University of Minnesota researchers find AI can advance business — but not without humans

Some wonder whether — and who — machines will replace in the workplace.

Alok Gupta, information technology chair for the University of Minnesota's Carlson School of Management, says that's the wrong framing for the issue.

Instead, people should ask how humans and artificial intelligence (AI) can work together more — and what companies and employees can do to prepare or respond, said Gupta, also a senior associate dean.

"There are plenty of stories about AI failures in complex job situations," Gupta said. "Certainly, where you need to make decisions based on a variety of different inputs and where it's not clear ... in those kinds of situations where judgment comes into the play, AI has become quite good but it's not perfect, and it won't be anytime soon."

Artificial intelligence (AI) — or computers that imitate human intelligence in performing tasks — already has a growing presence in several industries. And expanding adoption of AI could fuel a 14% boost in gross domestic product in North America by 2030 through increases in productivity, product quality and consumption, according to a PwC report.

Counterintuitively, AI for the near future is likely to take on low-skilled jobs, Gupta said. It may run chatbots to answer simple questions but send more difficult ones to humans.

For example, the latest AI tool, ChatGPT, writes and does coding, but not at an advanced level — at least not yet.

With AI handling rote functions in the workplace, Gupta said humans can gain higher-quality experience and build their skill sets. Humans in teams become much more creative because they feed off the diversity of thinking that they have.

When humans work with machines, Gupta's research shows, they lose creativity because of their over-reliance on the machine's decisions or inputs.

"I don't think anybody will be successful by just taking off the shelf AI technologies and applying them," Gupta said. "That's a recipe for disaster, because they're going to slow down the creativity and entrepreneurship in their own businesses, in their own corporate environments."

Alexandre Ardichvili, a professor in the U's Department of Organizational Leadership, Policy and Development, found that AI use in accounting resulted in a loss of human expertise.

Based on that finding, Ardichvili now is working on broader research to identify different types of effects that AI may have on people in the workplace.

He is looking at where machine learning-based AI — also referred to as narrower AI because it focuses on a specific task — already is replacing or helping people in industries including in retail, banking, finance, insurance, medicine and accounting.

Ardichvili specifically is seeing how organizations and employees may look more at the tasks that need to be done instead of how those tasks work together.

With AI in the workplace, employees likely won't see how the more narrowly defined tasks they perform fit into the work of their organization or their team, Ardichvili said. And they may only work individually on tasks rather than in teams. The downside for humans in the workplace will include loss of interest and loss of meaning in their work and loss of community building.

Ardichvili's research on AI in accounting found that people lost expertise because they had fewer opportunities to learn from deliberate practice. They also had fewer opportunities to raise their skills by performing more advanced tasks and to learn from experienced colleagues and mentors who could help coach them on their way to becoming an expert.

"All of these opportunities for learning could be significantly jeopardized if not completely eliminated by the AI taking over significant parts of the tasks, parceling them out and delegating the performance of more simple tasks to individuals," Ardichvili said.

While he continues studying the effects of AI on humans in the workplace and how to cope with this new reality, Ardichvili harbors some skepticism over the technology.

"Unfortunately, there is no evidence yet of strong benefits of AI," Ardichvili said. "When you look at industry reports coming out in the past several years, there is quite a lot of disappointment and frustration that the AI systems don't deliver on the promises."

Gupta, however, points out that AI today has only limited data on which to learn how to solve increasingly complex business problems. AI and algorithms, though, are much better than humans at judging what they know and what they don't know and can solicit input from human experts.

In some cases, humans and AI working together will outperform either working alone.

"As [people] move up the food chain they have to be able to have a dialogue with AI rather than having it make decisions," Gupta said.