Ai Agents are dead, sell this instead (crazy opportunity)

What's up guys? In this video, we're gonna be going over the difference between building simple AI agent systems that you're probably all used to building and how that differs from building actual real AI infrastructure in businesses. And if you don't know what I mean by AI infrastructure, don't worry, we're gonna get into all of that in the video. I've put together a presentation to sort of walk you through each stage of this. Specifically, we're gonna be going over the simple AI agent systems that we all know that are sort of the one-off, one-use-case AI agent systems or just AI automations. That, like I said, we all know and we all know how to build, and sort of the cons of them that you probably don't even really notice when you're building them. And if you haven't actually started selling these to companies yet, some of the things that are going to hold you back from actually scaling your agency with systems like this. There's 4 main things that are problems with these systems. We're going to go into detail on each of them and what makes it a problem and how we can get around this with building AI infrastructure rather than just these simple one-off agents. Also going to get into what AI infrastructure actually is and the difference between that and these one-off agent systems in a little bit more detail. Then we're gonna go over how to identify problems in businesses and how to actually build these AI sort of infrastructure systems. I'm also gonna go over a demo example that I've put together where in one case a company comes to someone and one person builds a one-off agent system, and we're gonna have a look at how that impacts that business. And then we're also gonna look at another person that builds a proper AI infrastructure around their problem and implements that into the business, and we're going to see the effects that that had on the business. And then we're lastly going to look at why, as an agency owner, you should actually care about this and why you should be doing this, and also the next steps for you and how you should go about implementing these changes I'm going to be showing you into your current offer and the way you run your agency. Okay, so firstly, we're going to be looking into these AI agents and tools that, like I said, we all know and we've all been building for the past however long you've been doing this for. So AI agents and tools, most of the time they're going to be single use case, they're going to be plug-and-play systems, and they're also not going to need adjusting. What I mean by not need adjusting, by the way, is sort of similar to the plug-and-play system. Um, once it's done, you can kind of just implement it, don't really have to worry too much about it. And a really, really simple example of this is a system that I made recently, which simply just generates leads and adds them to a Google Sheet with all their contact information. So you can see here, we trigger it off by just clicking this, and it sends off a request to Apify, which then scrapes Apollo for, I think it was, uh, real estate companies in this case. And then it gets back a bunch of real estate companies with all their contact information. Then we just have a simple filter here, and the filter was for an employee count of 500 or below. So all the ones that have an employee count of 500 or below go through. We just clean up the company name, and then we scrape their website, pass it over to an LLM that's just going to write sort of a personalized compliment that we can use for each of these companies when we're writing outreach messages. And then it just adds all of that to a Google Sheet. So a nice little system, does the job, generates us some leads, um, but obviously it's a little bit too simple. So with a system like this and other similar systems, there are going to be 4 main issues that you're going to run into. So the first one, it's easy to replicate. The second one, it has a short shelf life. Third one, it's difficult to sell for a substantial amount of money. And then also it rarely provides real value. And we're going to jump into each of these and I'll show you what I mean by each of them in a little bit more detail. So easy to replicate. What I mean by this is that system I just showed you is an AI lead generation system. And something like that I could potentially sell for maybe $5,000 to a company that's looking for a system like that. But the issue with this is basically a person could just come along— might be one of you watching this video— person could just come along, watch a YouTube video on how to build a similar system or even the exact same system, and then in sort of 2 to 3 days, even if they're a beginner, they could pretty much create the exact same system that I've just showed you there. So this actually happened to me recently. I had a person get in touch with me. They're actually looking to see if I did any coaching. I just jumped on a call with them and was talking for a little while. and they basically told me that they just started out with the AI agency sort of model. Um, they'd only been doing it like a couple weeks or something. It might even been just a week, and they'd watched a YouTube video on how to create an AI lead generation system similar to one I just showed you. And like I said, they had like a week, maybe 2 weeks of experience. They just built the exact same system and then they started using it, and they said they generated about 15 to 20 booked calls in like that first week or for the next week, which is absolutely crazy and obviously means that system works really, really well. And this guy was literally able to build it in like a couple days with a couple of weeks of experience, which is crazy, which basically means that that same system that I was able to sell for $5,000, if that guy came along, he would just be able to sell it for, let's say, $1,000, because it's so easy for