Whether it’s helping a consumer buy the right jacket or managing relationships with suppliers, generative AI technology has obvious benefits. During a panel devoted to AI and customer intelligence at the Fortune Brainstorm AI conference this week, speakers cited improved sales, satisfaction, and loyalty among the many advantages that come with harnessing the technology effectively.

“What I’m most excited about is a little bit old-fashioned, but I think it’s really where the heart of the impact is,” said Best Buy chief digital, analytics, and technology officer Brian Tilzer: “How can we actually solve customer problems that weren’t solvable before?” For the electronics retailer that means helping customers solve the problems of using and troubleshooting newly purchased gadgets and appliances, and giving employees more tools to accomplish that goal.

When a customer’s washing machine breaks, for instance, Tilzer said that Best Buy can now predict with nearly 80% accuracy the problem being described over the phone or email, and ensure that the repair person has the right parts on hand when visiting the customer’s home.

Mastercard president of data and services Raj Seshadri highlighted a recently launched AI customer assistance tool that lets consumers tap into retailers’ product catalogs and inventory through natural conversations. “When you add Gen AI to the mix, it’s the next inflection point; there’s a lot you can do now,” Seshadri said.

For all the power and potential of AI, though, Seshadri advised approaching the technology with a sober attitude and thinking about proper guardrails such as governance, principles, measurement, testing, and training across organizations. Seshadri is in favor of tech companies working with regulators. “The [companies’] intentions are all very, very good and spot-on, but [it’s important to] work with regulators to make sure that what [corporations] do works for everyone,” she says.

Accenture’s group chief executive of technology, Karthik Narain, says that keeping a balance of “humans in the loop” and leveraging proprietary data is key to making AI work effectively and accurately.

“We are moving from the phase of sizzle, of ChatGPT … to going into real, strategic, high-value, business-differentiating activities aided, or enabled, by generative AI,” said Narain.

A business’s ability to adopt generative AI right now depends on the accuracy required for the particular process, he said. “As you move up the curve of high strategic aspects, the amount of accuracy that is going to be required is very important.”