OpenAI is making its biggest push yet to get AI agents out of the coding world and into the hands of accountants, doctors, and everyone else whose job involves a computer. The vehicle is ChatGPT Work, released last month at the $20 monthly tier, built on a modified version of the company’s Codex coding tool.
The pitch is simple: an AI that doesn’t just answer questions but completes multistep projects on its own. Andrew Ambrosino, lead engineer for OpenAI’s desktop app, has already handed over the keys — his inbox, Slack, phone, Notion, Figma, and more. He acknowledges the risk: “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.” But he says he hasn’t had to take that hit yet.
The commercial stakes are high. Agents that run for longer stretches consume more tokens, which translates directly into revenue per user. But adoption outside engineering has been dismal. An OpenAI-backed study found that while 98% of OpenAI employees were using Codex in June, just 17% of organizational subscribers and less than 1% of individual subscribers had touched it. That gap between internal enthusiasm and external indifference is the entire ballgame.
OpenAI isn’t the only one eyeing this prize. Vertical players like Harvey for law and Clay for sales have been winning non-technical customers with a model-agnostic approach — plugging in whichever AI works best at any given moment. Thibault Sottiaux, who leads OpenAI’s core product work, frames it as a mission: “It’s the very mission of OpenAI — to bring everyone along.” The tension is whether OpenAI can build a harness simple enough for non-engineers before those vertical competitors lock up the market. The company’s own engineers admit Codex was “actively hostile” to non-coders as recently as February, showing them technical readouts about empty diffs. Making that experience feel like Windows, not DOS, is the real product challenge.
