CHICAGO (AP) — Microsoft CEO Satya Nadella is opening a corporate conference in Chicago with remarks that could set the stage for where it is taking its artificial intelligence business.

AI developers are increasingly pitching the next wave of generative AI chatbots as AI “agents” that can do more useful things on behalf of humans. But the cost of building and running AI tools is so high that more investors are questioning whether the technology’s promise is overblown.

Microsoft said last month that it is preparing for a world where “every organization will have a constellation of agents—ranging from simple quick-and-answer to fully autonomous.”

Microsoft elaborated in a blog post on Tuesday that such autonomous agents “can work around the clock to review and approve customer returns or review shipping invoices to help companies avoid costly supply chain errors.”

Microsoft’s annual Ignite conference is aimed at its large enterprise customers. The pivot towards so-called “agentic AI” comes as some users see limits to the major language models behind chatbots such as OpenAI’s ChatGPT, Google’s Gemini and Microsoft’s own Copilot. These systems work by predicting the most likely next word in a sentence and are good at some typing-based tasks.

But tech companies have been working to build AI tools that are better at planning and reasoning from longer distances so they can access the web or control computers and perform tasks on their own on a user’s behalf.

Salesforce CEO Marc Benioff has criticized Microsoft’s pivot. Salesforce also has its “Agentforce” service that uses AI in sales, marketing and other tasks.

“Microsoft is renaming Copilot ‘Agents’? It’s panic mode,” Benioff said in a social media post last month. He went on to claim that Microsoft’s flagship AI assistant, called Copilot, is “a flop” that is flawed and spills corporate data.