OpenAI Launches ChatGPT Work at $20/Month for Non-Engineers
OpenAI has released ChatGPT Work, a $20-a-month product built on its Codex coding tool, designed to let white-collar professionals outside software development run autonomous AI agents across their digital workflows.
👁 116 reads
OpenAI released ChatGPT Work last month, a $20-a-month product available on the company’s lowest subscription tier that extends autonomous AI agents to white-collar professionals who are not software engineers.
- ChatGPT Work launched last month at $20/month, OpenAI's lowest subscription tier
- It is a modified version of Codex, extended beyond coding to serve accountants, lawyers, doctors and other non-engineers
- OpenAI's lead desktop engineer gives the agent access to his inbox, Slack, Notion and Figma to test its limits
The product is a modified version of OpenAI’s Codex coding tool, adapted to handle what Thibault Sottiaux, who leads OpenAI’s core product work, describes as end-to-end professional tasks.
“In this new factor, ChatGPT can actually do entire, very complicated tasks for you all autonomously in a way that is delightful and safe.”- Thibault Sottiaux, OpenAI
The stated ambition goes further: OpenAI’s marketing frames the goal as intelligence that moves “beyond answering questions to helping everyone turn their biggest ideas into reality.”
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →From coding tools to every desk
Software developers have had AI agent functionality for some time, but ChatGPT Work is the first systematic attempt by OpenAI to give that same capability to accountants, investors, doctors and others whose work is dominated by computers rather than code. The transition has been uneven internally. Andrew Ambrosino, lead engineer for OpenAI’s desktop app, told TechCrunch that the company’s non-engineering teams - communications and finance among them - began using Codex “at a time that it was actively hostile to them,” asking users about code and surfacing technical readouts meant for software changes. Work on making the tool more general-purpose began in February, he said.
Ambrosino himself runs the agent with access to his inbox, Slack account, phone, Notion, Figma and more - an unusually broad grant of control. He acknowledged the risk plainly: “If I’m asking it to write a document, is there a possibility that it’s going to pull from a private DM on that subject and not know that it’s not supposed to share some info? Yes. I’ll do it for the job.”
The commercial logic is straightforward. Agents that complete longer, multistep tasks burn through more tokens, making each user more valuable to OpenAI. Expanding beyond software engineers is therefore critical - both for revenue and for justifying the industry’s enormous investment in training and compute. Vertical-focused rivals such as Harvey for legal work and Clay for sales are already pursuing those professional markets with model-agnostic platforms, meaning they can swap in whichever underlying model performs best.
That competitive dynamic is a recognised risk. Christian Catalini, writing on a16z’s blog, warned: “If the labs cannot rapidly get ahold of the key complementary assets needed to scale AI in the market, value will accrue elsewhere.”
More on this from QUANTSLANT
news·Linux trick turns SQLite database file into directly executable binaryDeveloper Farid Zakaria has demonstrated a Linux technique that stores ELF executable components inside a SQLite database and runs the file directly as a binary.