OpenAI's ChatGPT Work is betting on AI agents for everyone-but trust and usability remain obstacles

OpenAI launches ChatGPT Work: Bringing AI agents to the masses, but trust and ease of use remain challenges

OpenAI has released ChatGPT Work, a $20 per month subscription service designed to allow non-engineers to use AI agents to automate complex, multi-step tasks in various digital tools. Launched last month, the product is an improved version of the company's code-writing tool Codex, marking OpenAI's bet that AI can move from answering questions to acting autonomously in the workplace. However, as early users begin testing the system, issues about control, ease of use and cost are becoming major obstacles to widespread adoption.

What is ChatGPT Work and why is it important?

ChatGPT Work is designed to allow white-collar workers-from accountants to doctors-to deploy AI agents to interact with existing software tools, including email, Slack, Notion, and Figma. As OpenAI puts it, its goal is to achieve a world where "intelligence is not limited to answering questions, but also helps everyone turn their greatest ideas into reality." For the company, the commercial value is extremely high: the longer the agent works, the more tokens it consumes and the higher the revenue per user. Expanding into new career areas is critical not only for OpenAI, but also for the entire AI industry-currently, programming remains the most successful application scenario for AI.

What is the difference between ChatGPT Work and Codex?

Codex was originally designed for software engineers, but non-engineering employees at OpenAI found it unfriendly-it asked code questions and displayed technical information like "white space differences." Over time, the company made it more versatile, culminating in ChatGPT Work. Unlike Codex, which is still popular among developers, Work aims to provide a similar autonomous proxy experience for non-programmers. The interface contains buttons for selecting projects and plug-ins, but is still essentially a "magic box": the user enters a request, and the agent handles the rest itself.

Early Adoption and Usage Patterns

OpenAI said that as of June this year, 98% of its employees were using Codex, but only 17% of organizational users and less than 1% of individual users were using the proxy programming tool. This gap highlights the challenge: While the technology is transformative for engineers, it has not yet penetrated widely into the wider professional world. ChatGPT Work was launched to close this gap, but early usage data is still limited.

What are the main challenges to adoption?

One of the biggest obstacles is trust. Giving AI agents access to your inbox, Slack and other sensitive tools requires a huge amount of trust. Andrew Ambrosino, chief engineer of OpenAI desktop applications, admitted that the model could inadvertently read private messages when it should not have done so. "I'll do this for my job," he told BitcoinWorld (note: the original text is sourced here, but the specific source has been hidden upon request), but he also admitted that not everyone can accept this trade-off.

Ease of use is another issue. Setting permissions for agents to access cloud drives or calendars can be confusing, and many settings are only available in web apps, forcing users to switch between desktop and mobile. Models sometimes fail to explain why an operation fails, frustrating users. As one early user said: "Unless you turn up the effort, you'll get the worst intern you'll ever meet."

Cost and efficiency issues

Cost is also a problem. In the $20-month plan, a regular four-day session consumes more than 80 million tokens, which models estimate will cost $65-a subsidy of more than three times the subscription price. OpenAI said it is optimizing efficiency and noted that the price of its Luna model has been lowered by 80%, but long-term economic benefits remain uncertain.

How does ChatGPT Work compare to competitors such as Claude Cowork?

OpenAI engineers are reluctant to compare their products to Anthropic's Claude Cowork, but the similarities are obvious. Both provide a proxy framework that connects users 'existing tools. However, Claude Code was initially popular because of its more conversational and iterative approach, confirming with users at every step. OpenAI's first version of Codex was more autonomous, but the error rate proved to be too high. Since then, the company has added more user interaction nodes, and now ChatGPT Work aims to find balance.

Industry analysts believe this competition is benign, but point out that verticals dedicated tools like Harvey (for the legal field) and Clay (for the sales field) are gaining advantages by remaining model-independent. As Christian Catalini of a16z writes, if OpenAI cannot quickly obtain the complementary assets necessary to drive AI's scale in the market, value may flow elsewhere.

What this means for the future of working

For AI advocates, ChatGPT Work is the digital personal assistant they dream of. It can organize messy email inboxes into a structured calendar, or generate financial metrics dashboards without writing a single line of code. But for ordinary employees, this commitment is diluted by configuration, trust and occasionally confusing restrictions. As Wharton's Ethan Mollick points out: "ChatGPT tends to work magic and do it directly for you; Claude compares and displays the results."

OpenAI's Thibault Sottiaux insists the world is ready: "We do see that the world seems to be ready," he cited "incredible adoption rates" as evidence. But whether this adoption rate will spread from early tech enthusiasts to the wider white-collar group remains an open question.

Conclusion

ChatGPT Work is an important step in making AI proxies available to non-programmers, but everything is still in its early stages. Trust, ease of use and cost are the three pillars that determine whether the tool will become standard in modern workplaces or just serve as a niche product for AI curious people. As OpenAI and its competitors continue to improve their respective frameworks, the key is balancing autonomy with user control and proving that its value outweighs the risk.

FAQs

Question: What is ChatGPT Work?
Answer: ChatGPT Work is a $20 monthly subscription service launched by OpenAI that allows users to deploy AI agents to automate tasks in digital tools such as email, Slack, and cloud disks. It is specially designed for non-engineers.

Question: What is the difference between ChatGPT Work and Codex?
Answer: Codex is a programming tool for software engineers, while ChatGPT Work is a more generic proxy product for white-collar professionals. They are based on the same underlying technology, but Work has a friendlier user interface.

Q: What are the main concerns about using ChatGPT Work?
Answer: Key concerns include privacy (allowing AI to access sensitive data), ease of use (configuring permissions can be confusing) and cost (heavy use can cause the cost of the token to exceed the price of the subscription).