How leaders can tell the difference between AI adoption, productivity, and organizational value
AI adoption is becoming an increasingly poor measure of AI success.
An employee can use AI every day without becoming more productive. A team can become more productive without creating measurable value for the organization. And a company can roll out AI broadly while inadvertently making some employees' jobs harder.
That distinction matters because leaders and employees don't always agree about what they're getting from AI.
Korn Ferry's 2026 Workforce research found that 79% of CEOs said AI had improved efficiency, compared with 51% of individual contributors. Gallup found that 65% of employees at organizations using AI said it improved their individual productivity and efficiency, yet only 14% strongly agreed that AI had changed how work gets done across their organization.
Those findings aren't necessarily contradictory.
Using AI, becoming more productive with AI, and creating organizational value from AI are three different things.
Three levels of AI value
I think about AI impact at three levels: adoption, work improvement, and organizational value.
- Adoption: Are people using it? Usage rates, licenses activated, and training completion tell you whether people are adopting a tool. That's useful information, particularly early in an implementation. But adoption isn't an outcome. An employee using AI five times a day may be saving hours of work. Or that employee may be spending almost as much time checking and correcting the output as they previously spent doing the work. A usage dashboard won't tell you which is happening.
- Work improvement: Is AI making the work better? This is where measurement needs to move from the tool to the work. Take a specific workflow and establish a baseline. How long did it take before AI? How long does it take now? What happened to quality? Are there fewer errors or more? How much rework is required? If the tool saves one employee 30 minutes but creates 20 minutes of verification work for someone else, the organization hasn't saved 30 minutes. Some of the work simply moved.
- Organizational value: Did the improvement matter? Did faster work create additional capacity? Did quality improve? Did costs decline? Did clients get faster responses? Did employees have more time for higher-value work? Not every benefit needs to be converted into a single ROI percentage. But leaders should be able to connect an AI investment to the business problem it was supposed to help solve.
Why leaders and employees may see AI differently
The 28-point gap between CEOs and individual contributors in the Korn Ferry research should get leaders' attention.
It doesn't necessarily mean executives are overly optimistic or employees are resistant to change. They may simply be seeing different parts of the picture.
Senior leaders see investment, strategic capability, cost savings, and anticipated productivity. Employees experience AI inside the actual work. They see what gets easier, what gets harder, what needs to be checked, and what still requires human judgment.
Both perspectives matter.
The goal isn't to decide which group has the right answer. It's to understand why their answers are different.
Does 95% of generative AI really fail?
You've probably seen the headline that 95% of generative AI projects fail. That's not quite what the underlying research said.
The statistic comes from a preliminary 2025 MIT Project NANDA report. Researchers found that 95% of the organizations studied had not yet seen measurable profit-and-loss impact from their generative AI initiatives.
That's worth paying attention to. But "no measurable P&L impact" isn't the same as "no value," and it isn't the same thing as saying 95% of AI projects failed.
A large 2026 NBER study adds some context. About nine in ten senior executives surveyed reported no impact from AI on employment or labor productivity during the previous three years, while those same executives expected larger productivity gains over the next three years.
The evidence is still developing. More importantly, adoption, individual productivity, organizational productivity, and financial return aren't interchangeable measures.
When AI saves time, where does the work go?
This may be one of the most important questions leaders can ask.
Imagine a recurring task that used to take four hours. AI reduces it to two. It's tempting to record that as a two-hour productivity gain. But what happened to those two hours?
Did the employee take on higher-value work? Did workload simply increase? Did someone else inherit responsibility for reviewing the AI-generated work? Did the organization actually capture additional capacity?
Whenever AI is supposed to be creating significant efficiency, ask three questions:
What work disappeared?
What new work appeared?
What happened to the capacity that was created?
Until you can answer all three, you may know that a task got faster without knowing whether the organization became more productive.
Don't mistake friction for resistance
When employees say an AI tool is creating more work, leaders should be careful about interpreting that as resistance to change.
Employees may be identifying a poorly designed workflow, inadequate training, unclear accountability, additional verification work, or a legitimate limitation of the technology.
Resistance is something leaders try to overcome. Friction is something they should investigate.
Managers are particularly important here because they're close enough to the work to see that friction. Gallup reports that only 36% of employees in organizations integrating AI strongly agree that their manager actively supports their team's use of AI. Employees who feel strongly supported are much more likely to use AI frequently and report that it has changed how work gets done.
But support means more than encouraging adoption. Managers need to help people decide where AI is useful, where human judgment still matters, what good work looks like, and what to do when the technology gets something wrong.
Giving people access to AI is a technology decision. Changing how people work with it is a leadership challenge.
Psychological safety affects the quality of your AI data
There's another issue leaders need to consider. What happens when senior leadership has already declared an AI implementation a success? How comfortable is an employee saying it isn't working?
Amy Edmondson's research on psychological safety has demonstrated its importance to learning behavior in teams. That's particularly relevant to AI because organizations are still learning where these tools work well and where they don't.
Employees need to be able to say, "This saved me an hour." But they also need to be able to say, "This created more work," "I don't trust this output," or "I made a mistake using it."
If employees believe leadership only wants to hear AI success stories, leadership will get success stories. It just may not get accurate information.
In that sense, psychological safety isn't only an employee-experience issue. It's part of how an organization gets reliable information about whether its AI investment is working.
Five questions leaders should be able to answer about AI
- Where is AI actually changing work? Don't just measure who has access or completed training. Identify which workflows have changed.
- Is the work getting better? Measure time, quality, errors, rework, capacity, or whatever outcome matters for that particular work.
- Where did the work go? Know who is checking, correcting, and taking responsibility for AI-assisted work. Efficiency in one part of a workflow can create work somewhere else.
- Are managers equipped to lead AI-enabled teams? Managers should be able to help employees decide when AI should be used, when it shouldn't, and where human judgment remains essential.
- Are employees telling you the truth about what's working? Can someone say an AI tool made the work harder, produced a bad result, or isn't useful without being labeled negative or resistant to change?
If leaders can't answer these questions, the organization may know how much AI it has deployed without knowing how much value it has created.
Where to go from here
I'm not skeptical of AI. I'm skeptical of conclusions that run ahead of the evidence.
There are real productivity gains showing up in the research, particularly at the task level. There are also organizations reporting little measurable financial impact so far. Both can be true.
AI implementation may start as a technology decision, but creating value from AI quickly becomes an organizational challenge.
So instead of asking only how many employees are using AI, I would ask senior leaders, managers, and employees the same question:
Where is AI making our work better, where is it making our work harder, and how do we know?
The differences in their answers may tell you more than an adoption dashboard does.
Don't just measure whether people are using AI. Measure what happened to the work.
If your leadership team and the people doing the work seem to be telling two different stories about AI, that gap is worth examining before you invest in the next tool. DILAN's organizational development work helps leaders align people, leadership, and systems so change translates into sustainable performance.
If you would like to talk through what that looks like in your organization, you can reach me at office@dilanconsulting.com.
