OpenAI Launches ChatGPT Work, a Digital Employee Built on GPT-5.6

OpenAI launched ChatGPT Work on Thursday, a new agent built into ChatGPT designed to carry out complex office tasks on its own, drawing on a user's connected apps and files. The tool runs on the new GPT-5.6 model and is rolling out first to the highest-paying customers.

ChatGPT Work turns the familiar chat assistant into something closer to a digital employee. Instead of answering individual questions, the agent is meant to understand a broadly stated goal, pull in company files and accounts, and then deliver a finished result on its own, whether that's a document, a presentation, a financial report or a website.

What's New in GPT-5.6

GPT-5.6 debuted in three versions tailored to different uses. Sol is the most computationally powerful variant, built for the hardest analytical and coding tasks. Terra is meant to balance answer quality against query cost, while Luna prioritizes speed, which matters for tasks that require many short interactions over the course of a workday.

OpenAI emphasizes that the new model handles agentic coding much more cheaply than previous versions. Sam Altman said that when companies adopt tools like this, they look above all at the quality-to-cost ratio rather than raw model power.

The model is 54 percent more token-efficient in agentic coding - Sam Altman, CEO of OpenAI

The Race for the Office Desk

The launch of ChatGPT Work comes just days after Anthropic expanded its Claude Cowork agent from a desktop app to phones and the browser, letting Max plan subscribers kick off tasks on their computer and track progress from their phone. Both companies are now fighting on the same terrain: not single chat replies, but taking over entire chunks of everyday office work.

Google, Microsoft and Salesforce are joining the same race, having spent months developing their own versions of office agents. The market for agentic office tools is gaining momentum faster than the market for pure language models, because this is where companies are looking for measurable savings in employee time.

Government Scrutiny Over the Launch

The rollout of GPT-5.6 wasn't free of political friction. The Trump administration had previously held up the model's launch, and Altman acknowledged that OpenAI made a series of changes to data security and privacy in response to talks with officials. Only after those arrangements were settled did the model and the ChatGPT Work built on it reach users.

For Polish companies, this means the new tool will appear first in the corporate and education tiers before reaching regular Plus and Business subscribers in the coming days. The strongest interest is expected in departments that already use ChatGPT Enterprise for reporting and analysis.

The key question remains how much of the work the agent genuinely finishes without human oversight, and how much still needs constant correction. Earlier rollouts of similar tools, including Claude Cowork, showed in usage studies that most sessions still involve relatively simple business operations like reports and checklists, rather than fully autonomous projects.

ChatGPT Work is set to expand to Plus and Business plans within the next few days, which will make it possible to judge whether the tool actually cuts down the time needed to complete office tasks or remains an impressive demonstration of the model's capabilities.

Sources: OpenAI Debuts ChatGPT Work Workplace AI Agent With GPT-5.6 (forbes.com), OpenAI Launches ChatGPT Work, Deepening Race for Workplace AI Tools (bnnbloomberg.ca), OpenAI Launches ChatGPT Agent That Executes Complex Workflows (pymnts.com)