2025 Roadmap for Ai Automation beginners

If you're watching this video, you're probably a beginner to AI automation. Maybe you've heard about it, you've seen other people talking about it, you know it's a big thing, but you don't really know where to start and you don't really know what you're doing with it. So in this video, I've put together a roadmap for you guys on how you can get started with AI automation in 2025. As I don't want to sound like a guru online because I'm not, but it's probably going to be the big thing in 2025. In 2024, it's really started to take off after it's all sort of settled down with the big AI hype after ChatGPT. It's kind of started to settle down in 2024 and the attention has been shifted to AI agents. In 2025, the thing that everyone will probably be doing is AI agent development. So if you wanted to learn a little bit more on how you can get involved with it, how you can get started, what the steps you should be taking are, then this is the perfect video for you. Just to give you a little bit of context on who I am, my name's Rory Ridgers and I run an AI development company called Autonomously, and we help people implement AI systems into their businesses. And I've been doing this now for about 6 to 8 months, which means not too long ago I was in the exact same boat as you. Hearing everyone talk about it and not really knowing what to do or where to start. So I'm not going to waste any more time, I'm going to get into the video. There's a bunch of resources in there as well, and you can get all of them for completely free in the description of the video. I'll put a link to it. So let's get into it. So the 2025 roadmap for AI agent beginners, and if you've clicked on this video thinking I'm going to talk about AI automation and now I'm talking about AI agents and you've never heard of them, don't worry, I'll get into what they are in a second. So first of all, I wanted to go over some key terms that I'm going to be referencing in this video just to make sure we're all on the same page and we know what we're talking about. The first one is LLMs, and you've probably all heard of this. Stands for large language models, and these are things like ChatGPT, Claude, Google's Gemini, Llama, things like that. These are all AI models. They're called large language models. Next we have APIs, and these are a set of protocols and tools that allow different software applications to communicate. So basically what that means is APIs just let two softwares connect with each other. So If you wanted it so that ChatGPT could talk to a tool that you've made on another software, you connect ChatGPT and this other software using an API. Prompt engineering— this is basically the prompt that we're going to be giving to AI models or AI agents that tells them how to act, how to use the tools that they've been given, who they are, who we are, everything they might need to know on what they are, how they're doing it, that kind of thing. That's prompt engineering. And then training data/knowledge base This is basically just the data or the knowledge that an AI agent or an AI model has to be able to carry out whatever task it's doing. So this could be a chatbot on a website. It would have a knowledge base of just everything about your business. So this could just be uploaded PDFs or something like that, and it's just data that the AI agent can reference when it's carrying out whatever task it's carrying out. So first of all, what are AI agents? So AI agents, the way I like to describe them, is they're like AI employees. So they can work autonomously, and they can carry out tasks as employees would, but of course they're AI. So just to visualize for you how that might look, I've drawn out the diagram of an AI agent. Basically what they have first is an input or a trigger, and this could be anything. This could mean you type into the AI agent saying, hey, do this, do that, or this could be a file getting put into a certain folder on your Google Drive, or when your stock levels go below a certain number, anything like that. Just a definitive input to trigger the AI agent. And then the AI agent will always have a brain, which will be an LLM. So usually I use ChatGPT. This will be the brain for the AI agents, and then it will have tools equipped with it that allow it to actually carry out the specific task that it's been made to do. So these tools could be something like being able to send emails, or it could be being able to scrape a website, or being able to add events to Google Calendar. Anything like that. These are tools that are equipped to the AI agent that, yeah, just allow it to perform a specific task. And then there is an output. This could be a message back to you, say it scraped a website, then it could output back to you the scraped website, or the output could just be it telling you that it's done whatever task that it was meant to be doing. So to give you sort of a quick run-through of a really, really simple AI agent, the trigger might be you messaging it saying, hey, can you email Dave saying, what's up? That trigger would go along to the brain of the AI agent. The AI agent would look at what you said to it, send an email saying, what's up? And then it will use the send email tool to send an email to this person saying, what's up? And then it will output back to you saying, hey, send the email. Is there anything else I can help you with? That's a really, really simple