The Swedish buy-now-pay-later company Klarna has become something of a poster child for the potential benefits of generative artificial intelligence. The company relies on AI to create and tailor promotional images and to draft marketing copy, saving millions of dollars. Earlier this year it said an AI chatbot assistant was doing the work of 700 human customer-service agents, which it forecast would boost profits by $40 million this year.

Klarna’s approach highlights generative AI’s promise for powering businesswide systems, like customer service. It’s also helping individual workers with some of the tasks that form their daily work.

[time-brightcove not-tgx=”true”]

U.S. businesses are investing in AI, and they’re eager to see such gains. So far, the tools cover some of a worker’s tasks, saving time but not significantly impacting the business overall. They can, for example, help computer coders work faster, or save human-resources professionals time by having AI-powered chatbots answer colleagues’ questions about benefits.

To use AI as aggressively as an organization like Klarna, companies need to feel comfortable having AI systems access sensitive data or interact with their customers, rather than a human. “We defaulted customers into a [generative] AI chatbot experience and they hated it,” said Josh Silverman, the chief executive of Etsy, on a recent podcast. “It was one of the worst-performing experiments we’ve launched during my entire tenure.” Silverman said the online marketplace shifted to make it optional for customers to interact with the chatbot, but even then it’s likely costing Etsy some sales.

Then there’s the impact on workers. Polls indicate a majority of Americans are anxious that AI will reduce the quality and quantity of jobs. In surveys, workers say that they want their organizations to communicate with them about how AI is being used and will impact their roles; to be trained on the new tools; and to be involved in figuring out how to use AI in their companies.

Tech and finance companies are some of the furthest along. Accenture, for example, developed over the past year a custom version of OpenAI’s ChatGPT to assist employees in crafting sales proposals. The professional-services company involved its proposal writers in designing and testing the tool, and did extensive demos and training for the roughly 1,500 employees using it. “If we can eliminate the lower-value, time-wasting work, it allows writers like myself to be more valuable [with] things like strategic messaging,” said Sarah Szuminski, Accenture’s North America lead for such proposal writing.

IBM said it’s making sure that with AI-based systems there’s a “human in the loop” to make decisions, especially ones that impact people. Humans screen each of the 5 million résumés IBM receives annually. Academic researchers have found biases in AI systems, with some tools recommending African Americans for less prestigious jobs.

If nothing else, the intense interest in generative AI has heightened businesses’ focus on the skills that their people bring. Some 66% of business leaders surveyed by Microsoft and LinkedIn said they wouldn’t hire someone without AI skills. Businesses are still figuring out what the key AI skills are. Some say that having them will be as essential as basic online literacy was when the internet entered workplaces. Others say the ability to learn new things will be the most important skill.

They’re also having to think about how younger workers get critical work experience when entry-level tasks can be done by AI. Without those jobs, how does one learn the judgment, values, and culture of a business that are essential as they climb the ladder?

There are no easy answers, but executives remain optimistic. “The problems [are] largely solvable,” said Etsy’s Silverman, referring to issues like the time it takes an AI chatbot to respond. “It’s important that we keep doing it, because we’ve got to learn.”

More From TIME