Automate Your Agency

The AI Tool Breakdown: When to Use GPTs vs Assistants vs Agents

• Alane Boyd & Micah Johnson • Season 1 • Episode 64

Share episode

Send us a text

Are you drowning in AI terminology and wondering which tools will actually move the needle in your agency? Custom GPTs, AI assistants, AI agents—everyone's throwing around these terms, but nobody's explaining what you should actually be using.

In this episode, Alane teams up with Beau Cunningham to cut through the confusion and give you a crystal-clear breakdown of when to use what. No more wondering if you're missing out on the "better" solution.

You'll discover:

  • The real difference between GPTs and AI agents (and why it matters for your workflow)
  • How to train a custom GPT on client knowledge so new team members aren't starting from scratch
  • Why GPTs require conversation while agents actually do the work for you
  • A live example of their automation tool that's been running for years

Stop second-guessing your AI strategy. Beau and Alane break down exactly which tool solves which problem, so you can pick the right solution and start seeing results immediately.

Whether you're just getting started or ready to level up your current AI setup, this episode gives you the roadmap to make smart decisions about your AI stack.

See a custom GPT in action by testing out Willy the Workflow Wizard! He's powered by a custom GPT that is supported by automation. Check him out here! 

Disclosure: Some of the links above are affiliate links. This means that at no additional cost to you, we may earn a commission if you click through and make a purchase. Thank you for supporting the podcast!

For more information, visit our website at biggestgoal.ai.

Want more valuable content and helpful tips?

Explore our Courses/Programs:

Enjoy our Free Tools:

Connect with Us:

0:00:00 - (Alane): Welcome to Automate Your Agency. Every week we bring you expert insights, practical tips, and success stories that will help you streamline your business operations and boost your growth. Let's get started on your journey to more efficient and scalable operations. All right, I'm changing things up. I've got Beau Cunningham with me today instead of Micah.

0:00:23 - (Beau): I am so excited. It's typically me and Micah when we're doing this, so I'm super excited.

0:00:27 - (Alane): I know we ditched Micah and said, hey, let's do this together, because we never get to have fun and leave Micah out.

0:00:34 - (Beau): True. I'm loving this.

0:00:36 - (Alane): So whenever I was thinking about what we could chat about on the episode, you brought something up. And this has been a while ago and it. And it's been in the back of my head, and it is the idea of using GPTs or AI assistants that you train and how that can be used and how they are different than when we're talking about an AI agent.

0:00:59 - (Beau): I love this idea, Alane, specifically because for some reason, I guess with just AI agents coming up and being more commonplace in the workplace, I'm getting this question a lot. You know, how do I use an AI assistant? A GPT versus using an AI agent? What's the difference? This has been a hot topic recently.

0:01:17 - (Alane): And I think that's why, you know, you had even started putting together some resources for our team, because it's coming up in conversations so often with clients. And I see agencies have really, like, started thinking about creating GPTs that their team can access that are trained on individual clients, right?

0:01:36 - (Beau): Absolutely. They can do that. I mean, anything, whether it's just trained on individual clients or trained on their internal knowledge, really. I mean, the GPT is what you want it to be, right? And when we're talking about GPT, just taking a step back, you know, when we think about it, a lot of times when we talk about LLMs, we're talking about Claude, we're talking about talking about chat GPT.

0:01:55 - (Beau): Right.

0:01:56 - (Beau): Sometimes the line gets blurred a little bit and Claude will say, hey, I'm an AI assistant.

0:02:00 - (Beau): Right?

0:02:01 - (Beau): That's what Claude will say it is. But when you're in Claude, you can create what's called projects, which is essentially what a custom GPT would be within Claude.

0:02:09 - (Beau): Right.

0:02:10 - (Beau): Similarly for OpenAI, they actually call them AI assistants or custom GPTs. So what you're doing exactly like you're talking about is you can go in, provide that GPT or project with information on an individual client, whether that's all their policies, procedures, their Brand kit, their media kits, whatever it is, you can provide that to it. You upload it in there and that information just sits there and it's just an expert on that one thing.