breakdown of how an AI agent might work. Obviously, they get way more complex than that, but that is sort of the bare bones anatomy of how an AI agent works. So the next thing to ask is simply, where do I begin with creating these AI agents? So what I recommend and what I started doing is simply just watching tutorials on people making these AI agents. You can see I've linked some of my videos here. This was my first ever video. It's a great one for beginners, but we have the Building Your First AI Agent. That's probably the best one I can recommend if you're looking to just get into these AI agents. Adding custom actions to GPTs, and then the AI email assistant agent. I want you guys to just watch these videos. Don't have to be specifically these videos, but I've put them here to give you a good starting point. Watch these videos, make notes of the terminology, the concepts, and then the platforms that they are using as well. And also make note of how they are using AI agents to solve certain problems. So you can sort of— after you watch enough of these videos, you can notice trends in what people are actually using these AI agents for. So make a note of that. I recommend just writing it down and saving it in a file somewhere. But I just want you to get familiar with sort of all the terminology, how AI agents work, how people are building them, the platforms they're building them with, and also why they're building them. So here I've linked some of the documents I made when I was at this stage where I was just watching YouTube videos and seeing what other people were doing and how they were doing it. You can see I've got a tool creation PDF, Prompt and PDF, ChatGPT Coding PDF, APIs PDF, and then an AI Auditing PDF. And these will be available for completely free. I'll link this whole entire presentation in the description and you can just grab them. But these are the documents that I made when I was just watching YouTube videos and making notes of how AI agents work. So it should be really, really useful for you guys because they're coming from someone that's probably in the same position that you are in right now. One of the key things here though is that I don't want you to spend too long doing this because a lot of people— this goes with anything— a lot of people really, really easily get stuck in the trap of just watching YouTube videos thinking that they're actually being productive, you know, actually learning stuff, and just spending all their time watching YouTube videos, reading documents and stuff, and not actually doing anything. So obviously watch these videos to learn the basics and get a good grasp but don't spend too long just watching YouTube videos and doing nothing else. Start to actually take action and build some of these AI agents and automations on your own. So this brings me on to my next bit, which is building your first AI agent or automation. So first of all, you need to choose a platform to build these agents or automations on. I recommend Relevance AI because I think it's the most beginner-friendly and the most easy to get started with, but there are lots of other ones. There's NAN. I also use Make.com a lot. But pick one. By now you've probably been watching one guy consistently on YouTube, so what I did was I just picked the platform that he was using at the time, which I think was Relevance AI, so I just went with that. I don't want you to pick a simple automation or agent to build. I recommend giving this video a watch. This is when I added custom actions to GPTs. I recommend giving that video a watch and trying that out because as you can see here, it teaches you about APIs, which is really, really important, teaches you how to build custom tools, which is also very, very important, and it also teaches you how to use ChatGPT GPT more dynamically. And when I say more dynamically, I mean more than just asking it questions and it giving you answers back. In this video, I'll show you how you're actually able to use ChatGPT to carry out actions that you can't do just normally on ChatGPT. You have to make custom tools and connect them with APIs. So I recommend giving that video a watch, trying it out. Really, really simple to make, but you'll learn a lot of the basic skills from that video. As you can see here, I've said third step is just start building. So just actually start doing it. Don't just keep watching YouTube videos all the time. Actually get in there and start making these automations and these agents, because that is when you'll start to learn the most. So I've also said try to master one or two platforms rather than being an amateur at a lot of platforms, because you don't want to be spread too thin across all these different platforms trying to build all these different things, trying to build voice agents, then website chatbots, then AI agents on Relevance AI, then automations on Make.com, right at the beginning, you're just going to be spread too thin and you're not really going to be good at any of them. You want to try and get good at one or two of them. I chose Make.com and Relevance AI because they work really, really well together. I use Relevance AI to make my AI agents, and then I also have to connect it to Make.com quite a lot for sort of simple automations I use within the AI agents. So you don't have to use these platforms, these are just the ones I use. There are a lot of great platforms out there, but what I'm trying to say