OpenAI launches ChatGPT Work, an AI agent tool for white-collar workflows

OpenAI has launched ChatGPT Work, a subscription-tier tool designed to integrate large language models (LLMs) into the digital workflows of white-collar professionals such as accountants, investors, and doctors. The product is available for $20 per month and is positioned as a shift from LLMs that answer questions to agents that autonomously complete multi-step tasks.

The tool is built on a modified version of OpenAI’s Codex coding agent and is intended to provide non-engineers with the same autonomous functionality that software engineers already use. Thibault Sottiaux, who leads OpenAI’s core product work including Work, described the goal as enabling LLMs to "do entire, very complicated tasks for you all autonomously in a way that is delightful and safe."

ChatGPT Work grants LLMs access to user tools such as inboxes, Slack accounts, phones, Notion, and Figma, allowing the model to perform tasks across these platforms. OpenAI acknowledges the risks, with lead engineer Andrew Ambrosino noting the possibility of models pulling from private direct messages or other sensitive data, though he stated he has not encountered such issues in practice.

Commercially, agents that operate for longer durations consume more tokens, making them more lucrative for OpenAI on a per-user basis. The company views expanding agentic tools beyond software engineers as critical to justifying its investments in training and compute, given that coding represents a small subset of professional work.

OpenAI reports that 20 million people use the combined Work and Codex apps, compared to over 1 billion users prompting ChatGPT online. An OpenAI-backed study cited internal usage data showing 98% of OpenAI employees using Codex, but only 17% of organizational subscribers and less than 1% of individual subscribers doing so.

To make agentic tools accessible to non-engineers, OpenAI has focused on simplifying the interface, moving beyond command-line tools like CLIs that are familiar to developers. Ambrosino compared the approach to skeuomorphism, where digital tools are designed to resemble their physical counterparts to aid user adoption.

The company faces competition from vertical-specific competitors such as Harvey (for law) and Clay (for sales), which adopt a model-agnostic approach by integrating with multiple AI providers. Industry analysts, including Christian Catalini of a16z, have warned that labs must secure complementary assets to scale AI in the market or risk value accruing elsewhere.