News

OpenAI Wants an AI Agent for Everything, But Adoption Lags Far Behind

OpenAI is racing to put an AI agent behind every task a white-collar worker does, from coding to document drafting. But the company's own adoption numbers suggest most people aren't ready to hand that work over yet, according to a TechCrunch report published August 24.

The centerpiece of that push is ChatGPT Work, a modified version of OpenAI's Codex coding tool retooled for non-engineers and released last month on the company's lowest subscription tier, at $20 a month, TechCrunch reported. Coverage of the July 9 launch confirms the product combines Codex technology with the newly released GPT-5.6 model, rolling out first to Pro, Enterprise and Edu plans before reaching Plus and Business tiers. The idea is to let non-technical employees deploy agents across their everyday digital workflows, not just code.

The adoption gap is stark. Internally, 98% of OpenAI employees were using Codex as of June, according to the TechCrunch report. But among the company's paying customers, only 17% of organizational subscribers and less than 1% of individual subscribers have picked up the agentic tool, TechCrunch found. OpenAI's combined desktop and mobile app has just 20 million users, compared with more than one billion people who use ChatGPT online, per the report.

There are real reasons people are hesitant. Andrew Ambrosino, who leads engineering for OpenAI's desktop app, told TechCrunch he worries about agents overstepping their intended scope: "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." Thibault Sottiaux, who leads OpenAI's core product work including Work, told the outlet the goal is for "ChatGPT [to] actually do entire, very complicated tasks for you all autonomously in a way that is delightful and safe." Joe Gershenson, who leads engineering for OpenAI's agent harness, said the challenge is precision: being "more precise about what information the model really needs."

What this means for job seekers

The lesson isn't that agents are coming for your job wholesale — it's that they're coming for specific, well-defined tasks inside it, and companies are still working out which tasks those are. That distinction matters for how you position yourself in a job search.

Start by auditing your own role the way OpenAI's own numbers suggest employers eventually will: which parts of what you do are repeatable, rules-based, and low-stakes enough to hand to an agent, and which parts require judgment, accountability, or a human relationship with a client or colleague? The tasks in the second bucket are where your resume should lean.

The adoption gap TechCrunch documented also points to a real, hireable skill: supervising and auditing agent output. Ambrosino's concern about agents pulling from the wrong private data source is exactly the kind of failure mode someone needs to catch before it reaches a client or manager. Employers building out agent workflows will need people who can review what an agent produced, spot when it went off-script, and take responsibility for the result — not just people who can prompt one.

If you're updating a resume or LinkedIn profile, consider naming that skill directly: experience reviewing AI-generated work, setting guardrails for automated tools, or owning outcomes when an agent is part of the workflow. As more employers roll out tools like ChatGPT Work, that oversight layer is likely to become a distinct job requirement rather than an assumed skill.