0:02:37 - (Alane): Yeah. And thank you for explaining that. And just like, will you take a couple of seconds and just say, you know what if somebody wanted to go and create one of these? They go in into ChatGPT or OpenAI and they want to create it and then what do they feed it with?

0:02:50 - (Beau): Yeah. So essentially inside of OpenAI, you have to go into the OpenAI side of things. Not necessarily ChatGPT, you go into that back end and it walks you through the process of building it out. And it can be anything, it can be a CSV, it can be a PDF that you fill it up with, but it essentially gives you a section where you can upload documents to it. So again, docs, PDFs, CSVs, Excel files, whatever you want to provide it access to, you can do that. Similarly, same thing for Claude. You create a project and then you just upload whatever you want it to be.

0:03:20 - (Alane): So again, basically training it. That's. That's what we call it. You're training it.

0:03:23 - (Beau): Yeah.

0:03:24 - (Alane): Just like you would anybody, an employee on your team, hey, this is what you need to know about this client or this process. You feed them with information. We're doing the same thing with the AI agent. I'm sorry, with GPT.

0:03:35 - (Beau): Absolutely. I think the. At that point, I mean, we very much still train our AI agents.

0:03:41 - (Beau): Right.

0:03:41 - (Beau): We give it access to different databases and knowledge bases and information. But, you know, really, I think the key difference between what we see in GPTs and AI assistants and all of that and actual agents is we've talked, you guys have talked about this in previous podcast episodes, but agents have agency. They can go and they can tackle a task for you based off the information you provide it. It can go and tackle that task, it can do things autonomously.

0:04:06 - (Beau): But really where you're seeing things from an AI assistant and GPT perspective is those are actually going to require some form of interaction.

0:04:14 - (Beau): Right.

0:04:15 - (Beau): Still have to chat with it.

0:04:16 - (Alane): I mean, it's still a one on one conversation.

0:04:19 - (Beau): Totally. Yeah, totally.

0:04:21 - (Beau): Right.

0:04:21 - (Beau): And I mean, we see a lot of people that'll create a GPT and they use it in the sense of, you know, if they're creating a brief on a client or they're creating a, you know, something that they need to achieve for the client, they're going in and they're just saying, hey, I need to come up with this can you give me information on this client? And it can help them brainstorm, write content just like an agent can. But you're still having to query it, right? You're still having to chat with it, essentially.

0:04:46 - (Alane): Yeah. So a great first step. And then because they are helpful and they're helpful with that team member that needs to get work done. But let's, let's talk about how they can work with AI agents. And like you just said, like an AI agent can go and get work done for you. And I think maybe let's use the client brief example just to make it easy.

0:05:05 - (Beau): Yeah, totally. So if you think about an AI agent, right, AI agents allow you to. You're essentially giving it a brain, right? That can be a LLM like Claude or Chat GPT or whatever model you want to utilize for that. And then you can also give it access to additional things. But, you know, and with that you can actually have a GPT or workflow that somebody triggers that particular workflow, whether that's from a CRM perspective, they move it into a deal.

0:05:30 - (Beau): Once they've moved it into that deal stage where it says, hey, we've won this, at that point in time, it can then trigger your GPT which understands what types of prompts it's typically going to receive. You've trained your GPT custom GPT on this is what I want your output to look like. And you've given it whatever information or knowledge you wanted it to have. And from there it can query that, process the information that's pulled through, and then be able to provide you with some sort of result. It's a static result, though.

0:06:00 - (Beau): It's a. I did it, you know, and it, it gives it to you. That's it? Yeah, that's it. But you can then take that and feed it into an AI agent. All of that information can go into that. And then it's able to, then the agent can use its tools, it can use its LLM that it's connected to, or it, its AI model it's connected to, to then take the autonomy to do different things with it. So at that point we could say, hey, we want you to create a task inside of ClickUp, Asana or Monday, and we want you to go and create a document out of what we just produced from that GPT and then we want you to attach that doc inside of Monday or maybe we want you to drop it into a Slack channel or create an email from that, an email draft in our email account.

0:06:43 - (Beau): So again, you can kind of pair the two together and just let the GPT do the lifting of the creative however you want it to be trained, but then let the agent actually go and take the action from there and do the job for you.

