OpenAI Unveils AI Agent Development Tools for Businesses

# agents # openai

On Tuesday, OpenAI unveiled a suite of new tools aimed at assisting developers and enterprises in crafting AI agents—systems capable of autonomously performing tasks—using OpenAI's proprietary AI models and frameworks. These tools are part of OpenAI's newly introduced Responses API, which enables businesses to create custom AI agents capable of conducting web searches, sifting through company files, and navigating websites, similar to OpenAI's Operator product. The Responses API is set to replace the company's Assistants API, which OpenAI plans to phase out in the first half of 2026. The buzz around AI agents has been growing rapidly in recent years, despite the tech industry's challenges in clearly defining and demonstrating what "AI agents" actually entail. A recent case in point is the Chinese startup Butterfly Effect, which gained viral attention for its Manus AI agent platform, only for users to find that it fell short of the company's bold claims. The pressure is on for OpenAI to deliver on the promise of AI agents. "It's pretty easy to demo your agent," said Olivier Godement, OpenAI's API product head, in an interview with TechCrunch. "To scale an agent is pretty hard, and to get people to use it often is very hard." Earlier this year, OpenAI introduced two AI agents within ChatGPT: Operator, which navigates websites on your behalf, and deep research, which compiles research reports. While these tools offered a glimpse into the potential of agentic technology, they left much to be desired in terms of autonomy. With the Responses API, OpenAI aims to provide access to the core components that power AI agents, enabling developers to create their own versions of Operator and deep research-style applications. The goal is for developers to build applications that feel more autonomous than what's currently available. Developers can leverage the Responses API to use the same AI models powering OpenAI's ChatGPT Search web search tool: GPT-4o search and GPT-4o mini search. These models can scour the web for answers to questions, citing sources as they generate responses. OpenAI asserts that both GPT-4o search and GPT-4o mini search are highly accurate. On the company's SimpleQA benchmark, which evaluates models' ability to answer short, fact-based questions, GPT-4o search achieves a 90% score, while GPT-4o mini search reaches 88%. In comparison, the much larger GPT-4.5 model scores only 63%. The Responses API also includes a file search utility that can rapidly scan company databases to retrieve information. OpenAI assures that it won't use these files to train its models. Additionally, developers can access OpenAI's Computer-Using Agent (CUA) model, which drives Operator, allowing them to automate tasks such as data entry and app workflows by generating mouse and keyboard actions. Enterprises have the option to run the CUA model, which is currently in research preview, on their own systems, according to OpenAI. The consumer version of CUA available in Operator is limited to web-based actions. It's important to note that the Responses API won't resolve all the technical challenges facing AI agents today. While AI-powered search tools are more accurate than traditional AI models—an unsurprising outcome given their ability to simply look up answers—web search doesn't eliminate the issue of AI hallucinations. GPT-4o search still answers 10% of factual questions incorrectly. Moreover, AI search tools often struggle with short, navigational queries like "Lakers score today," and recent reports have questioned the reliability of ChatGPT's citations. In a blog post shared with TechCrunch, OpenAI acknowledged that the CUA model is "not yet highly reliable for automating tasks on operating systems," and it can make "inadvertent" mistakes. However, OpenAI emphasized that these are early versions of their agent tools, and the company is continuously working on improvements. In addition to the Responses API, OpenAI is releasing an open-source toolkit called the Agents SDK, which provides developers with free tools to integrate models with their internal systems, implement safeguards, and monitor AI agent activities for debugging and optimization. The Agents SDK is an evolution of OpenAI's Swarm framework for multi-agent orchestration, released late last year. Godement expressed hope that OpenAI can close the gap between AI agent demos and practical products this year, stating that, in his view, "agents are the most impactful application of AI that will happen." This sentiment aligns with a statement made by OpenAI CEO Sam Altman in January, predicting that 2025 will be the year AI agents join the workforce. Whether or not 2025 indeed becomes the "year of the AI agent," OpenAI's latest releases demonstrate the company's shift from showcasing impressive agent demos to developing tools with real-world impact.

Related article

Comments (20)