is Don't spread yourself too thin. Don't try and learn all the different platforms. So just pick one, learn it, master it, and you will be absolutely fine. So as I said, there is no best platform. All these platforms have their own strengths and weaknesses, and ultimately the best platform will be the one that you work the best on. Because for instance, Relevance AI and NAN are both big AI agent development platforms, but you could build the same AI agent on Relevance AI that you could on NAN. But someone who's used Relevance AI for ages will obviously be able to build it better than on NAN, and then someone that's used NAN for ages will obviously to build the same agent better on NAN than on Relevance AI, you know. So just stick to the one that you like the most and you know the most, and you can build pretty much any agent on any platform. So now I want to get to niche, and this is going to be mainly for the people that are actually looking to sell these AI solutions and these AI agents. You would have heard niche being an important thing, sort of being thrown around in the industry. So there are two ways you can go with this. You can either go broad or you can go niche down when you're looking to sell these AI solutions. So I've given a sort of a little breakdown of both these sides, the broad and the niche. The broad, the strengths mainly are it's easier for those with less experience, there is a larger addressable market because you're just looking at everyone rather than a sort of niche market. And I've said there's also more opportunity to learn at the start because there is a wider range of projects. So I know this from going broad in my agency, that you'll be doing a really, really wide range of projects at the start because everything will be new and you'll be working with people from all different industries. So at the start, every project will be different, which means you're going to learn way more than just sort of doing the same thing over and over again. The weaknesses obviously for this are that you can't charge as much and you'll probably close less as you are less specialized. And I know you guys are going to hear this and just be put off broad instantly, but when I say that you're going to close less, this isn't necessarily a bad thing because you have a bigger addressable market. You're going to be talking to more people. So even if you have a lower conversion rate when it comes to closing, you still might be able to get more clients than if you go niched. And hearing the fact that you can't charge as much might also not be great for some of you guys, but I promise you, at the beginning, you just really want to learn how all of this works, how to create AI agents, and how to actually solve problems. So don't worry too much. You can still obviously charge a good amount of money, but when you're niched down, you can specialize a little bit more and you can charge more for what you do. I've also said a downside for the broad approach is that you often won't be able to reuse your agents/automations. You could also see this as a strength in that every time you'll have to make a new agent or an automation, so you'll probably learn way more. But with niched, a lot of the time you can make an agent or automation and you'll be able to use it more than once, because in your niche down market, a lot of the time people are going to be looking for the same solutions. So yeah, that's something you can't do in broad. If we look at niche, obviously the strengths are kind of just the opposite of the weaknesses in broad. You can charge more as you're more specialized, you'll have higher conversion rates, and you can likely reuse your agents and your automations. But then the weaknesses here obviously is that there's going to be a smaller addressable market, and it probably requires experience in a specific industry. So obviously you can look at this, take of it what you will. You might already know straight away, yes, I want to do broad, or yes, I want to do niche. But if you aren't too sure, the way I look at it is go broad unless you have a good amount of experience in a specific industry, and I've put at least 5 to 10 years here. But obviously take the 5 to 10 years with a pinch of salt. That isn't a one-size-fits-all. The main thing here is that if you don't really have any experience in a specific industry, then better just go broad and try and go for everyone. But if you have, say, you've worked in finance for 10 years, or you've been a lawyer for 15 years, or you've been high up in a specific company for a long time, it may be better to go specific. And the reason for that is because you will likely know a lot of the pain points of these industries. When you start to learn how to use AI and AI agents, you'll quickly learn how you can solve these problems with AI agents. So that's why I recommend if you've got experience, it might be better to go niched. But if you haven't, like I didn't, it's probably better to just go broad. So believe it or not, the whole process of learning the basics, choosing whether to go broad or niche, and building your first AI agent slash automation should only really take you about 1 to 2 weeks. And this goes back to what I said before, about people getting stuck in the trap of just learning and learning and thinking they're doing something and not actually doing anything. This whole process should really only take you 1 to 2 weeks, maybe 1 week