0:06:58 - (Alane): Yeah, and we, we did this, gosh, like it's been. How old is like a willie, the workflow, like two years, maybe longer. I don't remember now. So for, for our listeners, if you want to get an idea of what an input could look like and an output can look like with the GPT in the middle, now you're not going to see that piece, but you're going to get to see the beginning and the end result. We have. It's a free tool, it's the 25 automation idea starters and that is a dynamic tool for you to use where we ask you, I think it's eight questions on which software platforms that you use.

0:07:35 - (Alane): And then we've, we've trained a assistant, a GPT on automations and typical automations in a department software, all these things and the output that you receive is an emailed report custom to you with automations that you could put in right now in your organization based off the software that you use.

0:07:57 - (Beau): Yeah, and it's, it's pretty slick too because it really is. It's just utilizing an automation to where you type in that type form exactly like you're saying. And to Alane's point, we've gone into that GPT and just said this is what we want you to produce. Here's all of our knowledge be this portion of the process.

0:08:14 - (Beau): Right.

0:08:14 - (Beau): And then it queries it and outputs it, drops it into, you know, a document and sends it on its way.

0:08:20 - (Alane): Yeah, I mean it's been running for years now and it's super smart and intelligent and, and you can continue to train your GPT and, and edit and tweak it. You know, we went through a, I'll call it a massage. We went through a massage area where we were, you know, tweaking the prompts, tweaking its knowledge, making sure that we liked the outputs.

0:08:40 - (Beau): Yep. So as far as this is concerned, I think that there's, there's really some key differences, ways to think about the differences in all these things. You know, for me, you know, I think that chat or chatgpt or just Claude being able to chat with it just regularly, that's all well and good. That's a conversation that you're having, that's somebody you can brainstorm with.

0:09:00 - (Beau): Right.

0:09:01 - (Beau): And be able to just have a conversation with and get creative with and Produce the output, you're telling it what you want it to do and it does it. That's the kind of key deciding factor on what that is. And then the custom GPT is, it's kind of filling a role from a knowledge perspective. It's, you've trained it with knowledge and you can chat with it and it's just specialized in that one particular thing.

0:09:21 - (Beau): But then the agent is actually what can go and actually do work for you.

0:09:25 - (Alane): Now, one thing that I was thinking about is when, when it comes to clients and you have an individual working with a client, it's not a just, it's not just about the work that you're doing, it's also understanding the nuances of the relationship and the things that they find value in, which is a hard knowledge transfer between people if you make changes. And so I'm going to make, I'm going to make up something crazy. But let's just say you do, you're an agency and you do social media marketing and your client does not like LinkedIn marketing and do not bring it up.

0:10:00 - (Alane): So for a new team member coming on board, they may see that and go, what an opportunity. They're not doing LinkedIn marketing not realizing that this has been a history of the client and they don't believe in it. Well, that GPT could be trained to say, when you do a media plan or social media marketing plan, do not include LinkedIn. This client does not like it.

0:10:22 - (Beau): Yeah, you could absolutely do that. And I think something interesting as well, you know, when we talk about situations like that, you know, you're, when you're in a relationship with a client, you're working with them, that relationship does evolve quite a bit. They change their minds a lot. They, they'll say, I don't want to do this right now, but in the future, two months from now, they're like all about LinkedIn all of a sudden.

0:10:43 - (Beau): Right.

0:10:44 - (Beau): And so I think it's also important to remember that, yes, absolutely, you can train it on the, you know, idiosyncrasies of those clients, but at the same time there's kind of an evolution that's happening as well. So on a custom GPT, you do have to go back to it and you're going to have to retrain it somehow.

0:11:01 - (Beau): Right.

0:11:02 - (Beau): You're going to have to say, this is where we're at now. You know, we're now here with this client. And that's not to say that it can't be done.

0:11:09 - (Beau): Right.

0:11:09 - (Beau): You can train a GPT, you can utilize something like a make or an N8N to go back in and as you get additional information it can go in and update kind of the knowledge or the brain of that GPT. But when we start talking about, you know, agents and giving it access to your pipe drive notes or your, your CRM notes or your, you know, conversations that you're having, all of your call transcripts and all of that, and it just has easy access to it or a Google Drive full of that information, that's where you don't really have to go and manually update something. But with the GPTs you do. But to your point, if that information is loaded into a GPT, you can absolutely chat with it.

