Your AI Projects Won't Work (Unless You're Doing This)

Not too long ago, we had MIT come out and say that 95% of AI pilots within companies were failing to deliver a positive ROI. And that was obviously a bit of a shock to the AI space and the industry as a whole and started the whole like bubble talk, which we're kind of deep in the midst of right now. But on the other side of that, very recently, we've had a Wharton study that says 75% of companies are seeing a positive ROI. And overall, there's a whole lot more positive and optimistic about how AI is actually helping companies to see some kind of positive ROI on their investments in the technology. for my agency, Morningside AI, I thought I needed to dig into both of these and figure out what was going on cuz we can't have two contradicting studies, right? So, I've done you guys the favor of digging in and reading these entire reports and figuring out who's telling the truth or perhaps there's parts of truth in each report. And that's what we're going to be breaking down in this video. And if you're an AI business owner like myself, I'm going to be sharing at the end my strategies and tips based off these reports as to how you can look into 2026 and make sure that you're making the right bets and really using the data to direct your strategy next year. Right?

So the main reason for the contradiction is that these two reports are actually talking about completely different things. So the MIT report is talking about success on custom development. The Wharton report on the other side is talking about general AI within companies. So that can include just using chatbt or co-pilot or claude or any of the kind of generic LLM based tools. Might be sort of AI creative tools here and there. But a lot of the time it's going to be offtheshelf or kind of generic tools. So that immediately kind of reveals like why there's such a big gap here. Because if we're talking about custom development, I can say firsthand from morning. AI AI that the development portion and particularly custom development is going to be much harder than just teaching the people how to use JBT. So first thing you need to understand is that the difference in these numbers is primarily because they're talking about different things. custom development and the success of actual transformation. The MIT report really talks about yeah there's a lot of like money being thrown around to these different initiatives but how much is actually resulting in transformation like these general purpose technologies like the internet and things like this where it opened up doors for completely new business models and like new startups rapidly gaining market share and like sort of flipping things on its head. If you think about the internet, you had e-commerce and people taking like a WBY Parker and their glasses model, taking something like retail for eyewear and then saying, "Hey, what if I rethought the whole business from the ground up to be e-commerce and internet- based? " That's the kind of thing that they are looking for. And if we're honest, we're not really seeing these AI first startups really taking over and taking massive chunks of market share at the moment. While I believe that stuff is going to be on the way, I think as of right now, as the MIT report has found, the actual difficulty of getting custom development and truly transformative AI plugged into companies is still very difficult. And they pinned that down to a few things primarily around the fact that these systems are inflexible and they don't really learn. And we've encountered this at Morningside where a lot of the time the development is only a small portion of it. You can see this in say things like voice agents and other where the development is actually only maybe a four week or 6 week process but the testing and optimization and taking feedback from the client and integrating that into it to really make it get to the point that it's performing how they want to. That is actually a whole lot more difficult and that at the end of the day is what MIT is talking about is needed for actual transformation and positive ROI within a business. And honestly to me seeing these custom development numbers are not a shock at all because I know one how difficult the development can be. software projects generally new software initiatives within companies fail at a ridiculously high percent. I'll put it on the screen somewhere. I always forget it. And then AI projects are even more difficult than your typical software project. And so they're failing at 80 to 90% depending on which studies you look at. But in this case, MIT is talking about 95%. And that's not in the technology. There's also the human difficulty of having push back from staff internally. There's layers and layers of fear and lack of skills and sort of lack of understanding of the technology that you run into when you try to deploy things into production. And I think that's what's happened is the world has just like been charging straight ahead on this AI train and just hit the brick wall of like how hard it really is to get proper transformation within companies. And so while we still have a long way to go on that side, we are seeing incredible numbers in terms of generic LLM usage and what managers and executives are perceiving to be a positive return from their team starting to use it. because we are seeing really good usage metrics like 82% of people in companies at least in the Wharton survey are using generic tools like chatb or co-pilot weekly and that's for use cases like data analysis summarization and editing and so what we're seeing with these two different studies is not necessarily a contradiction it's like an evolution of hey look let's just train up the team and teach them to use this and they can start to use it within their own workflows if you train people up with these skills to use these tools they will find ways people are inherently lazy and like water always runs downhill and with the right skills people can start to autom autoate or delegate to AI parts of their job. But the tricky thing about that is that these studies are saying that doesn't necessarily trickle down to your bottom line, right? You're not going to see that massively pop up because it's not rethinking a process. It's not cutting whole parts of a team. And that is one other thing that these studies are pointing out is that it's mostly augmentation. It's mostly humans being empowered by AI, not replaced by it, which I think overall is a great sign that this thing is going to be sort of a bit slower and give us enough time to adapt and train people rather than being like a hot knife through butter and suddenly everyone's

