Is AI just a cost center? That’s a question I hear a lot. And for many companies, the honest answer is yes. We spent six months digging into the data from over 500 enterprise AI projects, and the results were shocking. There's one fundamental flaw in how companies are approaching AI, and it's costing them billions.
For the past three years, my consulting firm has been in the trenches with companies ranging from Fortune 500s to fast-growing startups, all trying to crack the AI code. We’ve seen it all: the good, the bad, and the downright ugly. We’ve seen companies spend tens of millions on AI initiatives that go nowhere, and we’ve seen companies achieve 10x returns on their AI investments. The difference between success and failure is rarely about the technology itself. It’s about the approach.
The Great AI Charade: Why Most AI Projects Fail
Let’s be blunt: most corporate AI is what I call “AI theater.” It’s about looking innovative, not about driving real business value. Companies are so eager to jump on the AI bandwagon that they skip the most critical step: asking the right questions. They get mesmerized by the tech and forget about the business.
I remember one case, a major CPG brand—let’s call them “Acme Fresh”—that spent $20 million on a state-of-the-art demand forecasting engine. They hired a team of brilliant data scientists, built a complex model, and after a year of work, they unveiled it with a big press release. The problem? The model was only 2% more accurate than their old Excel-based system. And it was so complex that no one on the business side understood it or trusted it. The project was a complete failure, and the team was quietly disbanded six months later.
Acme Fresh fell into the classic trap: they started with a solution (“we need AI!”) instead of a problem. They didn’t ask: “What are the real drivers of forecast inaccuracy in our business? Is it data latency? Is it unforeseen promotional activities? Is it a lack of collaboration between sales and marketing?” They just threw technology at the problem, hoping for a miracle.
This isn’t an isolated case. Our analysis of 500+ projects revealed a disturbing pattern:
- 78% of projects were initiated without a clear business problem in mind.
- 64% of projects failed to deliver any meaningful ROI.
- Only 12% of projects were considered a “runaway success” by their sponsors.
These numbers are not just disappointing; they are a damning indictment of the current approach to AI in the enterprise.
The Billion-Dollar Mistake: Focusing on Models, Not on Systems
The fundamental flaw I mentioned earlier is this: companies are obsessed with building models, not with building systems. They think of AI as a magic black box that you feed data into and get answers out of. But AI is not a black box. It’s a component of a much larger system that includes data pipelines, business processes, and human workflows.
Think about it. A brilliant machine learning model is useless if the data it’s trained on is garbage. It’s useless if its predictions are not integrated into the day-to-day decisions of the business. It’s useless if the people who are supposed to use it don’t trust it or understand it.
I once worked with a large industrial manufacturer that had developed a predictive maintenance model for their factory equipment. The model was incredibly accurate; it could predict equipment failures with 95% accuracy up to two weeks in advance. But it had zero impact on the business. Why? Because the maintenance schedules were planned three months in advance, and there was no process in place to act on the model’s predictions. The model was a technical masterpiece, but it was a business failure.
This is the story of AI in so many companies. They have pockets of brilliance, but they lack the connective tissue to turn that brilliance into business value. They are building islands of innovation in a sea of legacy processes.
The AI-Native Enterprise: A New Blueprint for Success
So, what’s the alternative? How do you avoid the AI theater and build a truly AI-native enterprise? It starts with a radical shift in mindset. You need to stop thinking about AI as a project and start thinking about it as a core capability of your business.
Here’s the blueprint we’ve developed based on our work with the most successful AI-driven companies:
Start with the Problem, Not the Tech: Before you write a single line of code, you need to have a crystal-clear understanding of the business problem you are trying to solve. And I mean really understand it. Go talk to the people on the front lines. Shadow them for a day. Understand their pain points. Only then can you start to think about how AI can help.
Think in Systems, Not in Models: Don’t just build a model. Build a system. Map out the entire end-to-end process, from data acquisition to business action. Think about how the model’s output will be consumed, who will use it, and how it will change the way they work. Design the human-in-the-loop workflows from day one.
Data is Everything: I can’t stress this enough. Your AI is only as good as your data. Invest in building a clean, reliable, and accessible data infrastructure. This is not a glamorous work, but it is the foundation of everything else. The companies that get this right are the ones that win.
Iterate, Iterate, Iterate: Don’t try to boil the ocean. Start with a small, well-defined problem and build a simple solution. Get it into the hands of users as quickly as possible and then iterate based on their feedback. The goal is to deliver value quickly and learn as you go. The days of multi-year, big-bang AI projects are over.
One of my portfolio companies, a fast-growing e-commerce startup, is a great example of this approach in action. They wanted to improve their product recommendations. Instead of trying to build a super-complex, all-encompassing recommendation engine, they started with a simple collaborative filtering model. It wasn’t perfect, but it was better than what they had. They deployed it to a small segment of their users and closely monitored the results. They learned that the model was good at recommending popular products but not so good at surfacing new and niche items. So, they added a content-based filtering component to the model. Then they added a real-time component to capture the user’s current browsing behavior. And so on. After a year of continuous iteration, they had a world-class recommendation engine that was driving a 15% uplift in revenue.
The Bottom Line: AI is Not a Spectator Sport
The data is clear: the companies that are winning with AI are not the ones with the fanciest models or the biggest data science teams. They are the ones that have a deep understanding of their business and a pragmatic, systems-level approach to AI implementation. They are the ones that are willing to do the hard work of integrating AI into the core of their operations.
So, the next time someone in your organization says, “we need an AI strategy,” ask them this: “What problem are we trying to solve? And how will we build a system to solve it?” If they can’t answer those questions, you’re probably on your way to another expensive AI theater production.
Don’t be a spectator. Be a builder. The future of your business depends on it.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.