Why Most AI Projects Will Fail — And How to Find the Companies That Won't

In this episode of Motley Fool Hidden Gems Investing, Motley Fool contributor Rachel Warren sits down with Steve Lucas, chairman and CEO of Boomi, to unpack what Wall Street is missing: Why the next wave of AI winners won't be the flashy model makers, how to spot the difference between a real AI strategy and expensive spin, and the single metric that separates transformative technology from hype.

To catch full episodes of all The Motley Fool's free podcasts, check out our podcast center. When you're ready to invest, check out this top 10 list of stocks to buy.

A full transcript is below.

This podcast was recorded on July 12, 2026.

Steve Lucas: The one company that is unequivocally making money from AI is in video. That's the one company that seems to be making a ton of money. There's a lot of other companies that have proven they can build amazing models and lose extraordinary amounts of money.

Rachel Warren: That was Steve Lucas, chairman and CEO of Boomi, explaining who is actually making money in AI right now and who isn't, is a 30-year enterprise software veteran who previously turned Marketo into a $4.75 billion acquisition. I'm Motley Fool analyst Rachel Warren. Steve has sat across the table from hundreds of CEOs navigating the AI moment, and what he's hearing might surprise you. We discussed the ROI reckoning that's coming, what separates real AI winners from expensive experiments. Why the next wave of big beneficiaries probably isn't who you think. We hope you enjoy.

Welcome back to Motley Fool Conversations. I’m Motley Fool analyst Rachel Warren. Today, we're looking past the AI Hype Cycle to focus on execution, data infrastructure, and true return on investment. Joining us is Steve Lucas, chairman and CEO of Boomi. Steve is a multi-time CEO with nearly 30 years of enterprise software leadership, including senior roles at Salesforce and Adobe. Previously as CEO of Marketo, he drove a massive turnaround resulting in a $4.75 billion acquisition by Adobe. Now at the helm of Boomi, a data activation powerhouse, serving over 30,000 global customers. Steve is here to talk about the current state of AI, where corporate tech budgets are actually moving, and how investors can spot the real winners. Steve, welcome to the show.

Steve Lucas: Thank you, Rachel. Happy to be here.

Rachel Warren: For the last few years, it seems as though investors have largely rewarded companies for simply having an AI strategy using the right AI buzzwords and earnings calls. But it seems we're entering something of the next phase in that journey, where Wall Street demands understandably measurable business outcomes and ROI. So I'm curious what do companies need to do or keep top of mind to actually deliver to that end?

Steve Lucas: First of all, I think you're absolutely right. Over the past couple of years, we've gone from we didn't have AI, now it exists. To boards, pressuring executive teams, CEOs, and leaders at companies to put AI into their company, build an AI strategy. In the two years that we've seen that pressure mount, we've seen the birth of Agentic AI inside of businesses and all those things. I think that the pressure has now started to subside. As you pointed out, now it's about returns, and I've been quoted a few times as saying that ROI supersedes AI, and there is no doubt that that is the case today. I just think it's the enormity of the pressure on that left-hand side, coupled with rushing into a lot of AI projects. But we're not seeing the high rates of return that you'd expect from businesses. Now, that's going to change as AI matures, and how organizations manage AI matures as well, but we're definitely seeing a change in the winds.

Rachel Warren: Do you think we've reached a point where AI spending could become a drag on earnings for companies that fail to demonstrate meaningful returns on those investments, or do you think it's just too early to really make that determination?

Steve Lucas: If you look at the four major hyperscalers in the U.S. alone and the amount of capex that they put into AI last year versus this year, this is the canary in the coal mine. Last year, it was around 400 billion, 410 billion. This year, it's over 700 billion. For companies. That's an extraordinary increase in spending, and obviously, that is not reflective of the broader market, but it's an indicator of the broader market. The broader market organizations, their spending on AI is way up. Their spending on software applications is down, and spending on infrastructure is up, as well. AI and infrastructure seem to be the two big investment priorities for large organizations.

But I think we are, you said the blank check. I think that's a perfect phrase, and we are at a place where organizations I walk into board meeting after board meeting, CEO after CEO, and what I hear continuously is help me show return. After these projects that we initially pursued, they're not providing ROI. That is it is real. It's in the market. That's being discussed, and you're starting to hear other executives call it out, as well, which is, let's stop trying to scare all the executives with these scare tactics into investing in AI, and in business, let's help them show real rates of return.

Rachel Warren: You mentioned earlier that you'll go in these meetings and there's C suite executives basically saying help us show that we are making this profitable or on the path to making it profitable. Are we at the time where boards are aggressively holding C-suite execs accountable for these investments yet? Is there still a bit of a grace period that we're seeing?

