How to Survive the AI Bubble as an AI Agency (Do This NOW!)

Right now, the AI industry is sitting on an ice age, and over the next 12 to 24 months, we're going to see a big shift. For some, this is going to be a generational opportunity to build real wealth. And for others, it could mean your financial ruin if things go south. And the difference between these two outcomes is not luck. It's understanding what's really going on behind the headlines. So, long story short, yes, there is a bubble. And really, the only move left on the chessboard for us as AI business owners or agency owners is to figure out what our moves are from here. So, in this video, we're going to be diving deep into what this bubble actually is, what the studies are saying, where the opportunities are, and where the risks are. And at the end of the video, we're going to be breaking down my five-point playbook for how to survive 2026 if this bubble is going to pop, and perhaps maybe when this bubble pops. So, first things first, what actually is this bubble? What does it mean? Long story short, there is more money being spent on AI and invested into it than is being earned from it. And this is because big tech is spending $400 billion per year. That's Microsoft, Facebook, Google, companies like this are pouring money into data center buildouts at an insane rate. And I mean, that's quite normal to invest in the technology. It's like a pretty good play. They've got a lot of cash sitting on hand and they think this is going to be the future. And and they're not wrong. But the point at which this starts getting a little bit sketchy and bubbly is when you start looking at some of the circular spending. Now, you may have seen some of the images going around. If it's not a bubble, then why bubble shaped? And there's like all of the spending going around. And basically this is when companies are kind of passing money around between each other and then being able to note that down as sales or revenue and therefore making it look like their earnings are higher but they haven't actually received the money and the company that's sending them the money may not actually be able to give them that money in future. Stock prices are typically based on a company's earnings. So if earnings sort of beat their expectations then the stock price will go up and we've been seeing companies beating their expectations consistently which is leading the stock prices of these AI stocks to continue going up. And so as an analogy of the circular spending, imagine that you wanted to start a coffee shop. And so I give you $100 to start your coffee shop. And then you take that $100 I gave you and you pay me $100 for the coffee beans to start your company. So on my books, it looks like I've made $100 in bean sales. But has any money actually entered the system? No. And this is basically happening at a massive scale, which is really what's putting most people on edge? Well, the natural next question is why is this so dangerous? And that's because the stock market, if you look at particularly the S&P 500, 75% of the gains of the S&P 500 are made up of these AI and tech stocks. They call it the Magnificent 7 and I think it's even maybe even less now they're considering these these top tech stocks. Anything around these kind of AI infrastructure build out any companies that are are getting into this are seeing their stocks go and sort of tear away from the rest of the market. And you see like the graphs that show the growth of these top companies in the S&P 500 versus the S&P 493, which basically represents the rest of the top American companies. you start to understand that seemingly the whole economy is based off the growth of these stocks and if there are some dodgy accounting going on and people passing around money and checks and at some point maybe a company like OpenAI isn't able to make its purchases of say Nvidia's GPUs and then that revenue that was supposed to be recorded for Nvidia and was justifying its high stock price disappears then you start to have like kind of a cascade of these big tech stocks tumbling in value which because they make up the majority of the value of the S&P 500 and probably the most of people's returns in their retirement accounts and things like this there is the potential for the whole economy to fall down with them. And so the real question is why would these companies fall? And the answer is that they failed to get a return on their investment because at the end of the day all the spend in buying land and setting up power supplies and buying GPUs and rigs and operating the whole system, it has to pay itself off at some point. And that means we need to follow the trail of where the value is actually being created with AI. And there are really two main categories where the value is expected to come from. You either have tools like chat GPT and claude and and things like this that are considered kind of general LLM tools and then you have the business applications of AI being taken using the APIs from these different providers and building it into custom AI systems that companies get a lot of usage out of. And so to understand the risk of this actually being a bubble and whether or not this investment will get paid back, we need to dive into the data and the studies that are reporting the usage of these two different major categories. So first off, the concern here that people have is that we're going to see another dot crash. similar to when internet tech stocks were exploding. The bubble there was actually based around the telecom companies that were building out fiber cables. Like, hey, we're going to need this this fiber. At some point, people are going to need to use highspeed internet in their houses and homes. And so, these companies ended up pouring all of their money into the buildout of these systems. Now, the issue with that is that these telecom companies far outbuilt the demand for this fiber and they have what's called dark fiber. So, there was all of this fiber that wasn't actually being used. And while those cables are definitely being used now, the issue was that in the time that they expected, the usage did not materialize and therefore these companies were overvalued and had to be pulled back down to earth. Things are a little bit different today thankfully where the AI infrastructure that has been built out is being used by literally hundreds of millions of people. The insane daily and weekly active users that you get on something just like chatbt is ridiculous, let alone all of the other tools that are out there. So, it's clear that the value is real with this stuff and people are using it. But going back to those two categories we talked about, the problem is really where that value is being captured and if both sides are really carrying their fair share and the adoption is where we expect it to be. The main concern around this whole bubble thing is really centered around the MIT report that came out a while ago which you guys may have seen some of the numbers floating around but they reported that there's a 95% failure rate for a generative AI pilots within enterprises. The key point there is the enterprises part. So as I said we have these two categories. There's the consumer uses of AI tools like chatbt and so on. And then there is the business usage which again can be split into generic LLM usage tools like chatbt and claude and so on for businesses. And then there's the custom AI development within there as well which is what the study is really talking about. It's talking about custom AI development and true transformation where they're really rethinking a a process or a system with AI and whether or not they're able to get to the point where they are getting a positive ROI from it. And what we're seeing is that these huge companies, these enterprises from 100 million in annual revenue to billions and billions are too big and too slow to really be able to shake things up and get the most out of it right now. And that really is the core of the concerns around AI adoption and this value creation off the back of these buildouts. And so this report got everyone scared and worried for a while. And in this new report came out from Wharton that said 75% of companies are seeing a positive ROI. And so it's pretty confusing when you have one study saying that 95% fail and then another saying that 75% of companies are really happy with AI and they are seeing a positive return on their investments there. And so when you actually dig into these reports and see what both are talking about, it's that split that I talked about before, which is these generic LLM tools being used in this case within a business. It's chatb Claude Blexity, Fireflies, things like this that are kind of off the shelf and plugandplay are increasing employee productivity. Whereas the MIT report is talking about the custom development that is really trying to go in and break things and rearrange it and build a new AI first version of these systems, which of course is going to be a bit harder than getting an ROI with just giving your employees JGBT. And one of the key takeaways when you look at both of these studies is that Wharton data shows that smaller the company, the better the results. Firms in the $50 to $250 million per year range see a 79% positive ROI. And the giant multi-billion dollar companies are three times more likely to get stuck in the pilot phase with these custom development projects than a smaller firm. So basically, these huge companies with thousands of employees and clunky old systems in most cases are struggling to tear things down and replace it with AI systems. Honestly to anyone works in technology or particularly AI should know that that is not very surprising. But more importantly, the smaller the company, the easier it is to get results with these kinds of custom development projects that are truly transformational. And importantly, across the board, regardless of the size of the company, the generic tools like training their teams up on how to use Chad GBT and various other generic LLM tools is having a positive effect. And managers and executives are very, very happy with how things are going. Not only are they happy with the current results that they are seeing, but they are extremely optimistic about AI having a truly transformative effect on their company within the next 3 to 5 years according to the Wharton report. Okay, and so with all of that out of the way, what's the takeaway here? The data kind of proves our thesis of what I defined the AI automation agency model to be in the first place. When you go all the way back to 2023 when I first like named and pushed out this model, I said and defined it as the AI automation agency model is an online business model focused on helping small to medium-sized businesses. I knew even back then that these big enterprises and none of our business, they're going to have their own like internal team. They're going to have these big consultancies coming in and telling them what to do. Our focus has always been on the small to mediumsiz businesses. This data just validates that we've been in the right place all along. the 95% failure rate and the chaos and the difficulties in finding a true ROI is an enterprise problem. It's for the big boys, which I've always told you guys to avoid. And the tangible ROI and the results are going to be much more easily found in that SMB market because they're more agile. They're more nimble. They're willing to sort of tear things down and rebuild them from the ground up, which I think is a great thing that like basically the ball is in the small business's court where they have an opportunity because of their lack of gigantic teams and management structures and old technology. they have a chance to really wipe the slate clean and have a chance of true transformation because they can rebuild their systems from the ground up with the help of agencies like us. And interestingly, the MIT report, which is seemingly so pessimistic about things, literally says that this 95% failure rate in these enterprises creates unprecedented opportunities for vendors who are able to build AI systems that incorporate learning and feedback from their clients and able to optimize over a longer period of time. because that's really the core issue that they found that in order to be truly transformative, these systems need a lot more handholding to get to the point where they break their 5%. And also the report shows that companies are trying to do this stuff themselves internally and they're failing and now they need to reach out and buy rather than building internally. The MIT report shows that partnering with an external vendor like you and I doubles the success rate of the AI project. So basically it's business as usual for us. The the failure of AI in these large companies has nothing really to do with what we are doing and have been doing for a long time. And if anything, the MIT report is showing that the opportunity lies in the vendor's hands right now of being able to create systems that can push through to that 5%. And I'm going to go into that in a little bit. But we are basically going to be needed more than ever because of these difficulties and companies just can't seem to do it themselves. Okay, so now let's get into the fivepoint playbook looking into 2026 about how you can survive the AI bubble if things are going to pop. I mean, when you look at the data, it looks like we have enough consumer usage through chat and these consumer tools to support a lot of the buildout that's happening. There are extremely promising stats coming out of that Wharton report of like 80% weekly usage and these are anywhere from 50 to billions of dollars in revenue. These companies are seeing a 70 or even 80% weekly usage of AI by their employees. And when it comes to custom AI and automation for smaller businesses, the success rates continue to increase the smaller you go. So the only real issue we've got here is enterprise AI and building custom AI agent systems and multi- aent systems. Like of course this stuff is is going to be more difficult to achieve, but when you look at the data, it's like we got green lights on most things right now. And for us as AI agencies, the only red light is one that is nothing really to do with us. So going through these five points here, first thing is avoid the enterprise trap. Keep your focus on what this model has always been about helping small to mediumsiz businesses to understand and implement AI. That's it. At the end of the day, you can look at the SMB market, they're like speedboats, right? They can move very quickly. They're agile. But if you look at the enterprise is like a battleship and it's just like very, very hard to turn and move. And for us, we want to be working with people we can get results with. So if you're just starting out on your journey with one of these businesses, start very small. I've always said that start as small as you can. Then as you progress, start to look into the bin market going 100, 200, 300, 400, 500 employees around there. Number two is to get obsessed with ROI. The hype phase is over now with AI. And you guys know I've talked about the technology adoption life cycle a lot here on this channel. And we're in this phase where we've now crossed the chasm and we're well into the early majority. And the early majority expects results. They expect proof. And if you look at both these reports, they are screaming that companies are now tracking ROI internally. They are very, very serious about not taking anything on unless you can show a clear ROI. So that's the writing on the wall that we're no longer with early adopters. early adopters were going to go, "Hey, this stuff's cool." Like, "Yeah, sure. Here's 10, 20, 30 grand. Let's see what we can do with it." And they'll take a bit of a gamble. But now that we're getting into the rest of the market, the early majority and late majority, we have to come with the tangible ROI. So, you can do this a couple ways. If you are a more of a general development consulting agency like Morningside AI, we're just looking to double down on the ROI calculations done in our consulting process. Another great option is, of course, to look to niche down. Maybe 2026 is the year that you have to niche down. You need data to be able to prove an ROI. Like as I was saying there, it's more of estimates in the consulting process, but the ideal is that you've got tangible data from previous clients. And to do that, you need to serve one niche or solve one problem and you can collect the performance data from that. And then you can walk into these sales calls with undeniable proof of we got these results for certain people. We estimate that you will get this return. If you give us some of your numbers, okay, do a little calculation. Here we go. And that sort of stuff is needed now. And especially if the bubble pops, companies are going to be like, we're not spending enough and unless we see a clear ROI. Point number three is to look to shift from being just a builder to seeing yourself more as an optimizer. Because the MIT report, the 5% of successful projects, again remember in smaller businesses, you probably see 10 15% success rate when you go to custom be for smaller businesses. They found that the successful projects within these 5% of enterprises weren't set and forget. They required constant optimization. So taking in feedback from the team, being able to integrate that back into the system, adjust the prompts in order to move it more towards and iterate towards something that is really, really reliable and consistent as they need it to be. the real value for your clients is going to be unlocked by taking a very important system, kind of breaking it, putting AI into it, and then being there to help them through the process of optimizing over the long term. This also ties into the trend we're seeing with development becoming easier and easier and easier. We're getting better and better dev tools that make our job easier and quicker to set up and stand up these different systems and that's going to be great for us because we can build things and then we focus our time on a extended optimization window where it's like yes, okay, we take some feedback, we're going to test it for a week, try again, split test, get feedback. This process is where I see the value of AI agencies going over the next one, two, three years. Fourth point is to look into becoming, as we are at Morningside AI, more of an AI transformation partner, which means you're not just doing the dev, you're also able to offer education and training for their team on these AI tools, which as Wharton's proved provides some of the easiest ROI and lowest hanging fruit to get a a tangible ROI on their AI investment. So, if you're an AI agency, you can start to lead with education and consulting is kind of a foot in the door offer. You can offer things like an AI literacy audit where you're going to survey their whole team and tell them who's strong and who's weak, who's using it the most, what tools are commonly being used, and give them a bit of a read on where their AI literacy is within their organization. And from there, you can work on training workshops, you can sell them courses, you could push them into consulting, and then into development after that. This is exactly what we've been doing at Morningside AI. And I think when you use the Walton stats and you go to your clients and say, "The easiest ROI is going to be nibbling off some of this training. Let's just figure out where your team is at. It's to get them trained up and then we can move into these later stages of of actual custom development which requires a bit of identification, picking the best ones and then a development and long-term optimization period. You can get some quick wins and foot in the door with these education offers. And finally, in order to survive a bubble, seeing if you can make retainers more of a default part of your offer. If you have these one-off projects with all your clients and then the bubble pops and they go, "Wo, we're going to pull the spending back." Your revenue could also drop to basically zero. If you're able to build essential systems and go through that teething period of training, training, training and say, "Hey, we've really rebuilt our salesunnel here using the system, it's essential and it's a retainer of a few, thousand a month. If it's performing and things start going south, they're not going to be pulling that spin. They'll be looking to other parts like all the other agencies that they're working with. So, your goal is to really build systems that are so critical to your client's daily operations that they can't turn them off. And that's going to ensure that you've got money all through. Even if this is a complete recession or the bubble pops a little bit, deflates slowly, you can have some residual revenue coming in from these retainers. Okay, so we covered a lot there. just to kind of tie it all up nicely and put a bow on it. The bottom line is that the bubble is real. There is potentially some overinflated values going on because of that circular revenue. But when you drill down deep enough and look, okay, is there really an issue with AI itself at the end user level providing the the right kind of results and and ROI to make everything else up top justified? Well, not really. Cuz if you look at CHP and Claude, the consumer usage is crazy on that and it's only going to keep going up. When you look at things like Claude Code and all of these vibe coding tools, there's a ton of value being created through that. when you look at just the use of generic AI tools within companies chat declaw etc there's a huge amount of adoption weekly usage up to 80% tons of usage there so that's providing value as well and the only kind of concerning thing is the enterprise adoption of custom AI systems and as you go to smaller businesses that custom AI development is seeing a positive ROI at much higher rates to me things don't look that bad particularly for us as agencies so if we just stay focused on the small to mediumsiz business market the SMBs that this model has always been about then we're all good the key to surviving and thriving in 2026 and beyond as an agency even if this pops is to look at some of the the writing on the wall here, look at the data and be like, "Okay, let me maybe readjust things a little bit. Let me look to try to get a few more retainers on. Let me start to diversify my income a bit into offering some training because there's obviously a clear ROI for companies there. Let me get a basic consulting offer so that I can better identify the custom development use cases so that I can ensure that my clients are going to get results from that rather than just throwing at the wall and seeing what sticks." I've got a lot of questions about this recently, guys. So, I hope that's been helpful and a clear breakdown of what's going on here, what your opportunities are, and what the risks are. I'm personally super excited for 2026 and I think there's a ton of opportunity just a case of if you can keep your cool and start to look into some of these things I've mentioned in this video. So, if you're by chance a business owner watching this, you can work with my agency, Morning Side, down in the description below. And if you're an AI agency owner, want a bit more source, you can get into my free and paid communities down there as well. If you want a bit of a deeper dive into both of these reports, I've got a full breakdown of that up here. But that's all for the video, guys. Thank you so much for watching and I will see you in the next