I just wrapped six weeks teaching AI in Financial Services at Columbia Business School — and I walked away with a sharper sense of where AI's real value is created. It isn't inside the model. It's inside the person who knows how to wield it.
The course covered the full arc of what organizations need to understand to compete with AI: the fundamentals, practical use cases, how to build and deploy AI frameworks, data quality and governance, the infrastructure layer — chips, energy, hyperscalers — and, perhaps most timely, the future of work in an AI-accelerated economy. We finished with a capstone project that challenged students to do what most organizations are still struggling to do: build something real.
The results were genuinely impressive. Teams designed and deployed working applications for financial services — risk analytics tools for investment decision-making, credit card optimization platforms, debt management assistants, analytical support capabilities, and intelligent tools for the retail investor. Each group built a business case, architected and deployed a functioning tool, engineered guardrails, and established a robust evaluation framework. That is the full enterprise lifecycle, compressed into weeks.
Where Teams Set Themselves Apart
Here is what became clear as I watched each team present: the technology itself is no longer the differentiator. Today's vibe coding platforms and foundation models have dramatically lowered the barrier to building. Every team was capable of assembling a working prototype. That is remarkable on its own — but it is also now table stakes.
What separated the best teams was domain knowledge. The groups that produced the most innovative work were the ones who stopped treating the model as the product and started treating it as a surface for their own expertise. They went deep on the real pain points their end users face. They anticipated edge cases that a generalist model would never encounter on its own. They knew what wrong looked like in their domain — and they built guardrails accordingly.
"Models are powerful. But it's the expert who makes the model shine."
This is a message I carry into every boardroom and executive workshop I lead: organizations that win with AI are not the ones who adopt the most tools. They are the ones who most effectively pair those tools with the deep institutional and domain knowledge they already hold. The model is the instrument. Your expertise is the music.
Discipline Meets Creativity
The other thing that struck me was the balance students brought to the work. There was genuine creativity — novel framings, unexpected applications, elegant solutions to real problems. But it never felt untethered. Teams grounded their ideas in sound methodology: structured evaluation criteria, measurable success metrics, thoughtful risk and bias considerations. They were creative and rigorous at the same time. That combination is rare, and it is exactly what enterprise AI deployments require.
Four Things This Course Made Clearer Than Ever
Domain expertise is the moat. In a world where anyone can build, the advantage belongs to those who know their domain deeply enough to direct models toward genuinely novel solutions.
Edge cases reveal real mastery. Generic solutions handle generic inputs. It is how a model behaves at the edges — in the hard, specific moments — that determines real-world value.
Guardrails are a feature, not a constraint. The best teams built evaluation and safety frameworks from day one, treating responsible AI design as a competitive advantage.
Creativity and rigor are not in tension. The strongest work was both bold and disciplined. This is the combination every organization should be developing in its AI talent.
A Note of Genuine Gratitude
I want to take a moment to thank Columbia University and Columbia Business School for the opportunity to bring this curriculum to life in such a dynamic environment. CBS gave us the space to go deep on a subject that is moving fast, and the institutional support made all the difference.
Most importantly, I want to express my sincere appreciation to the MBA students who brought their absolute best to this course. You arrived with open minds, asked sharp questions, and — when it mattered most — brought your unique professional experiences and domain expertise to bear in ways that elevated the entire class. Watching you translate six weeks of frameworks and concepts into deployed, working applications was exactly the outcome I had hoped for, and you exceeded it.
The financial services industry is being reshaped in real time. The leaders who will navigate that transformation most effectively are precisely the kind of thoughtful, analytically rigorous, and creatively ambitious professionals you are becoming. The work you did in this capstone is a preview of what you will build in your careers.
For the organizations I work with, this experience only deepened my conviction: the path to meaningful AI value runs directly through your people and your institutional knowledge. The models will keep improving. The question is whether your teams are developing the expertise to direct them.
Back to Insights