Using AI every day does not mean your business gets value from it. When I ask leaders how AI is going in their company, they usually give me a usage number. A new Wharton study shows why that number tells you very little about results.

Wharton Human-AI Research and GBK Collective published the study in October 2026 under the title The Adaptive Advantage: Who’s Winning with AI and How. The researchers asked more than 1,400 leaders, managers and employees in large companies about their AI work. Usage has grown fast. 62% of leaders and managers now use AI every day, up from 46% a year ago. 80% say AI gives them a positive return. Yet only 28% report a strong return. And just 7% of companies succeed in all three areas the study looked at: business value, changed workflows, and people who support the change.

So the question for leaders has changed. Counting how many people use AI is no longer enough. What matters now is how much value AI creates, and what stops most companies from getting more of it.

 

 

 

The problem: lots of AI, little change in results

 

I see the same pattern in my client work. Companies roll out licences, people use the tools, and teams save some time here and there. On paper, the AI programme looks like a success. But when leaders look at the business, they struggle to name a result that AI has clearly improved.

The Wharton numbers confirm this. AI is everywhere, but the large gains stay with a small group. That gap usually has little to do with the technology. It comes from how companies measure success, how they design work, and how well leaders and employees understand each other.

 

Companies measure the easiest thing

The most common way to measure AI value is hours saved. 40% of companies use it. Companies measure extra revenue least often, at only 27%. This matters. Among departments that track at least one growth measure, such as revenue, output or customer satisfaction, 35% report a strong return. Among departments that only track efficiency, the figure drops to 13%.

This does not prove that measuring growth causes better results. But it shows a clear link. If you only look for saved hours, you will probably only find saved hours. And saved hours rarely show up in the profit and loss statement unless someone decides how to use them.

 

Companies add AI to old ways of working

73% of leaders and managers say AI has changed their workflows a lot. Only 28% say AI is fully part of their core processes. There is a big difference between a person who sometimes uses ChatGPT inside an old process and a team that has rebuilt a workflow around AI. In the second case, the team has thought again about tasks, decisions, handovers and the points where people need to check the work.

Just 20% of departments pass all three of the study’s tests for this kind of change: AI in core processes, use of AI agents, and workflows that have changed a lot. Of those departments, 67% report a strong return.

 

Strong results come from new ways of working

There is one more finding I find important. Departments with strong returns did not cut more jobs or stop hiring more often than others. Instead, they changed how they organise work. They adjusted management structures, tracked AI use, and included AI in performance reviews. The value came from new ways of working, much less from fewer people.

 

Leaders and employees see a different picture

78% of leaders and managers say AI training is good enough. Only 56% of employees agree. The same gap shows up for a clear AI vision and for how much workflows have changed. Employees are not against AI. 73% feel positive about its effect on their own role, and 84% of those who use it say it saves them time. The issue is that leaders think the rollout is going better than their teams experience it. When leaders base their decisions on that picture, they invest in the wrong places.

 

 

Practical recommendations

 

The study points to three conditions that need to come together: value, workflows and people. In my work with leadership teams, I turn these into three simple questions. If you cannot answer all three for an AI project, you know where the gap is.

• Value: What has improved in the business because of AI, beyond hours saved? This question forces you to connect AI to a business outcome. It could be faster delivery to customers, more sales conversations, fewer errors or better service.
• Workflow: Which workflow have we rebuilt, rather than simply adding AI to it? This question shows whether AI has changed how your teams work, or whether it sits on top of the old process.
• People: Do our employees experience the same progress that leadership reports? This question checks whether your view from the top matches what happens in the teams.

The study makes one thing clear. These three areas belong together, and companies that treat them as separate projects get less out of AI.

Here are three recommendations you can apply today:

 

1) Define value before you start

Decide what success looks like before an AI project begins. Time saved can be part of it, but it should not be your only measure.

• Choose one business outcome per use case, such as shorter response times, more output per team or happier customers.
• Decide in advance what happens with the time AI saves. Will teams take on more customers, improve quality or work on new tasks?
• Know your full AI costs. Only 40% of companies feel very sure they can track their total AI spending. You cannot calculate a return without knowing the cost.

2) Pick a few workflows and rebuild them properly

Spreading AI thinly across the whole company rarely creates large gains. Choosing a small number of important workflows and changing them fully works better.

• Start with workflows that matter for the business and take a lot of time today.
• Look at the whole process, including handovers, approvals and decisions. Then ask where AI should do the work and where people should.
• Choose on purpose how much freedom AI gets. In the study, 67% of departments use AI agents, but only 17% let them act without approval for every step. More freedom is not always better. What helps is a clear choice about who decides what and how you control risks.

3) Close the gap between leaders and teams

You will only know how well AI works if you ask the people who use it every day.

• Ask employees directly whether training, tools and guidance help them in their daily work.
• Compare their answers with what leadership believes. Where the views differ, you have found your next priority.
• Make one person clearly responsible for AI results. In the study, companies with their own Chief AI Officer report strong returns far more often than companies that add AI to an existing leadership role (58% compared with 24%). Again, this is a link and no proof. But it shows that clear ownership matters.

 

A note on the study

The Wharton researchers surveyed large US companies, and the people taking part rated their own results. The returns are their estimates, and nobody checked them against financial figures. European mid-size companies work in a different setting. Still, the patterns match what I see with clients in Belgium, Germany, the Netherlands and Switzerland.

The mood has also shifted. Fewer leaders describe themselves as excited about AI than last year, and more describe themselves as cautious or sceptical. I see this as a healthy sign. Companies have moved past trying AI out. They now look more closely at value, cost and risk.

 

 

 

What corporate leaders can do now

 

 

• Ask the three questions about value, workflow and people for your three most important AI projects.
• Add one growth measure to every AI use case, next to time saved.
• Decide what happens with saved time before a project starts, so the gain shows up in business results.
• Choose two or three core workflows and rebuild them fully instead of adding AI to many tasks.
• Set the level of AI freedom on purpose, with clear rules for when people check and approve the work.
• Ask your employees how they experience the AI rollout, and compare it with what leadership believes.
• Track your full AI costs, including licences, tools and computing power.
• Name one person who is responsible for AI results across the business.

If you want to move from AI usage to measurable business results, my AI Use Case & Strategy Sprint helps leadership teams choose the right use cases, rebuild the workflows that matter and connect AI to clear business outcomes.

Source: Wharton Human-AI Research & GBK Collective, The Adaptive Advantage: Who’s Winning with AI and How, October 2026.