OpenAI’s New ChatGPT Agent Tries to Do It All
Isa Fulford, the analysis lead for OpenAI’s new ChatGPT agent, wanted to order a bunch of cupcakes, so she requested the AI software to do it for her. “I used to be very particular about what I wished, and it was numerous cupcakes,” she says. “That one took nearly an hour—but it surely was simpler than me doing it myself, as a result of I did not need to do it.”
OpenAI has launched a brand new agent for ChatGPT that makes use of a digital browser to finish duties and might generate downloadable information, particularly PowerPoint shows and Excel spreadsheets. Whereas not a full alternative for the Microsoft suite of office instruments, the options included on this agent from OpenAI might obviate some customers’ reliance on Microsoft’s enterprise software program. The 2 corporations are longtime companions and presently in contract negotiations over ongoing entry to OpenAI’s fashions.
The discharge is a part of OpenAI’s ongoing efforts to show its almost three-year-old chatbot right into a money-making product. No small feat, regardless of the software’s thousands and thousands of customers, once you issue within the prices to coach and run highly effective AI fashions in addition to the exorbitant salaries required to retain top-tier workers members.
An agent, on this context, refers to an AI software that is ready to—or a minimum of makes an attempt to—navigate third-party software program and web sites and make choices on its journey to finish digital duties, following an preliminary set of directions from the person. “Agent” is the buzziest of buzzwords proper now for corporations seeking to promote generative AI instruments, particularly these with an eye fixed on enterprise clients.
“We’ve tried to construct a product with an entire lot of enterprise use circumstances,” says Yash Kumar, the product lead on the ChatGPT agent. Along with its file-generating capabilities, the agent can fill out on-line kinds, use a programming terminal, and make calls to public APIs to on-line companies like Google Drive and SharePoint.
This isn’t the primary agent launched by OpenAI in 2025. The brand new ChatGPT agent brings collectively facets of OpenAI’s web-browsing Operator and its long-processing deep analysis options, each launched earlier this yr and regarded to be brokers by the startup. “I used to be on the deep analysis group, and Yash was on the Operator group,” Fulford says. “We realized that the 2 merchandise are very complementary, and principally determined to mix groups.” The ChatGPT agent can swap between interacting with a visible browser, the place it could click on round like Operator does, and a text-based browser, the place it could course of a great deal of web sites like deep analysis does.
The rollout of the ChatGPT agent is coming first to Professional, Plus, and Staff subscribers, beginning at present for Professional customers. Enterprise and Schooling subs will possible obtain entry to the characteristic later in the summertime. At launch, Professional customers are typically capped at 400 agent prompts a month, with 40 prompts allowed for the opposite tiers of paying customers. It’s unclear when this characteristic will roll out without spending a dime customers of ChatGPT.