Ineffective Counsel

While generative AI promises to assist small businesses in myriad ways, a study of Kenyan entrepreneurs revealed it may hinder, rather than help, those who need it the most. Berkeley Haas researchers built a GPT-4-powered AI business “mentor” for East African business owners who could apply it to any problem—such as changing a menu or diversifying wholesale offerings. It gave top performing companies a 15% revenue boost, but struggling firms saw their proceeds fall 8%.

Doctoral candidate Nicholas Otis and assistant professors Solène Delecourt and David Holtz found that high performers asked for help with straightforward tasks, but companies with weaker revenues and profits sought advice on difficult tasks that might stymie AI—or even humans. 

“Generative AI has the potential to significantly influence business performance,” says Otis, the lead author. “Whether its impact is positive or negative depends on tasks entrepreneurs select for AI assistance.”

The results contrast with recent research that found that gen AI helped college-educated workers be more productive on well-defined tasks. What’s more, those with the weakest skills benefited the most. 

However, the Haas researchers concluded that while generative AI could benefit millions of companies in emerging economies, it may also widen the gap between high and low performers. 

“For gen AI to really add value to entrepreneurs in more open-ended contexts, they’d also need expanded access to complementary skills training and resources—including financial resources,” Holtz says.

Even so, carefully implemented AI interventions hold some promise for business development, Delecourt adds. “An optimistic way to view our results is that we had a positive effect for a subset of the population with a very low-cost intervention,” she says. “It’s just not a one-size-fits-all solution.”