Steve Lucas: Grace period for now, not for long. You look at what's happening. I think if you talk about this hidden cost. I published a white paper recently that talks about the cost of training GPT-2, which I think all of us are largely familiar with. It was the first time we encountered this bewildering technology called large language models or GPTs. The cost to train GPT-2, and this is public data, was around just shy of $50,000. Manageable, affordable. The cost to train the frontier models that we're seeing in 2026: over $1 billion. It's extraordinary, so we've gone from used car to aircraft carrier. Now, that cost is heavily subsidized by investors. OpenAI is burning $3 billion a month. That's a reported, reliable number. You can't lose $3 billion a month into perpetuity. You just can't organization can sustain those negative economics, no matter how transformative the technology, so those costs will be borne by someone. It'll be the consumer, it'll be the enterprise, the business itself.

We haven't seen the full cost of AI yet, but every CEO I talked to, they say the same thing, which is my spend on AI from last year to this year went up 10X, 20X. You read in the news, people are saying, hey, we got to put a halt or put a stop to this AI spending. Even Elon Musk who loves to spend money, put a cap on what his employees can spend at Tesla and SpaceX. That was a recently reported news item, as well. The point being, is that while there's a grace period for now, most CEOs in the very short-term, the next six months will start putting caps on the investment within AI internally, and then there's going to be this heightened demand to see real ROI at a board level before any company spends tens, hundreds of millions, billions of dollars on AI.

Rachel Warren: Now, I believe you suggested that as many as 40% of enterprise AI projects could ultimately be abandoned in the end, and that's a very interesting figure, and it ties into what you've been talking about. I wonder if you could dive into that mindset, but also what are the characteristics of projects that fail? What separates them from the ones that are actually creating lasting value?

Steve Lucas: We're in a heavy era of experimentation with AI. To a certain degree, you have to expect AI projects to fail, partly because they're really easy to start. It takes five minutes, you crack your knuckles, and you're asking Cloud or OpenAI to do things with your business data. You can start very easily. But what happens is you ignore business requirements, strategic outcomes because you can just start iterating with AI. I do think to a certain degree, we're coding has become so easy now. Accessing or starting to build outcomes with AI good or bad has become easy, so we rush into these things and we forget the basics of outcome ROI productive results, so we rush in. That's part of it. Part of it is just experimentation, seeing what works, what doesn't, where there's ROI.

But even Gartner is saying that a number of these what we call agentic projects, that's just AI with agents working inside of businesses that these things either going to fail or fail to just return results and they'll be shut by the end of 2027. So you've got very credible analyst organizations calling this out. I absolutely believe that, as well, and we see it every day. I think right now, we're still on the edge of that grace period. I don't want to say blank check. But I think very quickly, these costs are going to get reined in. Especially when funding starts to dry up for some of these frontier model organizations, they're going to pass those costs on to the consumer.

ADVERTISEMENT: Tobi Lutke: Morning Brew Daily breaks down the biggest news in business every morning, so it fits seamlessly in your day. I'm Tobi Lutke.

Neil Freiman: I'm Neil Freiman, and each morning, we cover everything from the latest tech headlines to why nobody can afford a house right now. You'll leave each episode of Morning Brew Daily smarter and ready to take on the world around you, and some people are saying it's the best part of their morning. Because we know something you don't news doesn't have to be boring. Join millions of monthly listeners and check out Morning Brew Daily wherever you get your podcasts.

Rachel Warren: Looking back at your experience in the tech space and enterprise software, do companies tend to fail in this space because the technology doesn't work, because customers don't trust or adopt the tools? I'm curious how that can translate to the current AI revolution.

Steve Lucas: After 30 years in software, I know one thing, and that is if humans don't trust something, it will never be used. I don't forget AI. The reality is, I've seen thousands of business intelligence or analytics or data projects that fail because the data wasn't accurate, no one trusted it. All it takes is one time and one person sitting in a room. You've been in one of these meetings, too, I guarantee, and anybody listening to this, where they say, This is not accurate. This is wrong, and the moment someone asserts it's inaccurate or it's wrong, it degrades the entire system, the effort that everyone put into it. Suddenly mistrust begins, so change only happens at the speed of trust.

Here we are with AI, which it's not about data that humans rely on. This is potentially about AI that could do the job of a human. No, I don't think that. I think all that hype about AI taking people's jobs is just nonsense and fraud it's more trying to scare people into buying a product. Than it is reality right now, and certainly for the foreseeable future. That being said, I think that it's trust. That is the number one thing is, do you you, Rachel, do I, Steve, do we trust AI? If you're watching this right now, do you trust it? If the answer is no, except to build a good PowerPoint, then we're not there yet.

