Why employees don't use the AI bought for them
Companies invest huge amounts of money in AI tools, and then wonder why employees hardly use them. A recent study by researchers from Stanford, the University of Virginia, and Dartmouth shows: it is often not about the technology and not about people's resistance. It is about how this tool was presented to them.
First, the researchers surveyed 184 employees from a law firm, an advertising agency, and an IT company that had already implemented generative AI. The pattern kept repeating: where using AI was mandated for the sake of speed, people said that their work lost its meaning. And it was precisely these employees who lost the desire to master new skills.
Then, the researchers observed a law firm for nearly two years, where two almost identical paralegal departments received the exact same AI tool. The difference was in only one thing: how the managers presented it.
In the first department, AI was presented as a productivity tool: to do the same things, only faster.
In the second department — as a way to make work more interesting: to relieve routine and allow time for more complex tasks.
The result: in the second department, AI was used 58% more and experimented with 70% more frequently. The paralegals gradually took on legal research and case analysis — work they hadn't done before.
Yet, having just the right phrasing was not enough. In the second department, it was backed up with resources: training, weekly meetings where employees shared their findings, Q&A sessions with the tool's developer, mentoring, and protected time each week to calmly figure out the tool.
And the most interesting detail is about performance evaluation. The paralegals who took on deeper analytics began processing fewer cases. According to volume-based metrics, they looked worse than their colleagues, even though they were doing more valuable work. The authors also warn of another trap: if you measure how much an employee uses AI, the company will begin to encourage meaningless usage just for the sake of the numbers.
What this implies for HR and managers
— If AI is needed solely to do the same thing faster, people perceive it as a way to squeeze more out of them. If it helps make work more interesting, they start investing themselves in it.
— The right message does not work without resources: time, training, and the opportunity to make mistakes while mastering something new.
— When AI changes the nature of work, metrics must change along with it. Otherwise, the evaluation system will punish the very people who use the technology best.
— When a role changes, its weight should also be reconsidered. A paralegal who does analytics is already in a different position, even if it is called the same on the organizational chart.
The effect of AI is not determined by the model or the subscription. It is determined by what the leader said when they laid the tool on the table.