OpenAI AgentKit Tutorial: Build Your First AI Agent in 8 Minutes

Speaker A: I just built a fully functional agent in under eight minutes, no code, no API configuration. And this is right off of OpenAI's new agent builder. Okay. It has the ability to change everything. And by the end of this video, what you'll be able to understand is what this really looks like. At the same time, what is Agent Kit and Agent Builder, who's IT for and how it will affect the future of AI platform domination. For those of you who don't know me, my name is Ryan Staley. I have been working with PE backed, VC backed, Fortune 500 organizations to scale and grow their company without hiring through AI transformation for the last two and a half, three years, worked with over 2,500 different executives on it, working with some of the biggest companies in the world. And so today I'm going to get right into it, I'm going to walk through a real quick visual of these tools because basically what happened on October 6, OpenAI dropped agent Kit at Dev Day in San Francisco. And it's arguably one of the biggest developer announcements of the year specifically for OpenAI. In addition to another one they had where you could actually leverage apps specifically within ChatGPT. All right, I'm not going to get into that today, but it's something I want to call out and the centerpiece is Agent Builder, which is a visual drag and drop campus. And I'm going to show you and really walk through. Is this something that anybody could use? And ah, what are the future implications? So without other delay, here we go, we're going to get right into it. So let me share my screen and I'm going to walk you through exactly what's possible, how to use it, uh, all these different areas. So let's get right into it. And what you can see is you're going to see the starting page, right? So over here, what this is, is this is what it'll look like when you walk in. Now what you need to do is this isn't available through a normal ChatGPT account. You need to pull this up through the API, basically. So you have to create an API account, you could use it under the same email address and what'll happen is you'll get to this, what's called like a playground API playground. Uh, most of the time it'll default to chat, but Agent Builder is where it's at. Now remember, this is only in beta right now, but uh, I'm going to show you really quickly how you can make it agent in less than eight minutes. So what you're going to see here is um, there's templates that they already have. I think that is the number one fastest way to build a QWIK agent to see what it's like and test it out. So uh, at the same time there's going to be drafts which are just different agents I was working on and then workflows. Okay, so this is one specifically designed for planning and uh, making complex decisions. That's the one that I'm using right now. Now once again I use this simple by design. And the reason being is because of the fact that I wanted to not overcomplicate this. Once I start getting under the covers on this, there's a lot of real technical details that most people are going to walk right by. So what you're going to see here is let me walk through the left and then you'll see what's going on on the right. I'll explain this. So on the left basically this is just the flow. If you want to work within this, you just literally click on it and you can go with inside that node. Okay, so each one you can click and work through this. However, I started a preview of it just to test it out and see what the output quality is. So what you're going to see here is I go a little meta, uh, in terms of asking it to really identify use cases for an AI skill training and AI, uh, consulting company to use Agent Builder internally or externally for its clients with publicly traded and tech forward companies. Okay, so what's happening is it's pulling this up and it just like deep research. So if you haven't used deep research, this is one of the things that often comes up where they're like hey, we need to clarify exactly what you're looking for and give it more context. And so that was one of the things that it did there. Uh, so I basically forgot to answer this one. Um, but I answered what's the initiative goal? How many initial use cases, blah. And what it also did is I said okay, two to three people, three people for headcount. Uh, and then it went through in a process. And as you can see, uh, under the covers this is a reasoning agent. So there's different models you could pick to go through. But what it started to do is actually automate or identify the areas where I could use Agent Kit to automate different areas or Agent Builder if you will. So it's got automated internal knowledge base, project management, AI powered proposal and report generator, and then it's even got top use cases for sales. Right Sales enablement, marketing campaign, outbound agent, uh, customer support agent, risk and mitigations. And then at the same time it's got a proposed 90 day timeline. Right. I then said give me the step by step nodes if I wanted to build this. And as you can see my prompting is very kind of not very sophisticated in this. I was trying to go through this quick to see and test it. Now there's other ways where you could test it as well and I'll show you that in a second. Uh, but basically what it is is it then walks through step by step with the nodes and then I even had it include the prompts for each stage. So this is what it fed out from really simple instructions and confirmation. And then it built this out. Okay, so I used the template like I said before, but then I tested it with simple streamlined prompting that wasn't very complicated. Uh, so that's a way where you could look at it. Now that's very, very simple. I think one of the things that you'll see over here is when I click on it and you click on the different nodes, there's going to be different areas that are a little bit more complicated. So if you're not used to using this, it's going to be a lot. So if you're a non developer it's going to be a lot. I think you go through obviously picking the models isn't super complicated, but then as you get into capabilities like output format, um, you know you