Michael Olenick: At Duke Fuqua, Using AI To Strengthen The Bonds Between Humans

Latest in an ongoing series exploring how business schools are integrating AI into teaching and research, one school and one faculty member at a time. See also At Indiana Kelley, “AI Is,” Not “AI Will,” Training Students To Leverage Increased Cognition (Michigan Ross), Rebuilding The Role Play (Kellogg), Learning Requires Friction (Wharton), The Deck Is Dead (Stanford), and We Teach MBAs To Apologize About The Best Tool They Have (INSEAD).

Sharique Hasan teaches at Duke’s Fuqua School of Business, arriving there after time at Wharton. He recently chaired the AI task force at one of management’s top journals and co-authored More Versus Better, a paper on AI’s effect on academic and student output that’s been downloaded more than 30,000 times and covered in Nature and The Economist. His core finding: AI has made production radically cheaper, but not better, and the gap between the two is where business education now has to live.

Every school I’ve talked to for this series is doing something different. Stanford leans into prototyping. Wharton is working through what it calls cognitive surrender. Kellogg built an entire department around the question. Duke’s answer turned out to be the most surprising one yet: rather than chasing prompt engineering, Duke has built fully AI-enabled classrooms designed to strengthen the bonds students form with each other, not replace them.

Below, we talk about that, plus why judgment, not production, is becoming the scarce resource in business, a simple framework for sorting what AI does well from what only humans can do, and why Hasan believes the future of expertise still lives exactly where it’s always lived: inside people.

Poets&Quants: What is Duke doing with AI in the classroom?

Sharique Hasan: One of the most interesting things we’re doing that I haven’t seen replicated anywhere else is building fully AI-enabled classrooms. At the end of a session, the professor knows whether students actually understood the concept, what’s missing, how it links to other concepts, based on the classroom conversation itself. There are privacy issues, obviously, but we’ve worked through them carefully. I haven’t seen anything like it at other schools.

P&Q: Why build it that way instead of just using AI for the usual things, grading, prep, and so on?

SH: What I like about our approach is we have a different theory of business education than most places. A lot of what students learn doesn’t come from professor to student. It comes from peer effects, the interaction, the sharing, the learning from each other. If AI can strengthen those bonds between students rather than replace the human parts of education with something more like a cyborg experience, that’s a huge win. And now the professor has a granular view into how that’s actually unfolding.

P&Q: You co-authored a paper called More Versus Better. What’s the core finding?

SH: The core finding is that AI has dramatically entered the writing process for academics, and it happened for students too, right after ChatGPT launched. But it’s resulted in more, not better. For researchers and for students. People produce a lot more, faster, but they don’t fully know what they’re doing. Mistakes get buried in the output. It looks like a research-shaped object, but put it in front of a faculty member with thirty years of experience and ask them to explain it, and they can’t.

P&Q: What does that mean for how you teach?

SH: Production is now very cheap. Anyone can do it. Judgment, the ability to tell whether something is actually good, is incredibly rare. Cultivating judgment is going to be a huge part of MBA education going forward.

P&Q: Is this the same problem you’re seeing outside the classroom, with executives?

SH: We’ve been talking to executives about this too. Two years ago, the problem was adoption, how do we get people to use AI. Now adoption is basically universal, and managers are stuck in the middle. Below them, massive volume: 2,000 lines of code becomes 20,000. Above them, leadership demanding cost reduction and headcount cuts. Business education has to live in that tension. How do you lead when production is nearly free but judgment is the binding constraint?

P&Q: Why do you think some people can’t tell AI-written work apart from human work?

SH: Some people genuinely can’t tell the difference because they’re not reading closely enough, or they haven’t developed the tacit knowledge to know it when they see it. Someone buying a small business who has never run one before can be handed a projection showing a $30,000-a-year business easily becoming $3 million, and they won’t know it’s nonsense. Same thing happens with code, with research. You end up with so much volume and complexity that nobody understands it anymore, so people just shrug and say it works, good enough.

P&Q: Does that create a real risk problem for companies selling AI solutions?

SH: The foundation model labs won’t guarantee any specific accuracy rate, which leaves whoever is selling AI solutions downstream fielding questions about downtime, error rate, what happens if a tail-risk event wipes out a client’s data. Nobody has good answers yet. The next set of skills that matters is systems thinking how does this fit into an organization, what risk does it introduce, how do you design around it.

P&Q: How do you frame the strategic question for organizations right now?

SH: The real question isn’t how we use AI strategically, it’s how do we use humans strategically, since AI is cheap and getting cheaper while human judgment is the expensive, binding constraint.

P&Q: You mentioned a framework for sorting what AI does well from what humans do well. Can you walk through it?

SH: It comes down to a simple framework. What AI does well: searching, structuring, thinking (in the sense of reasoning, calculating, comparing, predicting), and translating, whether that’s language to language or a dense paper into a takeaway. What humans do well breaks into three categories. Leading: setting vision, inspiring, mentoring, motivating. Owning: being accountable, defending unpopular decisions, judging and weighing quality (there’s a large literature now on LLM-as-judge, and LLMs are genuinely bad at it). And integrating: building relationships, negotiating, convincing people, empathizing.

P&Q: Does that match how people actually spend their time?

SH: It’s telling that when you ask CEOs and mid-career managers to sort their actual day-to-day work, most of it clusters in the leading and owning row, exactly the part that’s hardest to automate and the part they say they wish they had more time for.

P&Q: Is the anxiety around AI justified?

SH: Part of it is FOMO and part of it is what I’d call FEMO, fear-mongering. People are heavily online and there’s a lot of noise about AI destroying jobs, and honestly a good amount of it is marketing. It’s not obvious AI is the smoking gun behind hiring slowdowns. There are plenty of confounders, and economists are actively debating it.

P&Q: What about the students themselves, what are they anxious about?

SH: For students, the anxiety is real but complicated. They’re spending a lot of money on the degree and want a job, and they’ve absorbed the message that AI capability is now a job requirement. But nobody actually knows yet what the effect of an AI-capable workforce is on productivity. What I do think is true: the value students add isn’t going to be prompt engineering. It’s going to be how they deal with people, build relationships, and convince others, with AI as the thing that frees them up to lean into that more.

P&Q: Should MBA students be trained as prompt engineers?

SH: We should be making great leaders, not prompt engineers.

There’s a concept worth holding onto here, the idea that most real knowledge in society isn’t codified, it’s embodied in people: a carpenter, a sushi chef, a master craftsman. AI doesn’t have that kind of material engagement with the world. It has training data and internal processes, but it isn’t actually doing the thing. And it’s general in a way that trades off against depth. You can’t be a jack of all trades and a master of all of them too.

P&Q: Last question. Where do you see expertise living, going forward?

SH: Expertise lives in the future where it’s always lived: inside people. AI enhances that expertise, but it doesn’t relocate it. Someone who’s run a business for fifty years knows more than they could ever write down. If every scientist who ever wrote a paper vanished tomorrow, could the next generation reconstruct all the knowledge those scientists held just by reading the papers? Obviously not.

I’m bullish on AI. I’m even more bullish on humans.

---

Michael Olenick, JD, is a former INSEAD research fellow and founder of VSTRAT.ai. He has spent four decades building AI and expert systems; his research has been used at INSEAD and, before then, cited by Congress and media on how to resolve the 2008 era financial crisis. Michael currently lives with his partner, Anastassia, in Austin, Texas.

© Copyright 2026 Poets & Quants. All rights reserved. This article may not be republished, rewritten or otherwise distributed without written permission. To reprint or license this article or any content from Poets & Quants, please submit your request HERE.