without a job. Okay, so given all this, what are the plays that are on the table for us as AI business owners or AI agency owners? Well, I'm going to run you through them now. So, the first one is going to be a training based play like teaching people how to use the generic tools cuz these things are saying that generic tools are providing a great lift to companies. So, if you can just package up a nice education product or a course or program or whatever it is and go to companies and say, "Hey, look, I can come in and do workshops. I can train your staff to use these generic tools which in three out of four companies are providing a positive ROI and at the moment are the least risky way for you to invest in AI in order to see some gains and not be left behind ultimately. I talk about that quite a lot and I think there's real value and if you could go to companies and offer a training program which will teach one sort of the basics of how to use chachi you might throw in a claude or a couple other tools in there as well and then by department you train them up on say here's how a marketing team can use cachebt for like deep research for the different types of models and the different thinking levels here you can use agent for different tasks maybe like looking up leads or you want to like scrape some things off Google maps and then you make a mini course for each department you've got a very valuable educational product on your hands that you can go to companies and say lowest hanging fruit is to train your teams on it. I've got the generics like here's the base of how to use this stuff and I'm going deep on each department on how to get the most out of it given your current role and given the kinds of things that you're

going to have to do. The second play is the talent play. This means helping companies to get the talent they need because 49% according to the studies are struggling to hire the right AI talent and find people with the AI skills that they need. Whether that's basic AI literacy that they're looking for in roles that they're hiring in any role these days or it's specific Gen AI like AI automation developers or even AI engineers. All of these types of skills are needed in the workforce right now. And companies are willing to pay the big bucks to get the right talent. So your options here if you're in the AI space, maybe you're looking to start an AI business or you have one already, you can go into maybe upskilling. How can I make an AI upskilling program where either I'm selling to people who are like individuals looking for it or I can again go to companies and offer this like upskilling internally or I can go into recruitment and I can start to either make my own program teach people how to be sort of skilled in certain types of AI and then I can go and give those as placements or I can go out and actively headunt and search for AI talent and be able to connect that between companies and the individuals. There's so many plays there. Maybe you want to go onto a Discord server and sniff around for people who look like they're really good. like you find these certain developer communities, see who's contributing a lot, and then you just need to go to companies and say, "Hey, are you looking to hire AI roles? If so, let me know your budget. " And then you can just start to do some matchmaking. I know that's a massive oversimplification, but there are places where people are probably not willing to get their hands dirty. Go into school communities that like mine, I've got 260,000 people in there. Go into these more developer focused AI engineer communities if that's the sort of stuff you're looking for, but the talent is in demand. you can just chat and have a brainstorm about how could I maybe start to fill some of this demand with a recruitment service or some other kind of upskilling etc. The third play is

going deep in custom development and trying to help companies to get to that 5%. Now this is of course going to be a little bit more difficult than just making a chat GBT course but one of the key takeaways out of the MIT report even the skeptical one the one that's kind of all over the stuff and saying that it's only 5%. They found that external partners succeed two times more than internal builds. Meaning that companies who work with agencies and external partners and in any AI service provider are seeing higher success rates than if they try to do it internally. That's some great ammo to go to your clients with or potential clients and say, "Hey, look, let me paint the whole picture of what's really going on. There's custom development, there's the education piece as well. That's the lowest hanging fruit. " But really, if we're talking about transformation and getting a competitive edge, it's going to be some sort of custom development that we need to work with you on. And honestly, my biggest takeaway as an agency owner here is that we need to be one putting our prices up and sort of underelling the stuff a bit more and saying, "Look, I'm going to be fully honest. It's going to be a lot longer. There's of sort of feedback and iteration loops depending on the size of the company. It might be a 2 to 4 week audit. And then you can do the development process, maybe four to 6 weeks, but then after that, it's like we're just letting you know it's going to take another 4 to 8 weeks of optimization and feedback from you to get this to where we needed to go. " and being fully open and honest with your clients about what it really takes to see success with this technology and how they can be in the 5% and not the 95. A key quote from the MIT report that should give you a lot of encouragement for going down this route is the shift from building to buying creates unprecedented opportunities for vendors who can deliver learning capable deeply integrated AI systems. And now if you're

