Your ultimate guide OpenAI’s new Agent Builder
OpenAI Agent Builder
& AgentKit from OpenAI
Some Background
OpenAI launched their visual Agent Builder, which is a flow builder to create agentic applications or orchestrate multiple AI Agents in one agentic app.
Earlier this year, I wrote a detailed blog on the difference between Chatbots and AI Agents.
Three areas where agentic flow builders are emulating chatbot build frameworks are:
- The classic graph-based method (think nodes and edges, aka traditional flow building) seemed destined for obsolescence, overshadowed by AI agents that dynamically spin up task sequences on the fly. Yet, a wave of innovative AI Agent flow builders has flooded the scene: Flowise, LangFlow, LangChain’s LangGraph, LlamaIndex, Stack AI, and countless others. OpenAI’s entry only amps up the excitement.
- Other elements which comes from chatbots have continued their importance, these are conversation classification or intent detection, and the
- the importance of managing of context.
Another element to take into consideration is that the term “AI Agents” are not trending anymore, and has been superseded by the term “Agentic Workflows”. Some refer to this as RPA 2.0.
The OpenAI Ecosystem
There are so much hype around the entry of OpenAI into the Graphic Agent Flow Builder market, but there are a number of things to keep in mind…
The agentic flow builder market is already crowded! And with many highly capable platforms.
Consider this, you have chatbot development frameworks which have all pivoted into this space. Together with new entrants, as I have mentioned earlier in this post.
And then traditional model providers like OpenAI is also entering this space.
If I was a startup focussing on a flow builder, the message is clear that you are in it for the long haul.
There are two things one needs to also keep in mind amidst the hype…
Firstly, the Agent Builder is really a low fidelity UI builder. Considering the image below, on the left is the list of available nodes, once nodes are placed in the workspace, a properties window can be accessed to configure individual nodes.
Secondly, considering the image below, one must realise that the Agent Builder forms part of a much larger OpenAI Agent ecosystem. the AgentKit, ChaKit and Widget builder are all stellar components in the OpenAI strategy.
One really need to look at the wider picture…the technologies OpenAI has been focussing on are:
- Building out an AI Agent ecosystem (as seen above).
- Orchestrating multiple small language models (SLM’s)
- Model distillation / fine-tuning
- AI Evaluations (Evals)
To my point of OpenAI building and ecosystem bit by bit, look at the image below. Under the properties of the Agent node, is a Tools option. File search, web search, MCP integration and more all part of the tool collection which can be used.
The ability to perform web search with options like Country, Region, City etc. are all elements which can be used to create some kind of contextual reference.
The widget builder lets you describe the widget you would like to add to your conversational user interface. As seem below, you can describe the widget in natural language.
Below is the widget which is created…
And here it is incorporated into my agtentic workflow.
In all honesty, every time I tried to run a workflow, I get the error below…
If you have made it this far in the article, thank you for your time and interest.
The hype is real, but one needs to step back and consider the complete picture…
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. Language Models, AI Agents, Agentic Apps, Dev Frameworks & Data-Driven Tools shaping tomorrow.
https://openai.com/index/introducing-agentkit/
https://platform.openai.com/docs/guides/agents/agent-builder
