MIT research found roughly 95% of AI projects don’t achieve meaningful ROI. A lot of leaders are quietly asking why a few create real value while most never get past the demo.
The idea
From what we see in our work, it comes down to three differences. None of them are about the model. Today, technology is rarely the limiting factor.
If 95% of projects miss ROI, it’s not because AI is overhyped. It’s because many organizations aren’t structured to extract value from it.
Why it matters
- Real business ownership. Start with a painful, measurable problem tied to cost, revenue, risk, or efficiency, with an executive owner, KPIs, and accountability. Without an owner, it becomes innovation theater.
- Narrow focus, embedded workflow. Pick one bottleneck and put AI directly into the tools people already use. Not a new portal. Adoption lives where the work already happens.
- Data and organizational readiness. Projects fail on messy, inaccessible data and on organizations unready to change. You need usable data, aligned leaders, and change management from day one.
Three questions to ask
If you’re leading AI, ask: Who owns the business outcome? Is the use case narrow enough to embed in an existing workflow? Are we truly ready, from both a data and a people perspective?
What have you seen that separates the wins from the stalled pilots?
Join the conversation: see the comments on LinkedIn.
Read the full transcript
Hey, I’m Matt, and I was recently asked, can you cover what makes AI projects actually successful? And it’s a great question, especially after MIT somewhat recently released their research that shows that roughly 95% of AI projects do not achieve meaningful ROI. That number gets attention. And I believe a lot of leaders are quietly asking the same thing.
Why do a few AI projects create real value while most never move past a demo stage. From what I’ve seen in our work, it comes down to three big differences. Number one is real business ownership. A successful AI project starts with a painful and measurable business problem. Not just curiosity, not hype. A real problem typically tied to cost, revenue, risk, or maybe even efficiency.
And there’s an executive owner. There are KPIs and there’s accountability. If no one owns the outcome, it turns into innovation theater. Second is a narrow focus with an embedded workflow. The projects that work do not try to transform the entire enterprise. They pick one high-value bottleneck, one process, one friction point, and they insert AI directly into the workflow people already use. Not a new portal, not another tool to log into.
It shows up where the work already happens. That’s where adoption lives. And third, you need to have data and organizational readiness. Most projects do not fail because the model is weak. They fail because the data is messy or inaccessible or misunderstood or because the organization is not ready to change how work actually gets done. You need usable data. You need leaders that are in alignment. And you need change management from day one.
Today, technology is rarely the limiting factor. I believe that if 95% of projects are missing ROI, it’s not because AI is overhyped. It’s because many organizations are not structured to extract value from it. So if you are leading AI inside your company, ask yourself three questions. Number one, who owns the business outcomes?
Number two, is the use case narrow enough and can you embed it into an existing workflow? And number three, are you truly ready from both a data and a people perspective? I’m curious what you’ve seen that separates the wins from the stalled pilots. I’d love to hear your thoughts in the comments below. And thanks for watching.
Leave a Reply