OpenAI's Deep Research Agent automates white-collar office tasks

OpenAI has introduced a new tool called Deep Research, aimed at simplifying some office tasks. This AI agent can explore the web on its own, deciding what links to click and what information to gather for a report. Isla Fulford, a researcher at OpenAI, played a key role in developing Deep Research. Before its public launch on February 2, Fulford noticed strong interest from her colleagues when the tool went offline. Since its release, users outside of OpenAI have also shown excitement. Deep Research has already impressed industry leaders. For example, Stripe's CEO, Patrick Collison, shared on social media that the AI had produced six reports in one day and praised its quality. Additionally, Dean Ball, an AI policy expert, noted its impact on the policymaking community in Washington, D.C. Users can access Deep Research through the ChatGPT Pro plan, which costs $200 monthly. When given a task, like writing a report on a specific topic, the AI searches for relevant content, reviews it, and crafts a detailed report with citations and data. Unlike many existing AI tools, which are relatively simple, Deep Research employs a more advanced type of reasoning. It provides insights into its thought process as it works, allowing users to see how it decides what information is valuable. OpenAI sees potential for Deep Research to handle a wider range of office work. The AI could be adapted to produce specific reports or presentations based on internal company data. Researchers hope to develop it further for various tasks beyond web searches. Interestingly, users have also started using Deep Research to generate code, which has surprised the development team. They are still exploring the implications of this unexpected use.