to just learn the basics, spending maybe a few hours a day, even just an hour a day watching YouTube videos. After a week, you should have the basics down. Then you can start to decide whether you're going to go niche or broad. And then the most important thing, you can actually start to build your first AI agent or your AI automation. So Obviously, again, this isn't a one-size-fits-all with this, but for most people, this should really only take 1 to 2 weeks. Don't get caught up too much in all of this stuff. The main thing is just learning the basics and then just actually starting and actually making these AI agents and automations. And then once you've done all of this, all you have to do now is just sort of start to make more and more complex agents. That's what I did when I first made my very first AI automation. I did is I sort of looked at it played around with it, showed a couple people, said, yeah, that's cool, and then started just making even more complex agents and automations. And I promise you guys, that is the best way you're going to learn, is just by sort of pushing yourself further and further. But ultimately, you really have nothing to lose with this. If you make an automation and then you try and make a more complex version of it, or just completely different thing but more complex, and you fail, you don't lose anything. So just make these agents and these automations, show a couple people if you like, try it out, and then move on to the next and just try and get more complex and try and do different things with these AI agents. So here I've sort of put out a little checklist or a leveling system on what you can expect when you're trying to make more and more complex tools and agents and automations. So the first level, like I said previously, adding custom actions to GPTs is where I recommend everyone starts off because it teaches you a lot of the basics and it's relatively simple to do. The second level is going to be creating your first AI agent with at least one tool. Third is gonna be creating a tool or automation that uses an API. So I do this a lot on Relevance AI. I make a tool and then I usually connect it to a scenario on make.com via an API. And it's really, really useful. You're probably gonna have to do it all the time when you're making these AI agents and AI automation. So that's a really, really valuable skill to learn. Level 4, I said find a real world problem and try and solve it with an AI agent. And now this doesn't actually mean getting your first client, but this just means looking at a real-world problem that people have in business. It could be something really simple like your inbox being too messy or something like that. It doesn't have to be industry-specific. And then just try and create an AI agent that can solve this. And like I said, this isn't necessarily for a client. You don't have to try and sell this or anything, but just try and make it and try and solve this problem. And that is a really, really valuable skill. And then for level 5, I've just said create your first multi-agent system. And basically what this is is an AI agent, but instead of having tools attached to the AI agent, you have other agents attached to that agent that can call upon to perform specific tasks. I also have a video on that when you get to it, but this might take you a few weeks to get here. So when you come around to it, check out my channel. I have a video on making multi-agent systems. And then one of the most important parts of this, how do you actually get clients? So how do you actually sell these AI agents and these AI automations to people? So there's two ways. This isn't just for AI development agencies. This is just for all business pretty much. The main way you're going to get clients is inbound leads or cold outreach. And if you haven't heard of those, inbound leads is when people come to you and you get this through usually a personal brand. So this could be posting videos on YouTube, posting things on LinkedIn or X, or creating content in communities is also a really great way. And that is when people come to you and say, hey, I want this done, or hey, I want to implement AI into my business. That's the way I get my clients. The other way is cold outreach, and this means trying to get clients by actually going directly to people and trying to get them on board that way. So this could look something like email outreach, reaching out to friends or family, or cold calling. I'm sure you've probably heard of all of these, so you know what we're talking about. So put together a little strengths and weaknesses of both of these and which one I prefer. For inbound leads, I said the strengths are going to be higher conversion rates because If these people are coming to you, they already sort of know that you know what you're talking about and that you can do the thing that they're asking you to do. So the conversion rates are usually a lot higher with this. The leads are also going to be more qualified because they've taken the time to come to you. So you're gonna have to spend less time nurturing them and convincing them that you know what you're talking about and you know what you're doing. And they can be cheaper or free to get because, I mean, it doesn't really cost anything to make X posts or LinkedIn posts or create YouTube videos. That doesn't really cost anything. So a lot of the time they're cheaper to get. And one of the main things with