As AI moves from experimental tool to ingrained infrastructure, business educators face the increasingly urgent task of preparing the next generation of leaders to manage, build, and make decisions.
In 2026, the answer is not as simple as teaching students to use chatbots. Future executives and entrepreneurs will need to understand exactly how AI systems work—and how to use them responsibly.
At Columbia Business School, where students can choose from nearly 30 AI-related courses, that understanding has already reshaped how professors approach and engage with material in their classrooms.
On June 1, CBS convened faculty and experts from peer institutions across the US for a day-long symposium examining how generative AI, agentic systems, and human-machine collaboration must and will change graduate-level business education overall.
The symposium, titled The MBA Transformed? AI and Beyond, hosted by the CBS AI and Business Initiative, showcased best-in-class curricular innovation at a time when business schools must develop entirely new frameworks for leadership in an AI-native era.
Here are three key takeaways.
Go ‘one level deeper’ than you expect
With AI embedded across business functions, future leaders will need enough technical fluency to evaluate AI systems rather than simply defer to specialists. For Rama Ramakrishnan, an AI and machine learning professor at MIT Sloan, the right level of technical training for MBA students may be deeper than many business schools once assumed.
Ramakrishnan shared lessons from Hands-On Deep Learning, or HODL, a Sloan course he designed to teach MBA students how deep learning systems work and how to build with them. He designed HODL to give students enough technical fluency to understand what AI systems can do, where they fail, and how to work credibly with technical teams.
His guiding principle is to teach students to go “one level deeper.” That depth matters because AI is increasingly embedded in products, operations, analytics, marketing, health care, finance, and entrepreneurship. Future business leaders will need to ask better questions of technical teams, evaluate AI-enabled products, and make well-informed decisions about systems.
"The first thing I've learned is that teaching MBA students how things work at the right technical level creates a lot of employment opportunities,” said Ramakrishnan. And the right level is usually “one level deeper than you think they need.”
At “one level deeper,” he said, students “develop a granular understanding that allows them to put things together in creative ways to solve hard problems, and pursue ambitious projects. It also makes them much, much more credible with technical colleagues .”
HODL is designed to make that possible without overwhelming students who come from less tech-heavy backgrounds. Ramakrishnan noted that the course is “light on math” but heavy on intuition. Students complete a short Python prerequisite and must know core machine learning concepts such as train/test/validation, and overfitting. Then, with support, they build.
They code models and watch live demos. They make mistakes, and they learn that mistakes are part of the process. Most importantly, Ramakrishnan said, they gain “the confidence to do hard things.”
AI is much more than a chatbot or a shortcut
One challenge for business schools is that many students first encounter AI through chatbots, even though the technology is already reshaping workflows, analytics, and decision-making far beyond a conversational interface.
While an assistant like Gemini or Claude may be the most familiar interface for AI-users, it is a limited way to understand what AI can do, notes Guetta, who co-organized the symposium. In Business Analytics 3, an advanced analytics elective at CBS, Guetta helps students see the broader system underneath—encompassing models, embeddings, agents, tools, costs, guardrails, and workflows.
To make those ideas tangible, he built XLKitLearn, a tool that brings AI capabilities directly into Excel. MBA students already know Excel, which makes it a useful bridge between business intuition and computational thinking, according to Guetta.
With XLKitLearn, students can compare responses from different models, examine how much prompts cost, test model guardrails, work with embeddings, and see how agents use tools to complete tasks. This makes advanced AI concepts visible inside a business environment students already understand.
Guetta chose this approach to address the challenge of reducing unnecessary friction that comes from learning to code without removing the discipline of thinking carefully about how systems work.
That balance may define the next era of the MBA, according to Guetta. AI fluency will not mean simply knowing how to prompt a chatbot. It will mean understanding enough to work with their teams to create AI systems and integrate them into their company’s workflows.
AI education must teach students to build
The explosion of AI tools has created a new leadership challenge of knowing which tools to trust, when to use them, and how to adapt them to a specific business problem. At CBS, that crucial builder’s mindset shows up in new classroom tools. Professors Olivier Toubia and Malek Ben Sliman discussed two CBS initiatives designed to help students navigate and use AI more actively.
- MBA Copilot, which uses Retrieval-Augmented Generation to allow students to make their own AI assistant from external data sources like course materials, slides, notes, and other documents. By creating their own system, students encounter the infrastructure behind AI applications, including retrieval, databases, Application Programming Interfaces, and deployment.
- CBS’s AI List, which Toubia notes can be thought of as a “Metacritic of AI tools.” In a fast-changing market where new advances appear constantly, the platform allows students to discover, review, vote on, and organize various technologies by use case.
Both Toubia and Ben Sliman pointed to a larger shift in MBA education where students are not simply told which AI tools to use but are instead asked to understand a whole complex ecosystem. They will need to evaluate tools critically and customize systems for specific business needs.
That kind of fluency will matter in the workplace as AI becomes more pervasive. Leaders will need to decide not only whether a tool works, but whether it is appropriate, reliable, secure, cost-effective, and aligned with the problem they are trying to solve.