him to build. And there's going to be loads of people building the same thing. It's really, really easy to replicate. So we just see sort of a race to the bottom with the price. People can just come create the exact same system that you've just built because it's just a one-use case system, and they can just sell it. For example, from $5,000, they can just come in and sell it for $1,000. That's one of the main issues with a system like this is a short shelf life. It's a little bit of a tongue twister. Um, so this one is pretty similar to that one I just showed you, which it being easy to replicate. If we look at sort of a timeline here of these one-off use case systems like the one I just showed you, um, the timeline looks a little bit like this. So you could be selling your system for, like I said, about $5,000, and then someone could come in and start selling it for $2,000 because they'd be able to recreate that exact same system and it's pretty easy for them. So they see you're selling it for $5,000, they come in at $2,000, they're selling the exact same system So people are going to come to them and then people are going to see that they're selling that for $2,000 and they're just going to start selling it for $500 because they're again able to create the exact same system. So it's going to sell it for even cheaper. And then someone might come along and create a SaaS that can do the exact same thing and cost these people $20 a month. There, I mean, there's already loads out there like lead generation SaaS that uses, uh, AI to sort of enrich the lead and get the specific contact information, things like that. People can come along, make a SaaS that does that, charge $20 a month. And then this one doesn't always happen, but I just want to throw it in there. A tech giant could come in and just make a similar system for completely free. Let's say, um, OpenAI release an update where you can connect ChatGPT to Apollo and you can just ask it to scrape some leads in whatever industry you like. That's pretty unlikely for this one, to be fair, but you do see that happening with companies like OpenAI, Meta, Google. They just released one update to their product and it just completely wipes out everything everyone's done before where they're actually charging like a lot of money and it sort of dwindles down. Even when it goes to a SaaS, doesn't— that doesn't cost too much a month. Uh, you see these tech giants, they just release a feature for free that just completely kills it. So the third thing is this is difficult to sell for a decent amount of money. And the main reasons for that is because it's not tailored. If any of you have sold anything to a business before, it doesn't have to be an AI agency offer. If you've sold anything, you've done any B2B work, You know that of course companies are less likely to pay for something that isn't as tailored to them. Everyone knows that, that's obvious. The next one is there's not much real value. And because what I found with these sort of one-off systems, I've built loads of them myself, especially at the beginning, is they're kind of just more cool than anything. A lot of the time they're just kind of cool. And don't get me wrong, they are cool, but companies aren't going to pay a lot of money for things that just aren't cool. You can't really build a proper business of just selling sort of cool stuff. For maybe $1,000 or $500. It doesn't really work long term, and it kind of makes it really difficult to sell. You need to, of course, be providing real value and not just selling cool stuff. And then third reason is there's too much competition. So what I showed you before about just being so easy for people to replicate, it's just way too much competition. And likely, if people are selling it for less, companies are just going to go to them rather than you. Then the last thing is rarely provide real value. And of course, I put that in here, but I want to go a little bit more detail into why these things usually don't provide a lot of real value. They can provide a little bit, of course, But when you talk about real value for a company, the systems that these companies are gonna pay $10,000+ for, you can't really get that with these one-off use case systems. So little equation here. Everyone knows providing real value basically just equates to solving real problems, cuz that's what you're doing when you're working, uh, with other companies doing, running like a B2B business. Basically you are just solving problems and that's how you provide real value by solving real problems. And of course, to solve real problems, you need to find the issues first and then come up with a solution. Really, really simple stuff. To solve real problems, you first need to find those problems, come up with a solution, and then if you're doing that, you're providing real value because you're actually solving these real problems. But of course, if this is just a one-off system that you've made once, maybe it worked for you, maybe it worked for one company that you sold it to, and now you're just trying to resell it to a bunch of different companies, you're not actually talking to these companies properly, which means you're not properly finding the issues. There's always crux to all these issues and crux to these companies. Um, so you're not actually sort of doing a deep dive and finding the proper issues in these companies. So if you're not finding the issues, of course you can't come up with a specific solution for them. And if you're not finding the issues and coming up with a solution, you're not solving real problems. And if you're not solving real problems, you're not really providing real value. So now on to AI infrastructure. What actually is that? What does that even mean? So infrastructure, just as a term, basically means the fundamental and physical organizational structures, systems, and resources needed for its operations and growth. And when it says its that refers to