have text, JSON, widget, not super complicated there and then other areas that you can look at fine tuning this in. So it's really just kind of going step by step through this. Then there's conditionals and if else while user approval, different areas like that. And then just to kind of highlight it, the thing that Alano developers are excited about were guardrails and MCP connectors within here. Uh, at the same time you could actually create an eval for this. Which uh, what an eval is is basically uh, you could identify a test that you wanted to have to pass every single time so you could identify how good what you built is on a continuous basis. Once again, this is more for enterprise grade type solutions. Okay, so we have that and so I'm going to get back to like my overall summary. I wanted to show you a few other things with Chat Kit. So Chat Kit is another area that a lot of folks are really excited about that is related to the agent builder. And so what this is is as you could see, you can embed it in the front end, um, and then also use it with Agent Builder or you can run it on your own infrastructure. So what I see a lot of folks doing is putting these chatbots on their website or putting them within their products and building these directly in openhand. So they massively simplified this and as you can see it's got kind of a step by step walkthrough of it. Now they also had an example of a demo basically called Chat Kit World, which is a demo on here. So what I could always do is I could say okay, I could hit on one of these areas which, where should we visit next? And so I could say okay, where haven't I been? And it's going to show me a lot of places where I haven't been because I haven't. I only use this for a couple of different examples. So it's like you visit 121 out of 175. Okay, so let's look at, I could just say look at this. Well, I don't know why the hell I want to go to Antarctica. But what you're starting to see is this is how it works. It then zeros in on um, what the location is and it kind of talks to Antarctica what it is, what's involved there. And then it allows you to, you know, visit it. Right, which is really interesting. So if I wanted to visit it, trying to see here, nothing happened, um, allows you to spin the globe, rotate it, understand what's going on. So here's what I would say. I think this is interesting. I uh, think there's going to be a lot of utility on people's websites to leverage this in other areas within product. However, uh, like I said, it's one of those things where there does require technical capabilities to be able to build this. In addition, there's what's called Widget Builder. And so what Widget Builder is, is something that you could put as a tool effectively for the AI agents to use. And so they have these pretty pre built kind of tools that you'll see which are design structure, kind of output orientated. So uh, it's pretty interesting from that perspective. So you could basically just include this and have it uh, once you ask for an output, load up in one of these formats. Last but not least is Chat Kit Studio. So what this is, is this is a playground. I'll drop this in the, actually this section within the notes because it'll give you capabilities where you could scroll down here and build your own implementation. You could explore this and play with this, um, and then you could see working examples. So some pretty good areas like that the chat kit playground looks exactly like what you'd expect it to. It looks like a chatbot. Right. So you can put that in there and then like I said, embed this into your product, embed this into your website and that's really the core capabilities. Now you're probably wondering like, all right, what does this mean for me? What, what does this mean for what's happening with the Future and what OpenAI is doing in here? And so there's a couple things was pricing was not really revealed so we have no idea how much this is going to cost. At the same time a lot of folks are saying is that this is killing N8N or this is killing Zapier. Well, one of the key mechanisms that this uses is Zapier as a connection mechanism to talk to different apps. So I don't necessarily think that's the case. I think that's almost like BS headlining just to clickbait you into seeing what's going on. However, this is starting to look similar to what N8N is creating and I think there's going to be more and more heated competition with it. Ironically, N8N just got an investment from Nvidia and is valued at 2.5 billion now in terms of its valuation. So when we summarize this though, what I would say is the number one thing is Agent builder is good for startups. Building AI first products I would say is definitely more developer focused now, but even just by using a template, working with it, I think you'll be able to start building agents relatively easy. However, it's going to take some time. I just started using Claude code, uh, about a month and a half ago myself, am blown away with what some of the more technical coding agents are able to do that even the normal models at the highest level aren't able to do. Okay, so what I would say is I think this is going to be huge for the enterprise space because just like there is vendor lock in with other areas, folks might transition instead of building custom GPT to build an internal agents. Organizationally however, it's going to require a combination of people understand the business logic and the technical workflow. I think the technical workflow is going to get closer and closer to the business logic and that's the path that all these models are heading. So really appreciate you checking out this episode. What I'm going to do is, as you see, there's a link below where you could grab different props use cases, capabilities, and I'm going to keep building these out. But this is something that I felt really obligated to cover because I think there's so much misinformation on this right now in terms of all the tools. It's killing everything that's possible, whereas in reality, it's a beta product that has tons of potential. But also, uh, is going to take a while to kind of get going. So appreciate having you on this video and we'll see you on the next one.