Visit hbs.edu

Collaborating with AI

AI-human collaboration is a complex interplay

Based on the research of Fabrizio Dell’Acqua, Edward McFowland III, Karim R. Lakhani

Overall, our research indicates that you can’t just throw AI blankly at all of your employees. Some will benefit more than others, and your employees’ reactions might lead to unintended consequences.

Cyborgs vs. Centaurs

In our study with BCG, the consultants who integrated AI into their workflow (“cyborgs”) saw improvements, while those who merely divided tasks between themselves and AI (“centaurs”) fell short.

AI as a skill-leveler

Furthermore, the consultants who were performing the worst at the beginning of the experiment saw a 43% boost in performance, whereas those consultants who were the top performers at the beginning also saw performance increase but only at 17%. This suggests AI has an equalizing effect on your workforce.

The danger of falling asleep at the wheel

Another LISH study, which analyzed how HR recruiters use AI, found that as the quality of AI increases, humans have less incentive to exert effort and remain attentive. Thus, when humans collaborate with AI, there may be less collective output with higher-performing algorithms. Overall, the results show that maximizing human/AI performance requires careful attention to the quality of the AI and the skillset of the humans involved.


Read the full research here: https://ssrn.com/abstract=457332
https://www.almendron.com/tribuna/wp-content/uploads/2023/09/falling-asleep-at-the-wheel.pdf

Sign up for our newsletter to stay up to date with D^3 news and research: https://d3.harvard.edu/#join-our-community