0:11:51 - (Beau): That GPT is 100% the person, that person, right, that knows that knows everything about that client and can give you those answers. But you do still have to give it its brain, you know, consistently. You have to update it manually to do that. Unless you build an automation to do it for you.

0:12:09 - (Alane): Yeah, so I like the two examples that we've kind of talked about here. Like one of them is separate GPTs for each client, which is what I do see with our clients, right, that they have em, so you, you have each little nuance between your clients, you use it to chat with, build things. But also the other example with the 25 Automation Ideas tool that we built, it isn't a client specific, it is a knowledge that we have on based on what we do to provide.

0:12:38 - (Alane): So like when you started off saying Beau at the beginning, like it really can be anything. It can be anything that you want it to be, you know, within limits. But just to, to, for, for our listeners to be thinking about this is know what is your secret sauce? What is that you do that would be helpful to have a trained assistant on. If we had to individually take in every person that filled out our 25 Automation Ideas form, which is what we call Willie the workflow wizard, then we'd have to have a person manually taking those inputs and creating the report and then sending it off. That would be, I mean the person would be moved on and, and bored have another thing that they were doing at that.

0:13:24 - (Alane): It just wouldn't make sense. So we were able to train in a GPT on that and have that as part of an automated workflow with an AI agent just for you all to be thinking about. It doesn't have to be just client, each client having one. It can be knowledge that you have within the company and being able to utilize it in a, in a more broad sense than a very specific individualized sense.

0:13:49 - (Beau): I could not have said that better myself, Alane, like you probably could have. No, that was perfect. That was so perfect. But that, that's, that is the truth, though. You can train it to do what? Whatever you need to, you know, provide it with whatever information you need. And I'm, you know, I'll give an example here, kind of going down that road that you were going down. When we start thinking about anything that your business does and you start thinking about your pricing, your different services that you may offer. If you are an agency or something like that and you have multitude of services or professional services company, you can train it on that, you can train it on information about that.

0:14:21 - (Beau): You can have it help you write agreements, help you write your proposals. But obviously there are bigger things that you could create with that, with AI agents to where it can build those. But again, we have seen quite a few folks that will say, you know what, I just need help writing out an agreement. And they'll go in and they'll just put in, this is what we typically see, what we typically do. And they just train it on that specific thing.

0:14:43 - (Beau): So everybody knows I go to my agreement GPT, right. So there's still some lines that are blurred there. But with that, it's specialized in one thing. You're putting information in, it's producing an output and it's doing nothing else with it other than that, unless you do have it kind of in that string of automation.

0:14:58 - (Alane): And I'm glad that you said that because that was one of the things that I was thinking about is I think for a company that's going, I just need something that I can do that I, I don't need to hire another company for that I can do myself. This is a great option to start having an assistant there, for lack of a better word. Right. Are there more sophisticated ways to do things now between a RAG system, AI agents? Absolutely. And you kind of touched on that, you know, a little while ago with building this out with a library that an AI agent has access to in rag.

0:15:36 - (Alane): So there are more sophisticated ways to build these that are going to be a better tool for some things. But for you right now, if you're going, man, this sounds awesome, I just want to try it out. This is a great first step.

0:15:49 - (Beau): Yep, absolutely.

0:15:50 - (Alane): All right, perfect. If anybody wants to try out the tool that we've been mentioning, the 25 automation ideas, we're going to link to it in the show notes and reach out if you have questions and you're having fun creating your own GPTs. Thanks for listening to this episode of Automate Your Agency. We hope you're inspired to take your business to the next level. Don't forget to subscribe on your favorite podcast platform and leave us a review.

0:16:12 - (Alane): Your feedback helps us improve and reach more listeners. If you're looking for more resources, visit our website at biggestgoal.ai for free content and tools for automating your business. Join us next week as we dive into more ways to automate and scale your business. Bye for now.

People on this episode

Head Shot

Alane Boyd

Co-host
Head Shot

Micah Johnson

Co-host