register to Today’s TPRTexas Public Radio’s newsletter brings you the top news stories every morning.
A new study from the University of Texas at Austin’s McCombs School of Business and accounting firm KPMG looks at why some people get much more value from using AI than others.
When researchers analyzed how people actually use AI in the real world, they found that the high-influence users in their study produced better and more efficient work by using AI as a collaborator rather than a search engine.
Researcher Zach Kowaleski is an assistant professor in the Shulkin School of Accounting at the University of Texas at Austin.
“You have to be clear about what you want and what it will look like when you get it,” he said.
“And the people who are doing that in their workflow are also the ones who are repeating it. So like any good listener, when something comes back, they push up on the parts that don’t sound right to them and improve from there.”
The study found that so-called high-influence users provide clearer context and specific goals to achieve results. The greatest benefits from AI were seen by those who used AI in their daily work flow, not just occasionally.
The most sophisticated users approached the problem by treating the AI as a reasoning partner, asking the LLM to take on a specific role or perspective, and provide specific direction and examples. They showed the AI how to reason through a task and asked it to explain how the model got its response. Instead of accepting the initial output, we refined the model’s work through multiple exchanges.
Researchers spent eight months examining 1.4 million interactions with artificial intelligence in real-world workplaces.
This study is published in Harvard Business Review
