Play free
Menu

Research summary

Does a person and AI always make a better team?

Not always. Across 106 experiments, people with AI did better than people alone. But the pair often did worse than the better of the two working alone, most of all on decision tasks.

The study

Who: Michelle Vaccaro, Abdullah Almaatouq and Thomas Malone.

What: A systematic review and meta-analysis, published in Nature Human Behaviour in 2024.

Type: A research review. It combined the results of many experiments.

Evidence: 106 experiments from 74 research papers, published from January 2020 to June 2023.

What the researchers found

There are two different comparisons here.

Compared with people working alone, people working with AI did better on average.

Compared with the better of the two working alone, the person-and-AI pair did worse on average. Sometimes the AI alone was better. Sometimes the person alone was better. The pair often lost some of that advantage.

The gap was largest on decision tasks: tasks where someone chooses between options.

What this may mean for managers

This is our interpretation, not the researchers' finding.

Adding a person to an AI process, or AI to a person's work, does not guarantee a better result. The result depends on how the person works with the AI: when they use its advice, when they check it and when they overrule it. That is a skill people can practise.

What this study did not test

  • It reports averages across many different tasks, people and tools. One workplace may differ.
  • It did not test any training programme, including StepSim.

Sources

  • Vaccaro, Almaatouq and Malone (2024). When combinations of humans and AI are useful: a systematic review and meta-analysis. Nature Human Behaviour. arXiv. Source checked 1 October 2026.