Skip to content 
OpenAI is expanding what ChatGPT can do, and it’s no longer just about answering questions. In a step forward, the company has introduced an AI agent that allows ChatGPT to carry out tasks online, such as shopping for users or making early drafts of PowerPoint presentations. This signals OpenAI’s broader goal to embed AI agents into everyday digital workflows, enhancing productivity by handling tasks that typically require humans.
The new ChatGPT agent was unveiled during a livestreamed event on Thursday. It builds on two previously launched tools Operator and Deep Research and is now available to Pro, Plus, and Team subscribers. According to OpenAI, the agent can browse websites, click links, make choices, and even accept real-time feedback from users as it completes a task. The rollout comes amid growing interest from tech giants like Microsoft and Anthropic in developing similar autonomous AI tools.
Get to know the incredible journey of chatgpt from our previous post from ChatGPT to GPT-5
ChatGPT Agent Brings Operator and Deep Research Together
The ChatGPT agent integrates Operator, which enables the AI to interact with web pages like a user, and Deep Research, which handles lengthy online information gathering. By combining these tools, OpenAI aims to automate activities that require multiple steps — like searching for items, filtering results, and summarizing findings.
OpenAI’s product manager Neel Ajjarapu demonstrated the capabilities of the agent by asking ChatGPT to search Etsy for vintage lamps under $200 with free shipping. The AI was able to find suitable products, add them to a shopping cart, and provide the URLs. The model powering the tool performs all actions using a web browser interface, mimicking human input to complete assigned tasks.
Ajjarapu stated that the tool can take feedback mid-task, allowing users to adjust the instructions as the process unfolds. He noted that although the technology is still developing, it represents a meaningful shift in how users will interact with AI.
Presentations and Drafts: Early Stages of Productivity Automation
Beyond online shopping, the ChatGPT agent can also assist with generating presentation slides. While these are currently limited to rough drafts, OpenAI says the tool helps users get started with visual content they can later refine. Microsoft, which owns PowerPoint and has a stake in OpenAI, already offers similar drafting tools for professionals.
OpenAI’s Chief Product Officer Kevin Weil acknowledged during the event that the software remains a work in progress. “It is far from perfect,” Weil said, but added that the current capabilities would have seemed “exciting” if imagined six to twelve months ago. The emphasis is on time-saving and reducing the cognitive load of repetitive tasks, even if user oversight is still needed in many cases.
Also read on our previous post on how AI Driving Automation in US Manufacturing
Rollout Timeline, User Access, and Safety Limitations
The ChatGPT agent is available immediately to users subscribed to the Pro, Plus, and Team plans. OpenAI confirmed that enterprise and education customers will gain access later this summer. The company also clarified that while the tool can perform a wide array of functions, it does not operate unchecked.
Certain things like making purchases or giving legal, or financial advice need user permission or are deliberately limited. Open AI explained that the agent will ask the user before carrying out such delicate undertakings. In the case of emails and other such transmissions, the system is going to need user monitoring so that such would not be misused or sent inadvertently.
Such AI agents bring about novel safety and security dilemmas. OpenAI admitted that there might exist the possibility of abuse or AI errors and claimed that they are already guarded against to reduce any damage. These are task filters, permission protocols and in certain scenarios, a default rejection of high risk requests.
AI Agents The New User Productivity Frontier
OpenAI CEO Sam Altman declared that AI agents are going to become the next giant breakthrough as far as artificial intelligence is concerned. The mission is to develop devices able to process any multistep tasks in the digital world with the minimum of human involvement, enabling the user to think at the next level (or take the actual decision-making).
Other industry participants investing in agentic systems aside Microsoft, a lead investor in OpenAI, include competitor companies such as Anthropic. These are very different tools, moving through passive AI responses into active AI support, able to manage calendars, do research, make to-do lists, and, recently, online shopping, or slide deck composition.
Nevertheless, its present deployment is a preview of what can happen. OpenAI emphasized that their agent is in no way the final product. The company considers this stage to be sort of experimental, which is aimed at collecting user feedback and improving the system in terms of the applicability and stability.
The introduction of the ChatGPT agent is an important development of generative AI. It is likely to change the way users interact with digital tasks, given that, as they grow, they will not type any command, but just ask ChatGPT to perform the task.
