Unleashing Predictive Power| Eric Siegel | 548

And welcome, welcome, welcome. This is Peter Winick. I'm the founder and CEO at Thought Leadership Leverage, and you're joining us on the podcast Leveraging Thought Leadership. Today, my guest is Eric Siegel. He's a leading consultant and former Columbia University professor who helps companies deploy machine learning. He's the founder of the long-running Machine Learning Week conference series. The instructor of the acclaimed online course, Machine Learning Leadership and Practice: End-to-End Mastery. He's the executive editor of the Machine Learning Times, and he wrote the bestselling book. I love the title on this one, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die. He's been on NPR, he's been featured in Wall Street Journal, and he's obviously lowering his standards to spend some time with me today. So, uh, welcome aboard, Darren. Thanks.

Thanks, Peter. It's great to be here.

Now, I don't want to know the answer of predicting, you know, who will buy, die, or whatever. I, I think sometimes it's careful what you wish for, but let, let's start with how did you, so the machine learning is the domain expertise obviously, and that's, everybody's talking machine learning today, right? If you put anything .ai at the end of your company, your valuation quadruples, it seems like, right? But tell me about your journey into thought leadership from both the academic side and the writing side and the speaking side. What has that been like? What does that look like?

Yeah, well, I've been in machine learning for more than 30 years and I've been an independent consultant for 20 and 20.5, because I'm a data scientist, we gotta go— Right.

They don't say half, right?

And at Columbia, I was teaching the graduate-level courses in AI and machine learning. I've kind of gotten to the point now where I'm sick and tired of machine learning not being deployed successfully nearly as often as it could. Now, there's a lot of stunning success stories. And the Big Five and a lot of other industry leaders do a great job. You know, internet search from Google is powered by machine learning, targeting of ads, all these applications. Financial services are way ahead of the curve in general, but the majority of new machine learning projects actually fail to deploy. There's sort of this disconnect, which is why I've now written the second new book, The AI Playbook, which is meant to help capitalize on not just do awesome number crunching, but get it deployed so that it actually improves business. Got it.

So I, I like the idea of a playbook and we're seeing more and more of these in the marketplace now because there's all, there's a lot of abstract, there's a lot of theory, there's a lot of whatever. But as a busy business person, it's like, okay, what do I do and how do I do it? Which is really how a playbook comes in.

Yeah.

For the most part. Right. And, and it's not that they don't, absolutely. They don't respect the theory, but they don't really have the time.

Yeah. Well, I'd say that, you know, In general, outside of data scientists, business professionals haven't quite come to understand that there does require a very specific customized process, practice, discipline, paradigm playbook, like the name of my book, AI Playbook, for running machine learning projects in particular. They're different. They're a different kind of animal than a lot of other kinds of projects. So that's what I'm setting forth in the book. I call it BizML, the business practice for machine learning. Projects and get everybody on the same page so there's an established practice that's standardized and is well known, not only to the data scientists, but to the business professionals, because they play just as key a role. What's necessary in this process is a deep collaboration with the business side between sort of the tech business, the quants, their, their stakeholder, the data scientist, and, and their client, whatever you want to call it. Those two sides need to come together with deep collaboration And that's only gonna happen and work if the business professionals ramp up on a little bit of concrete semi-technical knowledge, which is fascinating, totally accessible. And really here's what it comes down to. You need to understand for any machine learning project, what's predicted, how well, and what's done about it. So as you know, as you read the title of my first book, yeah. Who's gonna click, buy, lie, or die? Any outcome or behavior that may be valuable to drive organizational decisions. Also, negative outcomes, whether you're going to commit an act of fraud—

Sure.

—cancel as a subscriber, turn out to be a bad debtor, or move to a negative health outcome in the case of healthcare applications. Whatever it is, prediction's the holy grail for driving organizational decisions, micro decisions, per-individual decisions, and those are directly informed by the predictions. So that's the use case. That's the thing that defines how you're getting value from a machine learning project is, What's predicted and what's done about it? Who's going to cancel? Offer them a retention offer. Which transaction's most likely to be fraudulent? Hold it instead of allowing the transaction to go through.

