Editor
Celine Low chevron_right
The buzz around AI agents has become around for some time now. Big companies like Google and Microsoft have already rolled out their own AI agents that can handle complex tasks and workflows autonomously with minimal human intervention.
Now, OpenAI has jumped on the wagon, officially launching ChatGPT Agent yesterday, 17 July.
How Does ChatGPT Agent Work?
According to The Verge, the company bills it as a tool that can complete work on your behalf using its own “virtual computer.”
ChatGPT Agent combines OpenAI's existing "Operator" (which browses the web) and "Deep Research" (which generates detailed reports) tools.
Yash Kumar and Isa Fulford, the product and research leads behind ChatGPT Agent, explained that this new tool is powered by a specially developed, unnamed AI model.
ChatGPT Agent can perform a variety of real-world tasks, such as:
-
Calendar Briefings: It goes through your calendar and give you a quick summary of upcoming client meetings.
-
Meal Planning & Shopping: It noy only suggests recipes but also helps plan and even purchase the ingredients!
-
Slide Deck Creation: It can analyse information (e.g., competing companies) and then create a full slide presentation for you.
-
Date Night Planner: Besides client meetings, it can cross-reference your free time on Google Calendar with restaurant availability on OpenTable to plan the perfect evening.
This new model was trained using a technique called reinforcement learning (the same method used for OpenAI's other "thinking" models) on complex jobs that require using multiple tools like a web browser, a visual browser, and even a "terminal" where you can import your own data.
Beyond Chatbots
Kumar shared that he started using the AI agent to automate small parts of his own life, like making sure he requests office parking every Thursday so he doesn't forget on Monday morning. Isa Fulford, a research lead, finds it incredibly useful for online shopping, as it's more thorough than previous AI Browse tools.
While the tool might be a bit slow for instant results, the team argues that even if it takes 15 to 30 minutes, it's a huge time-saver compared to doing these multi-step tasks yourself. You can "kick something off in the background and then come back to it later."
Importantly, OpenAI has built in safeguards. Before ChatGPT Agent performs any "irreversible" actions such as sending an email or making a booking, it will always ask for your permission first.
For now, financial transactions are restricted, and a "Watch Mode" protects sensitive pages by stopping the tool if you navigate away from its active tab on sites like banking portals.
Available For Use
Starting today, ChatGPT Agent is rolling out to Pro, Plus, and Team users.
You can access it by selecting "agent mode" in the tools menu or by typing "/agent."
It will be available to ChatGPT Enterprise and Education users later this summer, though a timeline for Europe and Switzerland is still pending.
Stay updated with ProductNation on here, Instagram & TikTok as well.
Read more related news here: