Researchers from Harvard, Wharton and other schools teamed up with Boston Consulting Group and gave 758 consultants eighteen realistic tasks. Some worked with GPT-4, some without.

On tasks the AI was good at, the results were impressive. Consultants with AI finished 12 percent more work, 25 percent faster, and their quality was more than 40 percent higher. Those who usually scored lower gained most.

Then came a task that was just a little too hard for the AI. Here, consultants who relied on it got the right answer less often than those who worked without it. The researchers call the invisible line between the two kinds of tasks the "jagged frontier."¹

Talent Takeaway

It reminds us that learning to use AI includes learning where it helps. A simple team exercise: sort your tasks into two piles, one where AI is reliable and one where it still needs a close human eye. Then agree on how you double-check the second pile.

¹ Dell'Acqua, F., McFowland III, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality (Harvard Business School Working Paper No. 24-013). https://doi.org/10.2139/ssrn.4573321

Further reading: Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044