Let me ask you this. When we look at your business model as a thought leader, you know, I think of it as a continuum, right? There's folks that will hire you as a consultant. Wow, you are one of the best people in the world to ask these questions of and tell me how to fix it, right? Here's My company at this point in time with this, with these variables and these issues, do your magic, you know, click your heels 3 times, whatever it is that you do, help me untangle the spaghetti and do it. That's one end of the continuum. And then the other end of the continuum is, hey, I'm a leader in this organization. And yeah, maybe we have some triage that we need to do to get out of, you know, solve a big problem that, you know, keep me up at night. But how do I take Eric's intellectual property, Eric's thought leadership, and move that into ways to develop the capabilities across tens, hundreds, maybe thousands of people in my organization that need to understand what to do with this, how to do with this in the, in the context of their role. Where do you fit on that?

Well, it actually kind of goes hand in hand with this sort of message I'm evangelizing with the book, which is you have to start with the business outcome. People too often are looking at AI, if you want to call it that, that's, that kind of oversells it, frankly. Or machine learning, this type of advanced analytics where you're learning from data to predict, and people kind of glamorize or fetishize the core technology because it is incredibly cool. And that's why I and most data scientists got into it in the first place. I mean, it's— there's no science that's cooler. It doesn't necessarily mean it's valuable. It's not valuable to you to deploy. It's potentially disruptive, but only if you disrupt with it. So actually planning from the get-go on the project, what's the out— what's the business outcome? What's the plan? What's the operational change that's going to be implemented and deployed? That's the carrot at the end of the stick. Step 1 of 6 that I outlined, 6 steps in the book called BizML, that practice. The first step is to define the last step, which is the deployment, exactly how you're gonna integrate it. 'Cause this is a business project first that happens to use machine learning as an important ingredient.

So, so for example, you know, it's like the, the, you know, whatever the, the 10 blind men and the elephant, right? So if I'm an enterprise salesperson, my thinking is, well, how do I use this to make my next sales encounter meeting richer, fuller, whatever? If I'm in supply and logistics, how do I do that more effectively? Right? So we're all looking at things even we're in the same entity from our perspective, our domain, our functional expertise, et cetera, and thinking, oh, this isn't really a hack, but like, tell me how to integrate this into X, whatever that is, accounting, finance, marketing, et cetera. So I like starting with the business piece. Then it's not about this is the cool, sexy thing that everybody's playing with. Yeah. Like, oh, I can do something better, quicker, cheaper, faster than I could before.

Right. So that's, I mean, that's the fast, the first, that's both the tactical and the strategic je ne sais quoi, right? I mean, you gotta start with that exact specific concrete way that's gonna be used, not just this fascination. And it's so tempting to get fascinated, especially the way AI is hyped today. Because the overall message is that we're headed towards AGI. Do you know what that is? Artificial general intelligence. That's the underlying message. And then literally the goal of, for example, OpenAI, computers capable of anything humans can do. They can run a Fortune 500 company. They could do anything. They're fully autonomous. You can onboard them and unleash them just the same as any human employee. I think that that's a wonderful fairy tale or ghost story depending on whether you're looking at the apocalyptic consequences of it. I think the apocalyptic consequences are totally real if you buy the science fiction premise, which is totally not real.

Right. But isn't it? I mean, I've seen that, I've read that, I've looked at that, whatever. But to me, once you sort of roll up your sleeves and start to play with it, it, like any other thing, it's just a tool right now. Yes. Is it eventually gonna take over the world and, you know, divert airline flights? I don't see that happening. No. What do I know? But I think it's like anything else. How do I use this to be a better X, whatever that is? Doc, I was medical professional, uh, you know, analyst, whatever. I think that's a better way to be thinking about it than the sci-fi of, you know, right now humans, humans run the world.

And if you want to, if you want to run the world, you use whatever tools are at disposal. Now, prediction is extremely valuable. It's unmatched, right? It's, it's the last remaining point of differentiation for improving large-scale operations. So, but it's a matter of exactly what operations. So what do you— now again, that goes back to that trio I mentioned earlier. What's predicted, how well, and what's done about it. And digging into some details, semi-technical. It's not the core rocket science, although you can learn some about that. It's how to leverage and capitalize on that core rocket science. Getting involved in that level of detail, that's the strategy. Yeah. Get concrete, you know, get specific.

