OpenAI has launched new tools to help businesses and developers create their own advanced AI agents. These automated systems can perform tasks independently, make workflows faster and more efficient.
Introducing the Responses API
The launch is centered on the Responses API, and serves as a replacement for OpenAI’s Assistants API, which will be phased out in 2026. This tool allows businesses to build AI agents capable of:
- Conducting web searches
- Scanning company files
- Navigating websites
Are AI Agents Important?
AI agents are gaining attention, but many companies struggle to define or showcase their true potential. Just recently, the Chinese startup Butterfly Effect launched an AI agent called Manus. The platform went viral but failed to meet expectations.
OpenAI’s Approach to Smarter AI
Olivier Godement, OpenAI’s API product head, acknowledges the challenges. “It’s easy to demo an AI agent,” he said. “Scaling it and ensuring people use it regularly is much harder.”
OpenAI came out with its AI agents ‘Operator’ last year and ‘Deep Research’ this year. The company has acknowledged that launching and getting users for an AI agent are two separate things. The former is relatively easy, while the latter may prove to be a great feat.
Both tools demoed AI automation, however, they lacked full autonomy. With the Responses API, OpenAI aims to take things further. Developers now have access to the same core components powering these tools. This means businesses can create more autonomous AI applications.
What’s Inside the Responses API?
Developers using the Responses API can tap into OpenAI’s latest AI models:
- GPT-4o Search
- GPT-4o Mini Search
These models power ChatGPT’s web search tool, they browse the web for answers, providing sources along the way. OpenAI claims these models are highly accurate. On the company’s SimpleQA benchmark, GPT-4o Search scored 90%, while GPT-4o Mini Search scored 88%. For comparison, the much larger GPT-4.5 model scored only 63%.
The Responses API also includes a file search tool that scans company databases. OpenAI assures businesses that these files will not be used to train its AI models.
Automating Computer Tasks with AI
Another key feature is OpenAI’s Computer-Using Agent (CUA) model. This technology enables AI to perform mouse and keyboard actions. Businesses can run the CUA model locally on their own systems. However, the consumer version in Operator is limited to web-based actions. Developers can automate:
- Data entry
- App workflows
- Other routine computer tasks
AI Agent’s Challenges
Despite these advancements, AI agents still face hurdles. AI search tools, while accurate, are not perfect. GPT-4o Search still gets 10% of factual questions wrong. AI search tools also struggle with short, direct queries like “Lakers score today.”
Additionally, OpenAI admits the CUA model is still in development. It is not yet reliable for fully automating tasks across different operating systems. Mistakes can still happen.
OpenAI’s Commitment to AI Agents
OpenAI is continuously improving its technology. Alongside the Responses API, the company is releasing an open-source toolkit called the Agents SDK. This set of tools helps developers:
- Integrate AI models with internal systems
- Implement safeguards
- Monitor AI agent performance
This follows OpenAI’s Swarm framework, a multi-agent orchestration system introduced last year.
Godement remains optimistic. He believes AI agents will be the most impactful AI application. OpenAI CEO, Sam Altman, echoed this sentiment earlier this year. He predicted that 2025 will be the year AI agents enter the workforce.
