Leaders think AI makes productivity go up
When factories first adopted electricity in the late 1800s, many owners did something that looks strange in hindsight. They removed the steam engine that drove the factory’s central shaft, installed an electric motor in the same spot, and left everything else alone. Things like belts, pulleys, and the rest of the floor layout stayed exactly where they were. The Stanford economist Paul David later showed that electricity took roughly four decades to register in productivity statistics. The factories experienced true productivity gains when manufacturers rebuilt their factory floors around a small motor at each machine.
Many companies are repeating that mistake with artificial intelligence today by bolting it onto legacy processes, without considering how they should be reimagined.
To learn more about this issue, my company recently commissioned a survey of 1,000 process decision makers and 5,000 employees at large organizations. Among the leaders, 89% said AI makes their teams more productive. Among the employees, only 64% agreed.
Other research finds even wider gaps. When the AI consultancy Section surveyed 5,000 knowledge workers, 40% of non-managers said AI saved them no time in a typical week, compared with just 2% of executives. McKinsey’s 2026 State of AI survey found that 47% of mid-level managers and individual contributors had experienced AI-related strains, compared with 31% of executives and senior managers.
WHAT LEADERS MISS
The gap between executive AI productivity perception and employees’ reality has a structural cause. Leaders see adoption rates, license counts, and the list of tools their teams have deployed.
Employees work inside the processes in which those tools run. In other words, when a company drops AI into a business process originally designed for people, employees are the ones absorbing the friction. Too often they end up re-entering data between steps that AI cannot connect. They also correct outputs because the original process never accounted for AI in the first place. In our survey, 44% of employees said they had to override AI outputs because nobody set up the underlying process correctly. A third said AI had increased their workload.
Many companies also skipped a basic step. Of the employees we surveyed, 69% said their employer never fully consulted them about how AI would support their day-to-day work. In fact, 41% admitted using AI simply to satisfy a corporate mandate when the task did not need it. That behavior inflates the adoption numbers that leaders most often pay attention to, which widens the gap even further.
The consequences of this approach are wide-reaching. Nearly half of employees (46%) in our survey said they would consider changing jobs because of poorly implemented AI. Meanwhile, 48% of organizations reported they had already rolled back AI use because it hurt employees’ ability to do their jobs.
External macroeconomic data tells the same story. A National Bureau of Economic Research working paper that surveyed nearly 9,000 executives found that 83% saw no impact from AI on labor productivity over the past three years. In McKinsey’s survey, 80% of respondents say AI improves their individual productivity. Yet the percentage reporting enterprise-level operating income impact from AI has held flat at 37%. While individuals might experience some productivity gains, their organizations ultimately don’t see an impact on the bottom line.
HOW TO CLOSE THE GAP
Leaders can close the gap by redesigning work with the people who do it.
McKinsey’s survey also points to what distinguishes the small group of companies that report significant returns. Nearly three-quarters of its AI high performers say they have fundamentally redesigned workflows because of AI, compared with about one-quarter of everyone else. Those companies rebuilt the factory floor.
Here are three leadership moves that can help other organizations do the same.
1. Put frontline employees in the room for process redesign. They know where processes can’t handle exceptions, or where colleagues retype data by hand. A redesign that skips essential employees will reproduce the same flaws from the original legacy process. It will also cost leaders the trust they need for the next change.
2. Measure the value and impact on the work instead of the usage. AI-adoption rates tell leaders whether people open a tool. End-to-end business process measures, such as cycle time, error rates, and cost per case tell leaders whether the work has improved or not. Once leaders track outcomes, employees lose any reason to perform their compliance obligations with AI.
3. Fund the redesign. In our survey, 37% of organizations allocated 10% or less of their AI budget to redesigning processes and managing the organizational change that goes with it. At the same time, 82% believe their AI investments will be a bust without more investment in process design. Leaders can resolve that contradiction by funding transformation efforts that go beyond surface-level AI integrations.
Electricity eventually transformed manufacturing for the owners willing to rebuild their floors. AI adoption will reward the same willingness. Leaders can start now by choosing the process that frustrates their employees most, inviting the people who run it to redesign it, and measuring the resulting changes to productivity. That single exercise is far more informative than any AI-adoption dashboard.
Jakob Freund is CEO of Camunda.
More
