OpenAI has introduced ChatGPT Work, an AI agent that is intended to read emails, create documents and automate entire workflows. The company is increasingly moving away from a chatbot for everyone and is primarily aimed at corporate customers. But initial user reactions show that there is still a gap between promise and reality. A commentary analysis.
What is ChatGPT Work?
- ChatGPT Work is said to be able to collect information from connected apps and workflows and create complete documents, presentations and even websites. According to OpenAI, the AI agent can working for hours on complex projectsbreak them down into sub-areas and work through them independently. Users should be able to track progress, ask questions and approve actions. The GPT-5.6 model serves as the technical basis. It includes the powerful Sol variant, the Terra variant intended for everyday tasks and the cost-effective Luna version.
- According to OpenAI, GPT-5.6, as the basis of ChatGPT Work, should process multi-stage orders more reliably and stick more closely to templates and reference data. In the desktop version, ChatGPT Work also has a built-in browser Research online and access web applications to be able to. Using the “Computer Use” function, the AI agent can operate programs, enter texts and move files. OpenAI is increasingly developing ChatGPT into an independent platform, a kind of super app that goes beyond a pure assistance system.
- ChatGPT Work gets access to information and context via a plugin directory. The agent can use this, among other things, to access applications such as Gmail, Microsoft Teams, calendar, CRM systems or Google Drive. Specifically, this means: ChatGPT Work can view messages, evaluate video calls, make calendar entries and create documents. According to OpenAI, such workflows can also be planned in advance. They could start at a specific time, regularly or in combination with a specific event.
Why OpenAI is lagging behind the competition
While Meta, Apple and Microsoft simply integrate their AI assistants into their existing products, OpenAI has so far been unable to keep up with corporate customers. That’s why CEO Sam Altman has one Change of course initiated. Specifically, this means: ChatGPT Work is intended to create a work agent and GPT-5.6 to create a work platform that goes far beyond small talk with a chatbot.
What is somewhat noticeable, however, how much the AI agent resembles Claude Coworkan assistant with web access, plugins and computer access that Anthropic had already launched in early 2025. But this also reveals that OpenAI is lagging behind the competition. Because in contrast to its biggest competitor, Anthropic, the company has wasted a lot of time adding emotions, emojis and nonsensical functions such as an erotic mode to ChatGPT.
Considering the disadvantages like that enormous consumption of resources as well as the potential for misuse of AI, the advantage lies more in the automation of processes, evaluation of data or optimization of work processes. This means that the cost-benefit factor of chatbots is not only low in private use, but also represents a major environmental impact.
However, you can really make money with AI agents and, above all, corporate customers. With research and healthcare in mind, AI agents like ChatGPT Work even have a social benefit. The biggest problem, however, remains the error-proneness of AI. In other words: If the chatbot gets confused or says nonsense, it is sometimes more harmless than if an agent messes up sensitive data, documents or messages.
Voices
- OpenAI quotes Will Daney, go-to-market manager at NVIDIAout of self-interest in an official ChatGPT Work announcement: “For events, I used to spend an enormous amount of time in Excel compiling account lists, tracking registrations, and processing feedback. With ChatGPT Work, I turned this into an automated workflow that I can run one event at a time and then share with global teams so they can adapt it to their regions. Before GTC, about 40% of my time was allocated to manual calculations and analysis. Now this process runs twice a week ChatGPT.”
- A Reddit user in a subforum entitled “ChatGPT Work doesn’t Work” is particularly upset about a promotional video: “I don’t understand why a company with a marketing budget in the billions can’t produce a video of their product working. They could have paid an intern a few hundred dollars to do all of these tasks manually – correctly – and then shown that as an ad for their product. Or they could have run their product a thousand times and just picked a run where it didn’t screw up the task. But that’s exactly what they decided to present to the world.”
- Thibault Sottiaux, product lead at OpenAIresponded immediately after the release of ChatGPT Work to the criticism of many users who pointed out deficiencies in the usability and cost structure: “The last 48 hours on Codex and ChatGPT Work have been very intense! Three important news: Temporary removal of the 5-hour usage limit for all Plus, Business and Pro tariffs. Introducing changes that make GPT 5.6 Sol more efficient overall and reflected in lower consumption, so you can do more with it. The exact impact is still being quantified and announced. We have reached 6 million active users and will be rolling out a usage reset in the next hour.”
ChatGPT Work as a new business model
Admittedly: ChatGPT Work is not just a small update. Because with the AI agent, OpenAI is increasingly moving away from the idea that ChatGPT is a chatbot for everyone. Sure: Of course it still exists. But corporate customers, sales, marketing and finance are increasingly becoming the focus of OpenAI.
Why? Because the company simply doesn’t make a profit. Private users will of course continue to be important. With ChatGPT Work, OpenAI reveals, where the money will be earned in the future. Compared to classic subscription models, corporate plans allow higher prices and entire workflows to be tied to ChatGPT.
The crucial question will not only be what tasks the AI agent can take on, but also how reliable and data protection-friendly it is. Also questionable: How much control are companies willing to hand over to an AI? One thing is certain: The The race between AI companies has now finally shifted. And that’s a good thing.
Because chatbots that are designed for entertaining conversations are perhaps a nice gimmick simply a waste of resources – both for the environment and for companies like OpenAI. As with chatbots, the greatest risk remains AI’s susceptibility to errors, which can have fatal consequences for companies – also with regard to their image and their customers.
Also interesting:
