Working with AI may beat working alone, but not the best solo option
Do not give the team credit just for adding AI. Vaccaro and colleagues' research synthesis reports a split result: human-AI teams beat humans alone on average, but lost to the stronger solo option. It draws on many studies, not a test of your own team's job steps.
What the source reports
Vaccaro, Almaatouq, and Malone built AutoAnalysis to gather and update studies on people working with AI. The November 2 digest says its latest run combined 1,955 measured results from 420 papers. It reports two average findings. People with AI did better than people alone. But they did worse than the stronger of the two solo options: person or AI. Beating one baseline can still mean losing to a better one. The digest used the paper's early abstract; it did not check each study or its methods. These averages do not tell us what would happen with a given team's work.
Original source
- Title
- AutoAnalysis: An automated, self-updating meta-analysis of human-AI collaboration
- Author
- Michelle Vaccaro, Abdullah Almaatouq, Thomas W. Malone
- Publication
- arXiv
- Date
- Friday, October 30, 2026