like a company or a business. So basically it's sort of the fundamental things that a company needs to operate and grow with. So the systems, the structures, the resources, that kind of stuff. That's what we mean when we say infrastructure. And then AI infrastructure basically means to either embed AI systems into a company's existing infrastructure or build new infrastructure using AI for a company. And that still might not mean too much to you, so we're going to get into an example Um, so that lead generation system I showed you before, like I said, it's just like a one-off system that you could like potentially resell a bunch of times for a little bit of money. But if you're looking at actual AI infrastructure that does a similar kind of thing, it might look something like this. So in this case, we've still got the lead generation system, but this time it can be more customized. So it can be doing sort of, uh, specific leads that are more custom to the actual business that we're working with. And we can also have some systems to reduce no-shows. Uh, we can have a custom dashboard set up for this business. We can have a custom follow-up system set up, an inbox manager agent, a RAG system for the sales team. You can see we're sort of layering on a bunch of different things to cover all the things, because obviously with a lead generation system, the main thing here is generating more leads and generating more revenue. That's the main issue that you're solving with this, and all of these things are going to contribute to that. It's sort of a more fleshed-out way, and there's going to be way less errors with this. And this is what it means to actually build real infrastructure in a business and solve real issues rather than just send these one-off solutions. So I made a little visualization so you could sort of visualize this. So to begin with, we have a cup and it's got a crack in it, and this is going to be an issue in a business. So it could be a lead generation issue. Company comes to you, they've got a lead generation issue. This is the crack in their cup of water or their glass of water, sorry. And then the next step, you're going to put a plaster on it, just bandage it up with that lead generation system that I just showed you. Um, to just sort of get them some more leads, solve the issue on the surface. Then you can see next down here, more cracks are starting to form after you've sort of bandaged it up. And this is going to be things like no-shows are high, unqualified leads, bad follow-up. So technically, yes, you've solved the issue, they've got more leads through. But when a company says they want more leads, of course that just means they want to increase revenue, all right? So they want more leads. They also want to close those more leads. More leads that you don't close means absolutely nothing. So you've given them the lead generation system, and then that's we've done, generating more leads now. But you can see these more cracks are starting to show. So bad follow-up, unqualified leads, no-shows are high. So likely their sales aren't even going up, their revenue isn't even going up. So this system was pretty much useless. Then if we look at the AI infrastructure approach with the same thing, company comes to you, they have a crack in their glass, they've got a lead generation issue. Rather than sort of trying to bandage it up and then more cracks appear, and then maybe trying to do more one-off solutions to sort fill the gaps up there. Rather than doing it that way, with the AI infrastructure approach, you're basically just making a new glass. So just making a better glass, putting the company into that, using that glass instead, rather than having this one that's got little cracks in and more cracks are going to form. Just make a new glass that's stronger and it's better, and there's actual real infrastructure there, and there's way less room for error, and there's not going to be all these cracks that come through because you've actually talked to these companies and found the real issues in their company. Next, we're going to move on to how do you actually identify these infrastructure opportunities, because now you know what they are, how do you actually identify what you need to build and how you need to structure these infrastructures? So the main thing here is just a business audit. Whether you're watching this and you're looking to sell this to other companies, or you were looking to build this in your own company, first thing you need to do is a business audit. So the first question you have to ask is what are the current issues in either your business or your client's business, your potential client's business, whatever it might be. What are the current issues? So from some high-level things, um, this could be like lead gen, client retention, customer support, scaling issues, management issues. So I'm sure most of you know if you run a business or if you talk to business owners, pretty much every business has one of these problems or a similar problem like this, either lead gen, client retention, customer support, scaling issues, management issues. So first step of the business audit is just identifying what the actual sort of general issues are in the business. And then the next step is why are these happening? They all happen for a reason. So why are these happening? I've put some examples down. Lead gen could be because of a lack of budget in the ads, for example. Client retention could be maybe poor results. Customer support could be poor usability of the product. Scaling issues could be a task management problem. And then management issues could be a lack of communication. So once you have what the issues are, why these issues are happening, then you have to move on to what actually needs to change to solve these issues. So again, I'll put some examples down. A lead gen issue caused by a lack of budget in ads— a solution for this is going to be smarter ad spend. Client