Rachel Warren: Many of the biggest gains from AI so far have gone to chipmakers, infrastructure providers. Obviously, those are entities that are really key to this continued build-out, but I wonder where you maybe see the next wave of AI winners emerging across the enterprise technology stack, just from your vantage point.

Steve Lucas: The one company that is unequivocally making money from AI is Nvidia. That's the one company that seems to be making a ton of money. There's a lot of other companies that have proven they can build amazing models and lose extraordinary amounts of money. Now, I am an AI protagonist. I'm a big believer in it, and I think the potential for AI is extraordinary. I think companies should be experimenting right now. I think trying and failing is a critical part of success, so I believe all those things. But the next wave of companies that will be I think AI beneficiaries. Certainly I believe our company's won simply because we are critical AI infrastructure. We enable organizations to connect to their data, radically improve the quality of it, and deliver it securely to models and AI agents. That matters.

But take a look at organizations like Snowflake, Databricks, Datadog. These organizations, they all center around the notion of data, and they're showing massive benefit. Their market caps are up. Their growth is up, and a lot of it has to do with they provide critical AI infrastructure and for business in particular. I think we've seen the vast majority of financial impact right now at the consumer level, and I think this next wave is going to be enterprise AI for B2B. But I think within that statement, it's going to be a whole lot of data, a whole lot of infrastructure. Those are the companies that we'll see accelerate. We're seeing that in our business.

Rachel Warren: I was going to say, how should investors evaluate opportunities in areas like data infrastructure, integration, automation, and governance? Those are less visible than some of the flashy AI models, but certainly really critical to successful deployment.

Steve Lucas: Whether you're an electric car or a petrol car believer, neither one of those goes without energy, and the data is the energy. Though, AI without data is meaningless. That's the plot that I think investors sometimes miss, and I think what's critical is understanding which businesses have data, data and graph modes, and I'm talking about a knowledge graph. Which businesses have unique data? They're either handling for their clients or they generate themselves, data that can't be automatically generated by AI or easily replicated. Those are golden nuggets for investors.

By the way, I think on the investment front, even when I think about this is you see a lot of organizations that are rushing into I characterize as like unique models where they're thinking about how do I provide some type of unique, like either hardware plus AI. A chip plus software, unique services like forward-deployed engineers, and I'll use the word ontology along with AI. People are looking for that. I’m not just a software-as-a-service type of message. You'll see that from every large organization, messaging to investors is, I'm not just software. That's critical for software CEOs because they want to deliver the message that I'm not easily disrupted by AI, so investors need to watch out for that. But I think data, I do think unique combinations of AI plus matter, but I'm still a believer in infrastructure. That's why I'm here.

Rachel Warren: I think that's one of the bigger questions. Obviously, some of the most well-known AI companies have not yet entered the public markets. There's rumors they will, if you think of anthropic or open AI, obviously, as a couple examples. But I think one of the biggest questions that a lot of investors have who are watching these companies right now is, how do these companies monetize long-term? How do they retain durable profits over the long run? It sounds like you're saying these are the models, the business models that they're developing in order to ensure that they're able to retain that financial growth and flexibility.

Steve Lucas: Again, I think for most investors, if you just step back and think about and let's not target or talk about any one particular company, but a frontier model company that's losing billions of dollars a month. The first narrative that we heard in AI narrative one. Oh, hey, AI is going to take all these human jobs, and you need to be ready for that. That didn't happen, at least not yet. But to a certain degree, I think a number of these public or frontier model AI CEOs, they were counting on AI taking these human jobs. They need it because there's just not enough software revenue to cover what it costs to train and build these models into perpetuity. They needed AI to be successful and consume some of the labor market, but here we are, and that hasn't happened.

What next? For AI to be successful long term, as I said earlier, these organizations, these big frontier models, they're going to have to convince enterprise organizations to use their model privately. That's a big step. This isn't just about consumer growth and what you or I use AI for at a small business level or at home. I'll pay my 200 bucks a month to Anthropic or OpenAI. But they've demonstrated that they need more revenue, so the enterprise shift is going to happen. I think there's so much more that needs to happen. I don't know how these organizations turn a profit without significant penetration into labor markets. Again, that's what they're counting on.

Rachel Warren: Very interesting. The other thing that this discussion brings up is we have heard a lot of companies, not just in the tech base, but certainly in the tech base, that have announced layoffs over the last, say, six to 12 months. In some cases, alleging AI efficiency, being a driving factor. I know you've talked a bit about today why you don't view AI as a replacement for human labor, and obviously, I think there are a lot of people that share that view. When we see companies that will announce layoffs, citing AI as a driving factor, is that trying to hide a hemorrhaging business, so to speak? Is that something that we should be perhaps reading between the lines a bit?

