OpenAI has launched GPT-6 Astra, its most capable model to date, designed to move AI beyond answering questions and toward completing real-world professional tasks. The model combines advanced reasoning with computer use, browsing, coding and document creation, allowing it to work directly across websites, desktop applications and internal business tools.
Unlike traditional AI assistants that mainly generate text or provide recommendations, Astra can take action on a computer. It can fill out online forms, update customer records in a CRM, organize calendars, conduct research, work with spreadsheets and presentations, install and test software, and troubleshoot issues visible on screen. It can also create websites and perform frontend quality checks.
That makes GPT-6 Astra particularly relevant for businesses where employees spend significant time moving information between different systems.
From answering questions to completing workflows
The biggest change is not simply that Astra is “smarter.” It is that the model can connect multiple steps into a single workflow.
A manager could, for example, ask Astra to prepare a weekly business review. The model could gather information from several sources, analyze the data, update a spreadsheet, create a presentation using the company’s template and highlight issues that require human attention.
The same approach can be applied to sales, marketing, finance, HR and operations. OpenAI says Astra can work across existing websites, desktop applications and internal tools without requiring a dedicated API integration for every system.
This is an important shift in how companies can think about AI adoption. Instead of adding an AI chatbot to an existing process, organizations can potentially delegate an entire sequence of repetitive actions to an agent.
Performance is also improving
OpenAI reports a significant improvement in computer-use performance. On the OSWorld 2.0 benchmark, GPT-6 Astra scored 72.6%, compared with 65.7% for GPT-5.6 Sol. In OpenAI’s latency simulations, Astra completed tasks in roughly 40 minutes on average, versus about 75 minutes for GPT-5.6 Sol — around 47% less time.
On ScreenSpot-Pro, which evaluates an AI model’s ability to understand and interact with graphical interfaces, Astra scored 92.7%, compared with 76.9% for GPT-5.6 Sol.
OpenAI is also positioning Astra as a model for producing finished business artifacts rather than just drafts. It is trained to create documents, presentations, spreadsheets and analyses that follow existing company templates, writing styles and formatting standards.
What it means for managers
For companies, the key question is shifting from “Where can we use AI?” to “Which workflows can we delegate to AI?”
The most immediate opportunities are likely to be processes that are:
- repetitive and rules-based;
- spread across several applications;
- time-consuming but relatively low-risk;
- easy for a human to review at the end.
That could include preparing reports, updating CRM records, collecting market information, processing routine documents, creating presentations or performing first-level analysis.
At the same time, Astra does not eliminate the need for human oversight. The more autonomy an AI agent receives, the more important permissions, monitoring, approval steps and clear boundaries become — especially when it can modify records or take actions in business systems.
GPT-6 Astra is being rolled out in phases, initially to a limited set of organizations, with access expanding to ChatGPT Plus, Pro, Business and Enterprise users as well as developers through the API, Microsoft Azure and AWS Bedrock.
The broader trend is clear: AI is moving from a tool employees consult to a digital worker that can execute parts of their job.
The next competitive advantage may therefore come not from having the best AI assistant, but from identifying the workflows that can safely be handed over to one.