retention because of poor results— obviously a better service. Customer support because of poor usability— maybe your app or your product needs to be more intuitive. And then clear ownership and accountability for the scaling issues. And then better systems for the management issues. So once you have all of this, current issues, why they're happening, and what needs to change. Now you move on to the actual solution. So you've identified the issue, identified the reason, identified what needs to change. Now you just need to build a system or a set of systems to fix that. And I appreciate that that is, um, easier said than done. A lot of you guys watching this might not be able to just snap your fingers and build a system that can solve these issues. I completely understand it. But from a high level, this is essentially what you're doing: identifying the issue, identifying the reason, what needs to change, and then building a system or a set of systems to fix that. So I'm going to give you an example to help you guys out to sort of see what this might actually look like. We've got a fake company here called Harper and Lane Interiors. The industry is interior design and project management. Team size, there's 18 of them. Pain point, so they've come to us with this pain point. They have this issue, uh, we're drowning in Slack messages and missed tasks. Things fall through the cracks constantly. Can you build us an AI assistant that reminds people of deadlines? So Pretty simple. They might not even say to you the 'can you build us this?' They might— majority of the time, actually, they're just going to come to you and say, 'I've got this issue, can you fix it?' Because that's sort of your job, is to figure out, um, how to actually fix this. You need to figure out what needs to change and how you're going to be able to do that. So if you're just doing like a one-off system, like I showed you, like that lead generation system, if you came at this with just a one-off system approach, it might look something like a Slack bot that can read from the Google Calendar, send daily reminders and nudges to stand task messages like, can you sort this by Friday, for example. Um, the pros of this is it's going to help people get reminders, it's also going to reduce short-term drop-off. So some nice little pros there. But of course, the cons of this is tasks will still get lost in Slack threads, no one knows who's actually responsible for anything, deadlines are missed because estimates are vague, designers overpromise because there's no visibility into real project load, contractors aren't looped in clearly enough. So basically there's just sort of a lack of layers here. This is just a one-off system, doesn't really solve all the problems. On a surface level, it solves the problem. It's built them an AI assistant that can keep the reminders, look at the Slack messages, and sort of remind people of what's going on. So on the surface, this works. But when you actually implement this and you build this for these people, and, um, you sort of— they use it for a little while, all of these cracks are going to show through. Like I showed you with the glasses, you've put the Band-Aid on, you sort of solved the issue, but then all these cracks are going to start to form where the tasks are still getting lost in Slack threads, uh, deadlines are missed still because the estimates are vague, yada yada. So if you look at it from an AI infrastructure approach, this is how it might look. You might have the first layer, which could be the data layer, which all Slack, Notion, and email threads are monitored and logged. Tasks are passed, categorized, and assigned metadata. So owner, urgency, project deadline, that kind of thing. The timeline data pulled from project management software and CRM. So that's going to be sort of the data layer. Getting all of the data into one place from every single possible point that the data could be posted or updated. And then we have an orchestration layer. So when a task is detected, it's pushed into a central task database, for example, like Airtable, Supabase, or a custom backend, whatever backend they currently use, or just sort of the best option for their use case. Uh, owners are tagged, estimated delivery time is predicted based on current workload. If a task goes overdue, escalation flows are triggered. For example, an alert message or reschedule downstream tasks. So now we have, so now we have the orchestration layer working in conjunction with that data layer. The third thing could be an intelligence layer. So a lightweight LLM layer interprets vague tasks like, can you sort this soon? The task could be sort fabric samples. The deadline could be 3 days. So it's gonna go through, look at everything that needs to be done, assign it a task, and then assign it a deadline. And then it's going to prioritize team tasks across load and urgency. Even reorders timelines automatically. It flags ambiguous messages for clarification. So now the data layer is sort of getting all that data, putting it in one place. The orchestration layer is sort of manipulating all of that data so we can actually use it. It's tagging it, it's adding estimated delivery, everything like that. The intelligence layer is actually doing stuff with that information, so it's sort of, um, sending reminders, prioritizing tasks, that kind of thing. And then the fourth, we could have an interface layer that might look like a weekly dashboard sent via Slack, an email, individual task inboxes, voice-based task logging, like, hey, log a new client to install Tasker Friday, uh, or private message the agent saying what's overdue and blocking progress on the Clark residence. So kind of just like the point of contact is really nice with these sort of infrastructure approaches to have custom dashboards and custom interfaces to be able to talk with the agent or check on analytics, that kind of stuff. So you can see from here, this is