AI’s capabilities are only limited by the questions we ask of it. Mostly.
The key to using AI effectively is to think about it in three levels.
In this episode of The Insightful Leader, we explore the basic, intermediate, and advanced ways to wield it.
Insight in your inbox
Receive our newsletters to keep up with the latest research and ideas from faculty at the Kellogg School of Management.
https://insight.kellogg.northwestern.edu/newsletter/insightful-leader
Podcast Transcript
Laura PAVIN: If you’re like me, AI is becoming a bigger and bigger part of your life, both at home and at work.
Maybe it helps you prepare for the day before you walk out the door.
PAVIN: Okay, Google, what’s the temperature right now?
GOOGLE HOME: It’s 13 degrees Fahrenheit and sunny in Highland Park. It feels like 2 degrees Fahrenheit.
PAVIN: Great.
PAVIN: You might be using ChatGPT to procrastinate.
[typing sounds and Laura talking out loud]
PAVIN: I need a dump-and-go meal to make in the slow cooker. Got 10 minutes.
Okay, we’ve got creamy chicken and rice. Barbecue chicken. Chicken-pot-pie filling. Honey-garlic chicken.
PAVIN: Wherever you happen to be on your personal AI journey, most of us are probably past the point of asking, “Should I use AI?” to asking, “What should I use it for?” And, “How can I use it effectively?”
We spoke with some faculty from Kellogg and Northwestern’s McCormick School of Engineering: Professors Brian Uzzi, Matt Groh, and Julio Ottino.
They’re kind of wizards with crisscrossing appointments in both schools. And they’ve been thinking a lot about those questions and what organizations can do to start adopting AI in a smart way.
We talked to them about it at a roundtable discussion we had at an Insightful Leader Live event.
And they say the questions we humans ask are actually a big part of getting good with AI.
OTTINO: AI is a collaborator that becomes as good as the questions that you pose.
UZZI: You don’t want to ask the bot for answers. You want to ask the bot how to think better about approaching a problem. You want to ask it for advice on how to solve a problem. Don’t solve the problem for me.
PAVIN: These professors work across the fields of engineering, organizational leadership, and management. And they say the key to using AI effectively is to think about it in three levels. Levels as in basic, intermediate, and advanced.
They’ve put together a playbook to help you level up.
GROH: The highest level where we think there’s the most value strategically is actually in this idea of advice. So it’s automation, answers, and advice. And advice is all about how AI can help us think through solving a problem.
What that means is AI can actually help us solve problems more effectively, creatively, thoughtfully.
PAVIN: Welcome to The Insightful Leader. I’m Laura Pavin.
In this episode: how to avoid common pitfalls and make the most of AI, step by step—or level by level.
And we’re not just talking about large language models like ChatGPT and Claude, by the way. It’s AI more broadly. Like machine learning, computer vision, ranking systems, reinforcement learning. The list goes on.
So if you or your organization have been Marco Polo-ing your way around an AI strategy, we’ve got some tips that can help.
That’s next.
…
PAVIN: Okay. So the first level I want to tell you about is the most basic one. It’s something you’re probably already doing with AI.
GROH: The very first level is automation. So it’s how AI can just do stuff for you without any supervision. And that’s often what many people think about.
PAVIN: That’s Matt Groh. He’s an assistant professor of management and organizations at Kellogg and an assistant professor of computer science at McCormick by courtesy.
He says when it comes to automation, you want to think about “set it and forget it” kinds of tasks. Tasks like scheduling social-media posts or sending email reminders to prospective customers.
If you’re looking to get started with AI in your workplace, automation is low-hanging fruit.
Here’s why. AI can help you automate what you’re already doing. You don’t need to reinvent the wheel.
UZZI: There’s a myriad of applications for automation, and it’s basically about letting AI do it for you.
PAVIN: Brian Uzzi is a professor of leadership and organizational change here at Kellogg and a professor of industrial engineering at McCormick by courtesy.
UZZI: So it’s not like you’re creating a new workflow. You’re really looking to replace the workflow you have now with automation. And when you do that, you free up time and resources for other productive activities.
PAVIN: So what kinds of workflows are we talking about?
Think about a visit to the doctor. Have you ever had a medical scribe sit in on your appointment? A medical scribe’s job is to take notes. Notes that go into your medical chart and aftercare plan.
These days, some healthcare providers are using AI instead of human scribes to capture the play-by-play of your doctor’s visit.
GROH: Turns out that companies like Abridge, now valued at $5.2 billion, or companies like Mobius, which is a smaller startup in that space but also doing really awesome work, can actually just come in, allow doctors to focus on the diagnosis, the interaction, and the treatment plan, and not have anyone else in the room.
So it cuts costs a little bit, but more than that, it actually organizes information in a better way than it’s ever been done before. So then you can track your patients more effectively.
PAVIN: Another example of automation is computer coding.
Matt Groh says companies like Amazon are leaning hard on AI to perform routine coding tasks. In 2024, Amazon’s CEO Andy Jassy posted on LinkedIn that Amazon’s generative AI software tool was “a game changer” and also a big money saver.
GROH: Their engineers can now do all the refactoring, all the kind of boring code stuff to keep up databases—not to create new products, but to make sure that the current products you have stay up to date with changing technology.
That would often take, let’s say, 50 days to do certain specific tasks. That goes to five hours. And because that happens for many, many employees, they essentially estimate a quarter-billion-dollar annualized savings.
PAVIN: So the promise of AI-driven automation is efficiency and savings. And freeing people up to do work that’s more creative and ideas-driven.
Julio Ottino is an Institute Professor at Northwestern’s McCormick School, where he was also the former dean. And he’s a professor of management and organizations at Kellogg.
And, while AI can feel like this magical wand, he says that automation may not be a win–win for everybody.
OTTINO: If you start at the level of automation, there’s no question that if you start putting AI-driven robots in a warehouse at Amazon, you will need fewer people.
The question is, can you assimilate those people into higher-level activities? Less-boring routine?
But this is where communication becomes an art, because you have to present a case that may not be wholly positive for everybody.
PAVIN: To recap: automation is the most basic level in this three-level AI-adoption playbook.
Think about the set-and-forget things on your to-do list you can put on autopilot. Maybe AI can help carry that load.
…
PAVIN: Okay, ready to move on to level two?
GROH: The next level is answers. So AI can give you answers to questions you have, but you might need to verify it in some way.
PAVIN: For instance, last year, Kellogg professor Brian Uzzi was searching for answers to a thorny problem.
Uzzi has a second home in Florida, and it was destroyed by Hurricane Helene in 2024. After the dust settled, he started calling around to see how much it would cost to rebuild the foundation. He needed it to be at least 15 feet above ground.
UZZI: I called engineers for weeks. Nobody had time to talk to me. I finally got one who would talk to me. He told me he couldn’t give me an answer for two months.
PAVIN: Two months. Uzzi wasn’t getting anywhere. So on a whim, he decided to hit up ChatGPT.
UZZI: So I basically gave a lot of textual information to a chatbot, and it came back with a detailed engineering estimate. The answer was out there, and ChatGPT helped me find it. And in the end, what ChatGPT gave me astonished me, because it was pretty close to what the engineer gave me two months later, for a very high price.
PAVIN: ChatGPT gave Uzzi confidence that the estimate from the human engineer was in the right ballpark, even if the price tag happened to be spendier than he was hoping for.
UZZI: AI can give you an answer to everything. So it does not only search quickly, but it also allows us to collect much more information simultaneously to get a comprehensive picture of the world around us and potentially fill in more blind spots at the same time.
PAVIN: In other words, AI can tap deep wells of information and distill the essential pieces.
It can also verify what we might already know. But even more than that, it can expand our thinking to include perspectives and information we might have missed or didn’t even know to consider.
Case in point: Kellogg’s Matt Groh has been doing research on how AI is being used in medical settings to diagnose skin diseases.
He’s seen how AI can complement a physician’s expertise by zeroing in on blind spots—and also the other way around.
GROH: We see a lot of examples where AI systems can actually beat physicians.
But then the thing is, there are often these blind spots. So we have current research that’s trying to map what a physician blind spot looks like, what an AI blind spot is.
And if you can combine the two, then you can make a complementary system that actually gets the best of both worlds.
PAVIN: For this complementary system to work, you need to be flexible—in a Goldilocks kind of way.
Not so flexible that you treat AI like some kind of fortune teller that always has the right answer. But also not so rigid that you default to dismissing what AI has to say.
Here’s Matt Groh again.
GROH: If the doctor just thinks they’re always right, then you’re going to lose all the upside of AI.
If the doctor is unsure themselves, then you’re going to get the upside of the AI, but you’re not going to get the instances where things went wrong.
PAVIN: To take a step back, at this answers level of the AI playbook, it’s all about approaching AI as a collaborator that can offer up information that widens your thinking.
More like an alternate point of view rather than a source of capital-T truth.
Engineering professor Julio Ottino says we don’t want to use cruise control when it comes to AI.
The quality of the answers we get from this technology will ultimately be determined by the quality of the questions we ask.
OTTINO: AI can be amazingly useful in expanding your space, but you have to ask the right questions.