If you're enjoying this episode of Leveraging Thought leadership, please make sure to subscribe. If you'd like to help spread the word about our podcast, please leave a 5-star review at ratethispodcast.com/LTL and share it with your friends. We're available on Apple Podcasts and on all major listening apps, as well as at thoughtleadershipleverage.com/podcast. So give me a sense— again, I want to move to sort of the business, your business as a thought leader, or just thought leadership in general from the business perspective. You've got the book, it's coming out, it's your second book. What have you noticed that is different about publishing now than it was a couple of years ago when you did your last book? And what are you doing to stay ahead of the curve from a getting the book out there in the right way standpoint?

Yeah, no, that's, I sure, you know, the first edition of my previous book was 11 years ago, Predictive Analytics, and then the Second edition, 2016, the updated edition, which is also the paperback. And the difference, it's more about trying to queue up for pre-orders now. Uh, it's a lot more podcasts than other kinds of media. Now, when you release a book, the way I take it is like a religious devotion. I mean, sure, there's a lot of potential career benefits for me here. I mean, my first book led to 110 paid keynote engagements, probably helped my conference. I don't have any measure of that. I've been running the conference. Since 2009. But as far as execution, both in writing the book so that it's accessible and sort of it's on a platter, right? You want to serve this, these concepts that are simple to you, but not to the rest of the world on a platter and bend over backwards to make it conducive and even entertaining and anecdotally driven. That's a, that's a really important—

So stay there for a minute. So, so when I look at a topic like your book, right? Yeah. On the one hand, I can say, Well, anybody in business today should be somewhat familiar with machine learning, right? And that, well, like that's really vague and whatever. Selfishly or logically, I would say if I were in your shoes, I'd be thinking, okay, I have limited resources. I can't compete with, you know, whatever Procter Gamble from a marketing standpoint. Who do I most want to get this book in the hands of to achieve selfishly my business outcomes? So how do you, how do you think about that?

Right. Well, the religious devotion side of me is just like, look, I'm excited about this message. I just want to get it out there as much as possible. So I can't claim to be, that's Eric the evangelist, but that's, that's Eric the evangelist. Yeah.

Yeah. I, I want to get in that sense of, I mean, yeah.

Who do I want to get in the hands of? The, the other 350 out of 500 Fortune 500 companies that really are just continually failing on these machine learning projects where you've got stakeholders and you've got data scientists and they just haven't bridged the gap yet. And this gap is totally bridgeable and people just haven't sat down. Look, it's not a soundbite. It's not even an hour presentation. It's a book's worth. It's a book's worth, but it's not a highly technical thing. It's semi-technical. You know, it's somewhat detailed. I said, what's predicted, how well, and what's done about it. Well, the how well is, is just arithmetic, but very particular arithmetic. You gotta get into this stuff. So I would love for these stakeholders, data scientists need this edification as well. Although for them, it'll seem like a very untechnical book. But there's a lot in there that they haven't thought through. But I want this book to be read by those, the rest of them, Fortune 500 stakeholders involved in, touch on, or, or somehow having stakes in the deployment where there's large-scale operations that can be improved with machine learning. Let's actually get it deployed. There's a lot of obstacles to that.

If you were to double-click on that and say, okay, 350 of the Fortune 500 fit into this category of those that, that aren't doing it well, or, you know, have, haven't gotten the fruits they want. And maybe there's 5 to 10 C-level execs involved in that decision at those, that company. Now you get laser focused and say, okay, there's like whatever that is, couple thousand people that really, really matter to me in terms of how do I get that book into those hands because they can unleash the power across, you know, at scale larger than a podcaster like I can. Right? So then the question is, well, how do you get to them? So I think there's, there's a couple of pieces when people are launching a book. There's sort of a hand-to-hand combat. Okay, well, do you name those names? Like, do you have the list of the 350 CTOs or CEOs or COOs of those companies? And in your interaction saying, hey, you know anybody at these places? And then there's the air cover doing something on a podcast, general media, the typical things one does, right? So what's, what's your approach there? What's your mix there? That's, that's a great question.

