4 Reasons Why OpenAI's Agent Builder is NOT the Zapier Killer

Hi everyone, it's Helena here and welcome back to my YouTube channel. And in this video, I'm going to give you my thoughts on OpenAI's latest release, which is the Agent Builder. I'm going to tell you why I think this is mediocre at best, and I'm going to tell you 5 reasons why I don't think this is the Zapier killer. So essentially, OpenAI has noticed like a lot of people have been using their API and using their technology off of their platform in their own automations or on their own software, and they make the connection through using automation tools like N8N, Zapier, or Make.com. So they're like, why don't we keep everyone in our own ecosystem? And then why don't we make an automation builder as well? So this is how Agent Builder was born. But I'm going to actually show you how to use this. And I'm actually going to show you why I think it's actually a very limiting tool. So to get to Agent Builder, go to platform.openai.com/agent-builder, you'll be taken to this page. And let's go ahead and create a new workflow. So at first glance, the UI is very clean. So you drag and drop the different nodes that you want in your automation and then connect them through the flow. So you can see what step 1, what step 2, what step 3. So in terms of ease of use, I think it's pretty good. Now the first downside of Agent Builder is that there's only one start node here, right? So if you use something like Zapier, you know, your trigger or your start node could be someone filled in a form, someone bought a product, It could be a specific date and time. It could be someone sent you an email. Like there's literally thousands upon thousands of triggers you could have. For the agent builder, there's only one and it's only a chat input. So essentially the agent builder can only be used for chatbots at the moment. Now the second downside of agent builder is when you look at the tools section, there's only 3. There's file search, guardrails, and MCP, and that's it. So if you look on Zapier, there's over 7,000 natively connected tools with all sorts of software. So it makes your automation building a lot easier. And if you look at N8N, there's like over 400, right? And the third reason I don't really like Agent Builder is that there's not even a natively built-in HTTP tool that you can use. So it's very limiting in terms of which softwares that you can connect this to your automations. I get that there is an MCP server node, but there are a lot of tools that don't have an MCP. So again, this is a limitation. And when you're talking to marketing people, HR people, they don't know what an MCP is. So it makes it harder for people in your organization to go in, update the automations. And really I'm struggling to come up with a business use case for Agent Builder. Even when we look at the built-in MCP servers, Like this is it. It's very limiting. You got Gmail, you got Google Calendar, you got, you know, Outlook email, Box, Square, PayPal. But this is pretty much it, right? So you could connect your own server. This does require a bit of tech skills in order to do that. And it's also an extra step because you have to now connect your HTTP request to an MCP server, get that MCP server up and running, connect that to here in order to use our automation. So I don't see the point of doing all of those extra steps when you can just use Zapier or n8n, quite frankly. So I want to quickly talk about like the difference between an MCP and an HTTP request as well. So if we look into the Gmail MCP, for example, that's natively built in by OpenAI. Now if we look under tools, we can see there's a whole bunch of tools here. So we can batch read emails, get profile, get recent emails, read emails, search emails, etc., right? So each and every single one of these tools It's an HTTP request. Okay, so MCP is a collection of different API endpoints. And now instead of you having to say, first you get the email, then you read the email, then you reply to the email, you give all of these endpoints to the MCP server, and you give this entire MCP server to an AI agent. And now based on your prompt, the AI agent can now decide which one of these endpoints it will need to use to complete the request that you have to have. So instead of making your AI automation linear— first you do this, second, then third, etc.— you actually allow the AI to determine which endpoint to use to complete the request in your automation. It's basically the next level up from an HTTP request. Now the limitation of it is because this technology is still so new, there's a lot of softwares that don't have an MCP server yet. And because Agent Builder does not have a built-in HTTP request, there's a lot of limitations in terms of what you can build inside of your Agent Builder workflow. And lastly, to deploy this automation that you just built, you press on Publish. And after it's published, you click on Code and you'll be given a ChatKit and Agent SDK. So you're given all of these code. If you're not techie, you have no idea what to do with all of this. And honestly, I didn't spend the time to figure it out because I just don't see the point of it. Like, why would I spend my time trying to figure out how to deploy all this code when I can make this entire automation on Zapier or N8N and I don't have to be limited in terms of my deployment? So those are all of my reasons why, like, I don't think Agent Builder is really production ready. Ready. I think it's a good concept. I can see where OpenAI is going with this, but the tool as it stands right now, I don't think it's that useful. Now I want to wrap up this video by talking to you about my thoughts of N8N versus Zapier. Now at first glance, N8N seems cheaper, and I thought so as well, but there are a lot of hidden costs to N8N that you need to be aware of. First of all, there's only 400 built-in nodes, so this means that if you're using software platform that's not natively built in, you have to create your own HTTP module, right, to make it work. So that takes time, it takes effort. And if you don't have the skills to do so, it's $100 to $200 to hire someone to build that for you. And all of a sudden, like, you're saving, you know, $10, $20 a month on your subscription bill, but you are spending more time building your automations. And you have to maintain those HTTP modules as well. So when you built it natively, when something changes, you got to go back in, you got to make sure it's all running again. So for me, I prefer the simplicity even though it costs me a bit more every single month. And when we start to look at like when you start running a lot of workflows every single month, I'm talking about like more than 100,000 or so, Zapier actually becomes cheaper than n8n. So n8n gives you the self-hosting option, but there are a lot of hidden costs to self-hosting that you need to be aware of. The first one is you need an engineer to set up all of the self-hosting for you to make sure it's secure and everything's running. So that could easily be $50,000 to $100,000. On top of that, you had to pay for the cloud infrastructure, you had to pay for the backup, you have got to pay for the cybersecurity, you got to pay for the hardware costs, you got to pay for the monitoring, alerting, etc. So it could easily add up to 6 figures when you're trying to self-host everything. That's something to consider. Consider. So you gotta consider all of these extra costs and do the pros and cons to see if it makes sense for you to self-host on N8N versus just having a cloud infrastructure that's compliant and works out of the box, easy to use, like Zapier. Okay, so I wanted to raise this point because I've seen a lot of YouTube videos going around saying like you can run workflows for free when you self-host with N8N, but you gotta be aware of these hidden costs. All right, so this wraps up my video. If you got anything out of it, please like and subscribe and also join my free course. The link is going to be below, productcams.com/free, and I'll see you guys next time. Thanks so much.