OpenAI’s ChatGPT Work, released last month on the company’s $20-per-month subscription tier, is the company’s biggest bet yet that AI agents can move beyond coding and into the messy reality of white-collar work. The product, a modified version of OpenAI’s Codex tool, connects large language models to a user’s inbox, calendar, Slack, and SaaS platforms like Notion and Figma, allowing them to autonomously complete multistep tasks like compiling weekly metrics reports or building investment memos.
But the launch raises a question that goes to the heart of the agentic AI push: How much control are you willing to give an LLM over your digital life? For Andrew Ambrosino, the lead engineer for OpenAI’s desktop app, the answer is total access—even if it means accepting some 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,” Ambrosino told TechCrunch. “I’ll do it for the job. I will take the personal hit here and there if I have to. And I haven’t had to.”
Also read: Hugging Face Reportedly in Talks for $13B Acquisition Amid AI Infrastructure Gold Rush
From coding tool to office assistant
ChatGPT Work is a deliberate pivot from OpenAI’s earlier focus on software engineers. While coding agents like Codex and Anthropic’s Claude Code have revolutionized how developers work, they represent a tiny fraction of the professional workforce. OpenAI’s marketing copy frames the goal succinctly: a world where “intelligence goes beyond answering questions to helping everyone turn their biggest ideas into reality.”
Thibault Sottiaux, who leads OpenAI’s core product work including Work, told TechCrunch that the product can “actually do entire, very complicated tasks for you all autonomously in a way that is delightful and safe.” He added, “It’s the very mission of OpenAI—to bring everyone along.”
Also read: Flock CEO calls for privacy-safety ‘compromise’ as police misuse cases mount
The commercial logic is clear. Agents that work for longer stretches burn through more tokens, making them more lucrative per user. Reaching new professions is essential not just for OpenAI but for the entire industry, which has invested heavily in training and computation. But vertical-specific competitors like Harvey (for law) and Clay (for sales) are already chasing those customers with a model-agnostic approach, plugging in whichever AI works best at the time.
Industry analysts see this as a major challenge. “If the labs cannot rapidly get ahold of the key complementary assets needed to scale AI in the market, value will accrue elsewhere,” Christian Catalini wrote on a16z’s “It’s time to build” blog.
The trust and usability gap
OpenAI’s own data highlights the disconnect. An OpenAI-backed study found that in June, 98% of OpenAI employees were using Codex, but just 17% of organizational users and less than 1% of individual users were using the agentic coding tool. The gap between near-total adoption inside the company and negligible adoption outside it is the challenge and opportunity for the company.
Making AI work for non-engineers requires more hand-holding. Ambrosino said OpenAI’s non-engineering workforce started using Codex “at a time that it was actively hostile to them—asking them about code and showing them, ‘oh, you have an empty diff for this thing.'” The team has since worked to make it more general purpose.
Early users, including TechCrunch’s own reporter, have found the experience mixed. The model can handle impressive tasks—like pulling a preschool calendar from email and adding it to Google Calendar—but setting up permissions is confusing and circular. “I tried multiple times to give it just ‘read’ access and received error messages,” the reporter noted. Many important settings are only available on the web app, and the model’s “effort” settings aren’t intuitive for new users.
Joe Gershenson, the engineering lead for OpenAI’s harness, acknowledged the issue: “There are things that we can do better to help them get the right level of reasoning.” He added, “Watch this space.”
The rivalry with Anthropic and the path forward
OpenAI’s engineers were reluctant to discuss competitors, with Gershenson saying, “I really don’t look at the harnesses that they’re building.” But the similarities between ChatGPT Work and Claude Cowork are hard to ignore—the first thing ChatGPT Work asked TechCrunch’s reporter to do was port over their Claude Cowork data.
The rivalry is rooted in a key design lesson. When OpenAI first developed Codex as a web app, the engineers bet on the model being smart enough to handle tasks entirely on its own—a “bit more AGI-pilled” approach, as Ambrosino put it. Anthropic’s Claude Code, built shortly afterward, was oriented around a back-and-forth conversation, checking in with users at each step. That approach proved more effective, and OpenAI eventually followed suit.
“[Our] product was a little ahead of where the model and harness was at the time,” Ambrosino says now.
Despite the competition, OpenAI’s engineers insist the key differentiator is the strength of their latest models. “The frustrating answer is that a lot of times it is the model,” Ambrosino said.
There’s also an open question about whether most people are ready for this level of autonomy. Ethan Mollick, the Wharton School professor who studies AI in the workplace, still sees Claude as more user-friendly, writing that “ChatGPT tends to want to do magic & just do it for you, while Claude does comparisons & shows them, repeatedly asking for input & feedback.”
Sottiaux disagrees, arguing that the conversational nature of the app is better than learning how to use an application. “We definitely see that the world seems to be ready,” he says. “This is why we’ve had incredible adoption.”
One of the biggest hurdles is cost. TechCrunch’s reporter used more than 80 million tokens in four days on a $20-per-month subscription, which cost $65 according to the model’s analysis—a subsidy of more than 3x the subscription price for casual use alone. Sottiaux says the company is “working every day to push the frontier on efficiency,” pointing to a recent 80% price cut for users of OpenAI’s Luna model.
For now, ChatGPT Work is a promising but imperfect glimpse of the future. It can save time on routine tasks, but it requires patience, trust, and a willingness to hand over access to your digital life. Whether that trade-off appeals to the average accountant or doctor remains the central question—and the one that will determine whether OpenAI’s bet pays off.
Be the first to comment