90% confidence rate sounds strong, but if half of employees say AI slows them down, confidence is not the right measure. The problem is that many organizations are tracking adoption and sentiment rather than whether AI reduces time, lowers cost or improves output. That helps explain why the Domino Data Lab figure has not improved since 2025.

Leadership Pretence Starts at the Top

The data suggests the credibility problem is not limited to frontline workers. More than 40% of senior leaders have pretended to understand their company’s AI strategy when communicating with employees, peers or board members, and 39% have approved or purchased AI tools they do not know how to use. If the people making procurement and strategy decisions are not proficient with the tools, it becomes harder to create a consistent tool policy or a credible rollout. That is consistent with the finding that 39% of employees receive conflicting messages about which tools they are allowed to use.

The Labor Market Is Raising the Stakes

The hiring data shows why employees feel pressure to appear AI-fluent. With AI skills listed in 73% of tech job postings, salary premiums of 11% to 15% on offer at nearly half of companies, and many executives valuing AI training over an MBA, workers have a strong incentive to signal competence. But if 33% admit to pretending they are more skilled than they are, some of that pressure may produce performative AI use rather than productive deployment.

Where AI Actually Worked: Guidance Inside the Tool

The strongest positive signal in the survey is simple: one-third of employees said AI was effective because guidance was available on screen while they worked. Employees also said they want better integration with existing software and role-specific best practices. Yet nearly a quarter have had no AI training at all, and 39% get conflicting messages about permitted tools. That suggests standalone training courses are less important than building support into the flow of work.

The ROI Gap: Measuring Work, Not Sentiment

The commercial consequence is that many companies have deployed AI without linking employee use to ROI. CIOs already have to defend AI spend, and the article argues they will increasingly have to explain business impact with hard data. Without measuring how individuals and teams actually use AI, companies cannot identify where work breaks down or where real efficiencies appear.

What CIOs and Executives Can Actually Do About the AI Adoption Gap

For CIOs, IT leaders and executive teams, the survey points to several concrete changes:

• Assign a named owner for AI enablement. The research says the conditions for AI outcomes have largely not been assigned to anyone; leaving it to IT or L&D alone is not enough.
• Move training into the workflow. One-third of employees said AI worked because on-screen guidance was available, so build role-specific help and integration into existing software instead of relying mainly on separate training programs.
• Publish one clear tool policy. With 39% of employees receiving conflicting messages about which tools are allowed, a single approved list and role-based use cases can remove friction.
• Measure actual usage and work time, not confidence. Since half of employees say AI can take longer than the manual task and 57% of enterprises fail to see ROI above spend, track where AI slows work and where it cuts time or cost.
• Require leaders to be users before buyers. With 39% of senior leaders approving or purchasing AI tools they do not know how to use, tie purchasing decisions to demonstrated internal use cases and require managers to use the tools they mandate.