OpenAI launches an AI agent in ChatGPT

OpenAI unveiled an AI agent for ChatGPT on Thursday to bridge research and action. It can now think and act proactively when choosing from agentic skills to complete a user’s task.

ChatGPT can perform different tasks, such as ordering a wedding outfit, while also considering the weather and dress code, using its computer equipped with tools that can interact with the web.

Users can also connect apps, such as GitHub and Gmail, allowing the chatbot to find prompt-relevant information.

AI agents are said to be the evolution of an assistant, which are now being globally embraced by giant companies, including Microsoft and Oracle, that are spending billions on the tech to improve productivity and streamline their operations.

The AI agent combines different qualities of its former agentic features, Operator, that can interact with websites and perform advanced tasks, such as multi-step research.

Pro, Plus, and Team users can activate ChatGPT’s agentic capabilities starting Thursday.

How to Use the New Agentic Capabilities

Users can activate it through the dropdown menu from the composer. They can choose agent mode (at any point) and just describe what they want to do (e.g.,  creating a slideshow, conducting deep research, or submitting expenses).

While performing a task, an on-screen narration provides visibility into what it’s doing.  Users can also interrupt and take back control of the browser when needed, ensuring tasks are aligned with their goals.

The AI agent can access a user’s connectors, integrating into workflows and accessing actionable and relevant information.

These connectors allow ChatGPT to see information and perform tasks, such as summarizing your inbox for the day or looking for timeslots you’re available for meetings, once it’s authenticated. But then, users will still be prompted to log in.

They can also schedule any completed tasks to automatically recur, like generating a metrics report every Monday morning.

Stay tuned for more AI news!