an AI agency owner, all of this stuff might be giving you a bit of the shakes and going like, "Oh, maybe this is going to be harder. I don't know thought it's not worth me going into. " No, it still is. It still is because we have some other great insights coming out of this stuff too. Like smaller companies of $50 to $250 million in annual turnover are getting an ROI much faster. No surprise, not rocket science. Smaller companies are being able to move on this faster and getting better returns. They're more agile. They make quicker decisions and they're able to get a positive ROI according to the study in around 90 days with a custom AI solution. Whereas for enterprises, it's taking more like 9 months. So stick with small companies if you want to be able to get them to that positive ROI quicker and it's going to be much easier. There's less stakeholders. there's this sort of technical depth. There's less big systems you got to work with. Smaller companies obviously smaller budgets, but if you're looking to get some really good case studies, staying smaller as they go. And also some great news about the investment wave that is coming. Actually found an interesting point where the perceived benefits of AI are different between management and executives. They're like executives are much more optimistic and sort of say overall they've had more impact. Then when you look at the managers responses, they are sort of a little bit less optimistic, a little more pessimistic because they're obviously a lot closer to it. They're seeing where it's struggling, where it's not. So there's a lot of enthusiasm from the top still and it's only increasing. We have 88% of the companies expecting budget increases aka they're going to be spending more on their AI initiatives going into 2026 and 70% of the companies expect revolutionary impact to their business in 2 to 5 years from these investments in AI. So basically companies are getting ready to get their wallets out and they've been investing in it this year. They're only going to invest more next year. So yes, the challenges are difficult. Well, I mean you can go and sell education and training if you want. uh we can call it like the easy money path where you're selling generic tool training to companies lower risk implementation you got 5 to 10k workshops or courses that's going to address that immediate need or you can go the more kind of strategic and long-term path where you were doing high ticket custom development but factoring in as MIT found that extra time needed for factoring learning into the system making them a lot more flexible and able to incorporate the feedback given from either users or from the team that you're working with to get it set up and if you're taking this approach you need to make sure you are charging more than you think because these optimization phases could take a while but that is the cost of true transformation is getting deep into these processes that require a lot of tweaking until they eventually get to the point that it's actually doing the job and as we said before there's this kind of other path that's emerging the talent services is also something very interesting that you should consider maybe you have connections maybe you've got experience in that area but there is essentially a huge amount of demand for AI talent and you can have a little exploration with HBT to see what angles you can maybe enter that from so this whole divide we're seeing is not necessarily about AI succeeding or failing it's about identifying and understanding where AI is getting the quick and easy wins versus where the hard problems that are ultimately going to be worth solving, but we're just being kind of brutally honest about the difficulty of some of these problems and starting to understand the nature of what true AI transformation looks like. As you guys know, my agency, Morning AI, we're positioned as AI transformation partners because we do all of these things. We educate teams, we do the consulting and identification process to find where AI can help them most. And then we do the development and that's kind of tying everything together. You won't be able to get there immediately with your agency here with your AI business. But starting somewhere, maybe it's recruitment or starting maybe with the education piece or starting with just consulting and eventually growing to be able to have that whole holistic sort of service package is where you really should be working towards because that's clearly what these companies need. It's not only the easy ones, but it's the hard stuff as well. And someone who can really stick it out with them through that learning and optimization process so that you can guide your clients to be in that 5%. So yeah, the point of this

video was to save you guys hours and hours of reading those reports. So I hope I've done that and I hope you got a much clearer read on sort of where we are in the space right now and what opportunities you have for your agency going into 2026. So if you are a business owner wanting to work with my team and do the whole AI transformation process with some of the best people in the space, you can get in touch with me and my team in the description below. And also if you're an AI agency owner not already in my free or paid communities or links down there if you want to get more information on that. That is all for the video guys. If you are interested about in how you can make money with AI avatars, one of the top technologies I think going into 2026, I got a full video breaking that down there. How you can make your first $100,000 according to someone who's actually done it. You can watch it up there. Thank you so much for watching. I'll see you in the next one.