Steve Lucas: I think that there's a whole lot of spin going on right now. That's what I think. I think you're absolutely right. These layoffs where's the data behind it? Simply saying, we've achieved so much productivity. Therefore, we need to lay off 9,000 people as a matter of convenience. There's no fact in it, or if there is, it's very little and certainly uncommunicated. What I would want to see as an investor is show me where the rate of return, efficient you're more efficient in your finance team, more efficient in your engineering team. We're producing ten times the amount of code in a third of the time, show me the numbers, and show me the money. If you do that, then I'm going to start to buy into that. Absolutely. This is a matter of convenience. I don't want to say that AI is the scapegoat. I think it's just a convenient foil right now.

Rachel Warren: If we're investors, we're trying to look beyond the AI headlines. What are metrics, signals, company characteristics, even that you would focus on as we're trying to identify the longer-term winners in enterprise AI?

Steve Lucas: It's going to go way beyond market cap. I think market cap is semi-fickle right now. It changes with wins, and you can have anthropic make a statement about some new feature in their product, and then suddenly dozens of publicly traded organizations tank market capitalization wise because of a perceived feature or detriment to a market based on AI.

I think signals that you can look out for, we talked about this briefly, but I think it's going to be organizations that are showing acceleration number one in their new customer acquisition. If you truly have transformative technology, the indicator for the success of that technology is not forcing a product down your existing customers’ throats. It's our new organizations coming to you, seeking that innovation, that transformative technology? You obviously, as a compliment to the frontier model businesses, the unbelievable growth going 0 to a billion to 20 billion and beyond in a matter of a few short years, that's unprecedented. That came through entirely new logo or new client acquisition. That's consumer growth and new business growth.

Now, that will change over time as anthropic and OpenAI and Mistral and others run out of new logos, they're going to start to mature and think about new feature selling and customer expansion and all those gnarly words that happen with mature businesses. But the number one for me is new logo acquisition. That's always an indicator of a sufficiently transformative technology.

Rachel Warren: One final question, I'm to border. What excites you the most about AI right now? What are you most cautious about?

Steve Lucas: That is a really good question, and I'll try not to get too philosophical on you, Rachel.

Rachel Warren: Be philosophical.

Steve Lucas: Here we go. I've been a Type 1 diabetic for almost 30 years, and we all play the hand that we're dealt in life. I wear a sensor on this arm, and I have an insulin pump on that arm, and I've had a small army of amazing humans, doctors, clinicians, nurses that have helped me live a full and healthy life, which I love. But as I sit here, there's data streaming between my phone and my glucose sensor and my insulin pump 24/7 nonstop, in my life depends on it. The reason for that background is because AI has not just the potential or promise, but it will transform lives. It will not only make my own personal life and the ability to manage type 1 diabetes profoundly easier, which I welcome. I look forward to that day, but it will cure it, along with the vast majority of the things that we think about as challenges to human life today. I think within the next two decades, and I genuinely mean this, and this is why I'm an AI optimist.

The things that we treat that we are challenged with healthwise, that we have to overcome will largely be managed and or solved by AI. I think we will live longer, healthier lives, and that I love. That's what I look forward to. For the billions of people that struggle with health challenges out there, I think that there's an exciting future to look forward to. That being said, what I don't look forward to is AI being aware of all of that data, which I know it has to be, and it's used to market products and services, too. That is the fine line that we walk every day, is how do we benefit radically benefit humankind? While not trying to sell you a cup of coffee.

Rachel Warren: I think that's the perfect note to end on, and you've given us all I think a lot to think about. Thank you so much, Steve, for your time today.

Steve Lucas: Thank you, Rachel. I really enjoyed it.

Rachel Warren: As always, people on the program may have interests in the stocks they talk about, and The Motley Fool may have formal recommendations for or against, so don't buy or sell stocks based solely on what you hear. All personal finance content follows Motley Fool editorial standards and is not approved by advertisers. Advertisements are sponsored content and provided for informational purposes only. To see our full advertising disclosure, please check out our show notes. For The Motley Fool Hidden Gems Investing team, I'm Rachel Warren. Thanks for listening. We'll see you next time.

Rachel Warren has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Adobe, Datadog, Nvidia, Salesforce, Snowflake, and Tesla. The Motley Fool recommends Gartner and recommends the following options: long January 2028 $330 calls on Adobe and short January 2028 $340 calls on Adobe. The Motley Fool has a disclosure policy.