And I agree that, that maybe that it does come down to sort of those couple thousand people if we're talking specifically about me looking for, you know, executive consulting engagements. But more broadly, in terms of the mission of the book, it's much more than 2,000. It's like everybody, not everybody, a good, or maybe a majority of all business professionals stand to, to, you know, really need this. It's, I mean, we're learning algebra now, but I'm, but I'm getting much more.

What, what I'm trying to push on is I think there's, there's two Erics here. There's Eric Evangelist, because in a perfect world, I'd love everybody to read this. And I get that. I, and, and you're, you're there, right? But there's also the Eric that says, listen, I'm a business guy, right? Like part of the reason I wrote this book is to attract net new clients and speaking gigs and all that sort of stuff. And the people that I care most about isn't the, you know, the two brothers in Cleveland that own three pizza shops, right? Like they, like, I can't afford to spend as much time, energy, and effort there. Right? So that's, that's I think the struggle that a lot of authors have.

Yeah. No, I think you're asking just the right question in terms of career, uh, tactics now. Yeah. I'd say that maybe if it is 2,000, you know, maybe 25% of that list are actually CEOs, chief data officers, chief analytics officers. So I don't know. I mean, you're helping me to think more tactically. For example, I know a guy in his career as a consultant where he literally knows about half of those people. Right. So I should probably take him out to lunch more often. Right.

Well, and there's probably places where, well, and it's also probably places where not all speaking events are created equally, right? So some are, hey, pay me my fee and I'll jump on the plane and do the thing. But then there's, if there was an association, I don't know the space as well as you, obviously, that said, oh, I'm gonna have 100 of those 250 people in a room. Would you like to go there, Eric? You're not gonna do that or not do that because of the fee or not fee, because the exponential impact or ROI of getting in front of that room is, it could potentially be huge to you. And I think this is where thought leaders need to be thinking of, you know, what does that room look like of people I would kill to get in front of? Because it could be a game changer for me.

So, so are you charging right now for, for this advice you're giving me?

You, you will be getting a bill. You, you will be getting an invoice after the end of this. I've never had a, I tell anybody listening, right?

Because I've never had a public coaching session before, but it's, it's legit, man. This is good.

There you go. All right. As we start to wrap, Any thoughts, reactions? What would you share with someone like you 2, 3 years ago? They've had a book, it's had some success, blah, blah, blah. It's waning a little bit in terms of the, there's always, hmm, you know, they have a, I don't wanna say shelf life, that's too punny, but like, you know, they don't last forever. 5, 10 years is a great run to get off of a book and you're starting to think about the next one. What, what would you advise to Eric minus 3 years ago?

Well, that's a great question. I mean, the first question is really, should you write a second book? I mean, writing a book is arguably, it's kind of like moving to LA to be an actor, right? Like, what are the chances?

It's, um, what does that mean? You better learn how to wait tables along the way. Is that what that means?

Yeah. Yeah. Or, you know, it's sort of, it has to come largely from a place of passion because it doesn't necessarily make a lot of logical sense. The time you have to invest. In order to execute on the, both the writing and then the public publicity of it and just the COVID of it, the title, all of that. If you spend it on other business development, it, it might be time much better spent if you're just looking at the outcome from your own business. So I think there's a really big question there. That's not to say, hey, it's a dumb idea to write a book, but I think if you, if you go through the exercise of really answering that question, it's gonna help you strategize about why you're writing the book and how to leverage it.

Yeah, great stuff. Well, I appreciate your time today, Eric, and appreciate the effort and best of luck with the launch of the book. Thank you. Thanks, Peter.

Great talking to you.

To learn more about Thought Leadership Leverage, please visit our website at thoughtleadershipleverage.com. To reach me directly, feel free to email me at peter@thoughtleadershipleverage.com. And please subscribe to Leveraging Thought Leadership on iTunes or your favorite podcast app to get your weekly episode automatically.