If you outsource your thinking to AI, you’ll become increasingly dumber. But if you start using AI as your collaborator, I think that synergy will produce better and better possibilities.
PAVIN: We hear a lot about the value of teamwork. But what does it mean to collaborate with AI in the way Julio Ottino is describing?
Let’s get into the final level of our AI adoption playbook.
We’ve covered automation and answers. And now we’ve reached the tippy-top level of our pyramid.
GROH: The highest level where we think there’s the most value strategically is actually in this idea of advice.
PAVIN: At this advice level, it’s all about human–bot collaboration.
So instead of deploying AI to expedite a process or fill in information holes, now we’re asking AI to help us figure out how to solve a problem.
It’s all very meta.
UZZI: So what you’re looking for at this level is to keep in mind that if you want to do something creative, if you want to boost your innovativeness, you’ve got to ask the bot not for answers, but for how to do it better.
So that you can then introduce your own special sauce, your own colors.
And that integration is what makes collaboration most profound.
PAVIN: Let’s consider what this advice level could actually look like in the real world.
Matt Groh’s research lab has been studying how AI can help managers when it comes to communicating difficult news—news like telling someone they’ve been laid off or passed over for a promotion.
Or say you’re a manager and you have to tell your team about a decision that’s unpopular or controversial. Something you know they’re not going to like.
GROH: Leadership is all about making decisions. And when you’re making decisions, often they’re hard decisions, because not everyone agrees with you.
PAVIN: This is the bread-and-butter stuff leaders and managers deal with all the time.
It turns out there are tested ways of having tough but effective conversations, where the people on the receiving end of difficult news feel validated and heard.
GROH: When someone’s sharing a trouble, there are playbook-type things in the active-listening space that you should do.
And there are things that you shouldn’t do.
PAVIN: Matt Groh says AI can help us level up our communication skills.
GROH: AI coaching can give you personalized advice when you practice with someone.
And you might ask, “Where can I practice with someone?”
Well, actually, you could practice with an AI.
PAVIN: In other words, AI gives us a way to rehearse these tricky conversations before they happen, while at the same time getting feedback so we can improve.
Groh has tested this out in his lab with over a thousand people. He says the results are promising.
GROH: We found that this can significantly increase people’s ability to communicate such that others feel heard.
This is where I see one of the biggest opportunities, actually, in upskilling people in communication.
PAVIN: This advice level is about leaning into how AI can enhance, not replace, human creativity and ingenuity.
As we’ve heard in this episode, AI can be a powerful tool for automation, answers, and advice.
But it’s not a replacement for the fundamental traits that make us human.
GROH: Large language models today, the Geminis, Claudes, ChatGPTs of the world, are general-purpose problem solvers. They can’t solve everything, but they are general-purpose.
What I try to do in my classes is provoke surprise and awe. So people think, “Oh, I never thought that way.” By thinking, “I’ve never thought that way. Matt’s showing me these things. I can discover these things also for myself.”
PAVIN: Surprise, awe, and also curiosity are the fuel that can help us make the most of AI.
GROH: You don’t necessarily need to get a PhD to be curious.
I have a one-and-a-half-year-old, and let me tell you how curious she is. She’s the most curious human being I’ve ever seen. Curiosity is inherent to who we are.
When you think back to your inner child at five or six years old, curiosity often drove the kinds of professions you wanted to explore.
And sometimes as adults, we tamp down our curiosity. I think that’s actually a big problem in the workforce. People aren’t engaging their curiosity because they’re busy picking up from daycare or whatever else.
PAVIN: Curiosity can go by the wayside when our brains are treading water, trying to stay on top of our to-do lists.
GROH: But if you make time for curiosity, then you allow for skill-building. You allow for creativity.
Part of this automation aspect is allowing for that extra time so people can reorient to the world.
We need to give space for that reorientation, which comes from a drive. And that drive has to be curiosity.
PAVIN: Curiosity, these professors say, is at the heart of their AI playbook.
It’s about exploring terrain beyond what we already know so we can dream up new ideas and new ways of doing things.
AI, they say, can be a partner in helping us get there, step by step.
[CREDITS]
PAVIN: This episode of The Insightful Leader was written and mixed by Nancy Rosenbaum. It was produced and edited by Laura Pavin, Rob Mitchum, Fred Schmalz, Abraham Kim, Maja Kos, and Blake Goble. Special thanks to Matt Groh, Brian Uzzi, and Julio Ottino.
Want more The Insightful Leader episodes? You can find us on iTunes, Spotify, or our website, insight.kellogg.northwestern.edu. We’ll be back in a couple weeks with another episode of The Insightful Leader podcast.