Stanford Researchers Link Job Enrichment and Support to 58% More AI Use

Paralegals offered time and support to explore more interesting work used an AI tool 58% more than peers told to draft documents faster. Stanford HAI published the finding on October 7, 2026, alongside employee interviews suggesting that a productivity-first rollout can weaken workers’ motivation to learn. The comparison involved different management approaches, not different tools.

The research team included Stanford’s Arvind Karunakaran, the University of Virginia’s Roshni Raveendhran and Dartmouth’s Tami Kim. Their two workplace studies examined how managers introduce generative AI and whether employees have room to discover uses beyond their existing assignments.

In the first study, researchers interviewed 184 employees at a law firm, an advertising agency and an IT services company. Across administrative, creative and technical jobs, employees required to use AI described a loss of meaning in their work. That feeling reduced their motivation to acquire new skills.

Same tool, different terms

The second study followed two law-firm divisions for nearly two years. They were closely matched in practice area, size, pay and promotion structure. Both received the same firm-wide license for a contract and nondisclosure-agreement drafting tool that the researchers called LawBot.

One manager emphasized faster drafting and shorter turnaround times. Paralegals there confined their use to existing tasks. The other manager asked which tasks workers found boring and what they wanted to do but lacked time for. Those paralegals began using LawBot for legal research and case analysis, shared tips and sought access to attorneys’ legal strategy meetings.

58% moreTool use

Paralegals used LawBot 58% more than peers in the productivity-focused division.

70% moreExperimentation

Paralegals experimented with LawBot 70% more than peers in the other division.

The promise needed resources

The enrichment-focused division received concrete support, not just a more appealing introduction. Managers provided:

• Formal training and question-and-answer sessions with the vendor.
• Weekly knowledge-sharing lunches and informal mentorship on more complex work.
• Dedicated time each week to explore the tool independently.

Because the divisions differed in both messaging and support, the comparison does not isolate the effect of the pitch alone. Karunakaran argues that resources make a promise of job enrichment credible; without them, employees recognize it as empty talk.

Fewer cases, more valuable work

The shift also challenged traditional performance measures. Karunakaran says deeper legal research and case analysis produced higher-value work but took more time, leaving paralegals handling fewer cases. Judging them only by quarterly or annual case volume would miss the value of their changed responsibilities.

He also warns against treating token consumption—the amount of text an AI system processes or generates—as evidence of value. Rewarding heavy use can encourage activity for its own sake rather than useful business work. His recommendation is to align evaluation with the new tasks employees take on, not simply count more output or more AI activity.