Stanford study finds AI adoption works better when workers get more than a productivity target

Stanford HAI researchers found that paralegals used an AI tool 58% more, and experimented with it 70% more, when managers framed it as a way to make work more meaningful rather than simply faster. Their study suggests that training, time to explore and measures that recognise higher-value work matter alongside the software itself.

Stanford HAI Watch analysis

What happened

The researchers studied AI deployments at a law firm, an advertising agency and an IT services firm. In one part of the research, they interviewed 184 employees. People required to use AI described feeling that their work had lost meaning, which reduced their motivation to learn new skills.

For nearly two years, the team also observed two similar divisions at a law firm using the same tool to draft contracts and nondisclosure agreements. One manager emphasised faster drafting and shorter turnaround times. The other asked paralegals which routine tasks they wanted help with, gave them time to explore, and backed that invitation with training, knowledge-sharing and mentorship. The second division used the tool 58% more and experimented with it 70% more.

Key findings

• The way AI is introduced matters
  In the law-firm comparison, a job-enrichment pitch was associated with more use and experimentation than a productivity pitch.
• Support made the invitation credible
  Training, weekly knowledge-sharing, mentorship and time to explore accompanied the job-enrichment approach.
• Usage is not the same as value
  The researchers warn that counting AI use or tokens can reward activity for its own sake, rather than useful work.
• Performance measures may need to change
  Paralegals doing deeper research and analysis produced higher-value work, but could handle fewer cases.

Why it matters

Organisations often promise that AI will take routine work off people’s hands. The study’s account points to the catch: if the time saved simply becomes a demand to do more of the same, the promised enrichment can look like a productivity target wearing a nicer hat. The researchers argue that workers need room to learn and take on different responsibilities, with evaluation that recognises those contributions.

The law-firm comparison offers a specific illustration, not a universal recipe. It covers two divisions in one workplace, while the wider employee interviews span three firms. Still, it gives managers something more useful than another slogan: a reason to look at how the tool is introduced, supported and judged.

Our read

The strongest finding is that adoption is a workplace design problem as well as a technical one. If employers want people to use AI creatively, they need to give them time and support to do so, and stop treating raw usage as a proxy for value. Managers can start by asking which tasks workers would actually like to change, then make sure performance measures do not punish them for doing more valuable work at a different pace.

What to watch

• Whether other workplace studies find similar effects beyond the law-firm comparison.
• Whether employers provide time and training when they promise AI will enrich jobs.
• Whether performance measures shift from case volume or tool usage towards the value of the work produced.

Discussion spark: If AI frees workers from routine tasks but employers still judge them by volume, is that a failure of the technology or of management?

Independent WittyWires tracker for public updates about Stanford HAI. Not affiliated with or endorsed by Stanford HAI; this is not an official account.