The primary danger of AI in business may not be that leaders use it too little. It may be that they trust it too widely.
Generative AI has made it easier than ever for managers to produce analysis and forecasts at extraordinary speed. Yet the hardest part begins after the output appears, when leaders decide whether the answer is useful, and what action should follow.
This is the problem that is tackled at the center of “Leadership Intelligence in the AI Era,” a Columbia Business School course.
As one of the more than 20 courses on offer as part of CBS’s AI curriculum, the class prepares students to lead in a business environment where AI is accelerating information but does not eliminate uncertainty or replace human judgment.
“The course is focused on the hardest thing that a leader needs to do — the most important thing, actually, that the leader needs to do — which is making decisions,” says Oded Netzer, the Arthur J. Samberg Professor of Business, who teaches alongside Christopher Frank and Paul Magnone.
Over the years, Netzer, Frank, and Magnone have taught students and executives how to make better decisions in data-rich environments. Once called “Leading in a Data-Driven World,” the course has evolved in response to developments in business, ensuring students leave with pragmatic skills they can immediately apply. First came the explosion of big data. Then analytics platforms, dashboards, and increasingly sophisticated models. Now, generative AI can produce summaries, drafts, forecasts, code, decks, and visualizations in seconds.
“The question we keep asking ourselves,” Netzer says, “is, ‘How do you best make decisions in a world of data, in a world of AI, and in a world in which uncertainty still resides?’”
As AI creates more output, leaders need good judgment
Business leaders are surrounded by information, and many struggle to turn that information into action. AI raises the stakes because it can make analysis feel faster, cleaner, and more authoritative than it really is.
Frank, a longtime leader at American Express and Microsoft, notes the irony of struggling to make decisions in a world where businesses are drowning in intel. He cited the cliché, ‘What gets measured gets done,’ and pointed out that, by contrast, “everything is measured today and it feels like nothing is getting done. AI is accelerating this paradigm. More information, more analytics and less time for judgement and action."
That gap between measurement and action is where the course begins. Students are asked to think beyond the mechanics of tool use and focus on the leadership work surrounding AI: What assumptions are embedded in the output? What context is missing? What risks matter and what decision follows?
The course frames the challenge as one of judgment in a world overflowing with analytics, dashboards, and AI outputs. Leaders have to sort through signal and noise, pressure-test assumptions, and validate claims. The goal is to give managers the ability to work with AI output productively while remaining accountable for the decisions they make.
For Magnone, who has held leadership roles at Google and IBM, that accountability is key. “We can think of AI fluency as balancing automated intelligence with human judgment,” he says. “The human should not abdicate judgment and the actual decisions to tools, no matter how good they are.”
Quantitative Intuition for the AI era
Netzer, Frank, and Magnone developed Quantitative Intuition, or QI, as a set of practical tools to help leaders make better decisions with data. In the AI era, QI gives students a way to interrogate machine-generated output with the same discipline they would apply to a market analysis, financial projection, customer insight, or strategic recommendation.
“What separates leaders who use AI well from those who just use it a lot are what we call the pillars of Quantitative Intuition,” Netzer said. “Precision questioning, contextual analysis. Do you provide the necessary context — your own knowledge — as context to AI in order to help you?
“And finally,” he said, “how do you react to the output of AI? How do you interrogate the output of AI? How do you synthesize that in order to move it into decisions?”
The course turns those challenges into a sequence of applied learning. During the block-week format, students move from the “new intelligence frontier” to precision questioning and framing, contextual analysis, synthesis and delivery, and a final practicum. They learn to ask better questions and prompts, identify algorithmic and cognitive biases, interrogate AI and analytics outputs, and move from passive reporting to active direction.
In many organizations, AI has intensified the rush to address a problem before leaders have properly framed it. The course slows students down at that critical first step. Before a prompt, a dashboard, or a model, there has to be a precise business question.
“When you talk about AI fluency, it is really the difference between what the tool does versus when you should use it,” he said.
That distinction comes alive in the course’s workshop format. Students work with real-world business problems, apply the day’s concepts to group projects, and prepare executive-ready outputs.
Leaders do not hand important work to a new team member without direction, expectations, review, and judgment. AI requires the same discipline.
In this way, the instructors encourage students to think of AI as part of the team. “Great leaders are great multipliers,” Frank says. So “how does [AI] have a multiplier effect on your team, on your culture, on your decisions?”