creating content rather than reaching out to people in order to get clients is that you can actually learn while you're doing it. That's why I decided to start my YouTube channel, because I knew that I would actually be able to learn how to make these AI agents and automations while marketing, which is really, really cool because it's kind of been proven, I think, that teaching is one of the best ways to learn. So that's why I started my YouTube channel, because I thought if I can teach other people how to make these AI agents and automations, I'll also be able to learn myself. The only real weakness here I could think of is it involves getting in front of a camera, which I know for a lot of people is a really big thing, getting in front of a camera and making content. And for me, it was a massive thing as well. But to be honest, once you do it, you'll realize that you should have been doing it earlier. I did anyway. And that is really the only thing stopping you from making content and getting clients this way, inbound leads. So if you can get over that fear of sitting in front of a camera and posting content on the internet, this is by far the best way of getting leads. The second one is cold outreach, which also definitely has its place. The main strengths here being that you can target a specific audience more accurately because you can target, say, just people from a specific industry or just CEOs of a company, that kind of thing. You can target them through cold calling or email outreach. Also, a lot of cold outreach can be automated with AI, especially email outreach. All of that can pretty much be automated with AI, so you can do it all on autopilot. And also you can get in front of a higher volume of people a lot of the time. Weaknesses here is that you're going to have a lower conversion rate, which isn't necessarily a bad thing because you're going to be reaching out to way more people, but your conversion rate will be definitely lower. And it can start to get expensive. Obviously, expensive— wrong here, don't know what happened there. Depending on how many emails you are sending. So there's some people that are sending hundreds of thousands or even millions of emails a month, and as you can expect, then this can get pretty costly if you're not signing clients off the back of it. So do be aware of that. And also a big thing is that you're gonna have to spend more time nurturing these leads because you've outreached to them, they're not too sure on you, maybe they don't know who you are, they don't know what you can do, so you're gonna have to spend more time nurturing them. So my opinion on this is create content and get inbound leads, as you'll be probably unexperienced in sales and unexperienced in AI agent development. And the reason I've mentioned unexperienced in sales is because, as I said, the conversion rate will be much higher for inbound leads and it'll probably be much lower for cold outreach. So if you're doing cold outreach, then you might need to have some pretty good sales skills to actually get these conversion rates up. So that's why I recommend inbound leads. If they've come to you, you don't really need to be the best salesman in the world. You just need to know what you're talking about, which you obviously already do because you've been making content on it. So that's why I've mentioned that part. And also I said if you're unexperienced in AI agent development, as I said, you can learn on the job, which is really, really cool. You can learn while you're actually marketing and getting these leads in. And also I said that personal brands, as a lot of you may know, they have high leverage for future endeavors. So if this isn't something you want to do forever maybe you want to launch a software or sell a course, whatever it is you want to do. Personal brands, so creating content on YouTube or LinkedIn or X, they have much higher leverage, so you can come and sell other things later down the line. So that's why I recommend just creating content and getting inbound leads that way. So now you should have your lead generation method, you should have your development skills, you should have a grasp on the basics, you should also have a target audience. And guys, once you have all this This is pretty much it. There is not really anything else that needs to be done apart from you actually just taking action and pushing yourself and trying to do more complex AI agent builds and AI automations and just getting yourself out there and trying to actually get clients. So I hope you guys enjoyed this video and you should now all be set up to create your first AI agents in 2025 and start to sell them to other people. As I said at the start of this video, this whole presentation will be available for completely free in the description. You can go ahead and grab it and grab all the resources that I put in there, the YouTube videos and the PDF files. You should see a link for it in the description, so you can go ahead and grab all of that. So if any of you guys have any questions, you weren't really too sure what I was talking about in this video, just shoot me over an email and I'll be happy to answer any questions that you have about AI automation, AI development agencies, or anything like that. And also, if you guys have any video suggestions that you'd like to see for 2025, then please do just let me know in the comments and I'll do my best to get around to doing them.