sort of building actual infrastructure in the business rather than just sort of making a one-off use case system that just sort of like I said, slaps a Band-Aid on, looks like it's actually fixing the problem, but when they come to use it a little bit, they realize that it didn't really fix the problem. They've just wasted their money. When you do it this way, you're building real infrastructure, and this can sort of grow, they can get better as it goes along, and hopefully no cracks are going to appear and nothing's going to seep through the gaps. So now you've got to the end of the video, you're wondering, why should I actually care about this as an agency owner? Why should you pivot to this sort of infrastructure approach rather than these one-off systems? Because of course I've gone through the benefits for the clients, which is absolutely amazing. And when you run an agency, of course you have to have the client in mind and you have to be trying to deliver the best service possible at all times for the client. But why should you actually care? What are the benefits of doing that when you're running an agency like this? So main benefits are going to be higher retainers because the infrastructure is going to need more support and it's going to need more maintenance. So you might need to be a little bit more involved in there, which means you can actually charge, uh, higher monthly retainers or quarterly, yearly retainers, however you do it. Um, you can have long-term contracts versus one-and-done builds. So likely once you've built this initial infrastructure for these clients, they might say, hey, I actually wanted to do this as well, I want something that can do this, I want something that can do that. And you're going to be able to build on that infrastructure. So it's going to be sort of a more long-term partnership kind of thing rather than just a one-and-done build like it says. So you just build the system, give it to them, never hear from them again. Easier upsells. It's kind of like the long-term contract things I said. These people might come back to you and say, hey, I want this, but you might also be able to go to them and say, hey, I think we can improve it by doing this. So because you've already built this initial infrastructure— infrastructure is made to be built on— so you can just build on this infrastructure that you've already built and get some nice little upsells in there. And then like I said as well, you actually become a strategic partner, not just a tool vendor. So you're not just making this tool for them, giving it to them, never hear from them again. You're actually becoming a partner with this person, which means you can grow with them you get more testimonials, you get more projects done, more experience, everything like that. Partnerships are just the best way to go with this, actually working as a partner with these companies rather than just sort of selling them solution, never hearing from them again. So the next steps for you, if you're watching this video, I've put together a little timeline. If you're just starting out with this, this is maybe one of the first videos you're watching, or maybe you've been in this for maybe a week or two, you're just starting out with this, I would actually recommend selling those one-off AI agent systems. Just purely to get a little bit of capital and to get experience. That's what I did at the beginning, just started selling sort of simple chatbots, that kind of thing. And like I said, gets you a nice little bit of capital and some experience building these AI agent systems and also working with clients, running a business, that kind of stuff. And now you're 1 to 2 months in, this is when you should start to look at the infrastructure-like system. So it's not just like there's a one-off system and there's infrastructure and then two completely different things. It's like night and day. Some of them are going to be more infrastructure-like, some of them are going to be more one-off system-like. It's not like black and white, like I said, it's not night and day. So you're going to want to start looking at the sort of more infrastructure type. There's more layers to it, that kind of thing. 3 to 5 months in, you should be focusing on pretty much just building infrastructure and actually being strategic partners with these people, building on the infrastructure, creating lots of layers to the systems that you're building rather than just one-off. You shouldn't really be thinking about these one-off projects too much anymore. And there's 6 months plus, so this is where you're really going to refine your approach and potentially niche down. You don't have to to niche down at this point. I haven't niched down. I've been running my agency for about 10 to 11 months now, and I haven't niched down quite yet. So you don't have to. Some do at this point, some don't. It's kind of up to you, and it depends where you're at at this point. So that is it for the video, guys. I hope you learned something, and I hope now you can kind of change your approach when it comes to your AI agency offer rather than these one-off agent systems that, like I said, we all know, we all know how to build, and we've probably all been building up until this point or up until recently. And kind of switch it for more infrastructure approach and building what's called AI infrastructure in these businesses, which basically just means building different layers of these AI agent systems and creating a whole new approach rather than just giving them a one-off system and then never speaking to them again and thinking that's going to solve all the issues. So that's it for the video. If there's any video suggestions you'd like to see in the future, please do just let me know. I have way more videos coming out very soon, so if you've got any video suggestions, just drop them in the comments or just shoot me over an email and I'll try my best to get around to doing them.