Outcompeting in the Digital Age with Rodney Zemmel, Global Leader of Digital, McKinsey & Company
Willy Walker: Good afternoon, everyone, and it's great to have Rodney join me here on the Walker Webcast. First of all, I have to say I had your colleague Carolyn Dewar talk about CEO Excellence last year Rodney, and it was a real pleasure having Carolyn on to discuss that book, which your colleague wrote. I'm deeply thankful to our mutual friend Gary Pincus for putting the two of us in touch with one another. And I'm very much looking forward to diving into our discussion. A couple of things before I jump in.
The first is that next week's Walker webcast is with Jeffrey Wright, who beyond being one of my childhood friends, has just been nominated for the Academy Award for his role in “American Fiction.” And Jeff and I recorded the Walker Webcast last week. And I would just say to anyone listening today, if you want to hear a really engaging conversation on everything from families, to race, to how to become one of the leading actors in Hollywood, and also playing college lacrosse, which we both did. And we played lacrosse together in grade school as well as in high school. It's going to be a great discussion and we'll have that out next week.
The other thing Rodney is, I'm going to New York next week, and I'm interviewing Gloria Steinem, who will be on the Walker Webcast the following week. And you have a quote in your book from Gloria, which as you can imagine I'm reading your book, and all of a sudden there's a quote from Gloria Steinem, which A. I didn't expect, and B. is a great quote, which we'll talk about in just a moment. Let me dive in here with an intro Rodney, and then we'll go.
Rodney Zemmel is a global leader of McKinsey Digital, serving clients across a range of industries on growth strategies, performance improvement, and value creation by harnessing the power of data and analytics, digital culture, and capabilities, modernizing core technologies, and digital business building. McKinsey Digital now represents more than half of the firm's client work, with more than 7,000 colleagues across 100 offices specialized in digital and analytics. Rodney previously led McKinsey's healthcare practice, working with clients in pharmaceuticals, biotechnology, and healthcare services. He also led the firm's support for private equity clients and other companies in consumer-facing industries. Rodney is recognized as a thought leader, writing articles that appear in leading business and technical publications, including the Harvard Business Review and Nature Reviews. He is a co-author of the books Go Long: Why Long-Term Thinking is Your Best Short-Term Strategy and Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI. Rodney is a graduate of Cambridge University where he earned a PhD in molecular biology.
Rodney, I got to start with this: How does a doctor in molecular biology end up running digital transformation at McKinsey?
Rodney Zemmel: It's all DNA. So in the old days, it was DNA and molecules, and now it's DNA for digital and analytics. But it's the same stuff. But the first time I ever used the Internet was actually when I was a scientist in the early 90s, I was working on HIV in Cambridge. And the first time I ever accessed the internet was to compare different gene sequences. And back then, I had no idea that it was going to come to define this chapter of my professional career at McKinsey. But I think there's tools that started out in the scientific realm. And it took a long time to 20 years+. And then I just got absolutely transformational implications. The business is just really exciting. But the honest answer is when I joined McKinsey, I said I was going to stay for two years and then go get a real job. So I think I failed on the get-a-real job part.
Willy Walker: We always said that the Baker Scholars at Harvard Business School all went on to be McKinsey consultants because they were too smart to do anything else.
So before we dive into your book, which is so good, and as I was getting ready for this, Rodney, there's one line after another line that I will put a bunch of this into our show notes so that people can see some of these things that I picked out. But there are so many really important frameworks that you and your colleagues put into the book. But before we dive into that, I want to talk for a moment. It seems like everyone today is jumping to just the benefits of digital transformation and AI, and we seem to have forgotten about the downside. There are a lot of people sort of forgetting about the fact that we are creating a technology that could be weaponized against us. I had Mo Gawdat, who used to run Google X on the Walker Webcast. I think it was now two years ago, talking about his book Scary Smart. I know half of McKinsey's work is on digital transformation, and you are truly one of the world's great experts on AI. How much does that downside concern you, Rodney? And how many clients are talking to you about not only the good side and how it can benefit their business but also the downside of what this could actually unleash?
Rodney Zemmel: So the downsides are very real. But pause them a little bit. The first downside is just wasting money. Most business executives have been promised amazing things through technology for the past 20 years plus. And one of the reasons we wrote the book is we actually think about 70% of digital transformations are actually not hitting their financial objectives. So the first risk, which sounds prosaic but is meaningful, is just the risk of wasting your money and wasting your effort. So that's one whole category of risk. I think you are asking me more about societal risks and maybe even the existential risks, of do we create these all-powerful machines that we'll end up sort of taking over humanity and so on. I think there are a few different categories in that. So there's a set of risks that really are features of the new technology, like people who get to talk a lot about hallucinations and how AI can make things up and so on. That's true. It's designed that way. You can create versions of AI that hallucinate less and just sort of stick to the facts more. Those turn out to be less engaging, and people enjoy interacting with them less. So you have a choice to how creative to make something or how straight down the line to make something. And obviously when you're making something creative, there's more a chance that it strays into fiction rather than fact. But that's a choice that's within a company's power and a choice you can control by how many technical layers you put on top of that, and ultimately having humans in the loop checking it. That's sort of one thing. Another thing then, is how AI can introduce bias. An AI is trained on human data. All human data sets contain bias. Whether that's deliberate bias or accidental bias. In particular, anything that's been trained on all the crazy things that have been written on the internet are bound to have all kinds of biases in it. Again, that's something that can be designed around. One of the great things about AI is you can actually measure what's coming out and do AB testing, and you can track the bias in a way that's been much harder in any other technologies. So I actually think it lets you shine a spotlight on bias what might you address it, if it's used well?
The two areas where I think are the biggest areas of risk, which are related. One is of course around deep fake and around we may never know what's real again. And how easy it's going to be to create fake content, plagiarized content, offensive content, all of those things. There are many technical approaches to try to watermark and try to help people tell the difference between what's artificial, and what's real. That's going to be tough. I'm actually not usually optimistic that we are going to be able to easily tell what's AI-generated content versus what's real content, and we might go back to an era where we actually care a lot about the source and about trusted sources, rather than just taking what we see on face value. That's one category of risk. Another category of risk is the whole area of persuasion. These are amazingly powerful persuasion technologies. Maybe we'll talk about that a bit more when we get into some examples. And of course, you can persuade people for good or persuade people for business purposes, but you can persuade people for bad as well. And that's an area which I think has got, which I think could be quite challenging for society. And I think it's healthy that businesses and regulators are beginning to pay real attention to that area.
Willy Walker: One quick aside before I ask you a question. When I was reading the book, we were talking about people in photographs and needing access to their photo libraries. It made me think that maybe Getty Images is worth a whole lot more than people thought because people want to be able to own content rather than going out and having to constantly pay licensing fees for it. But anyway, that's a whole different tangent. But it was interesting, as you were talking about some companies that are creating their photo databases that they did not continue to tap back to someone like Getty Images.
For a moment, Rodney just explained the difference between AI, machine, learning, and deep learning. For the people who are listening in, I think people hear either AI or machine learning or deep learning, and they don't quite understand the difference between them. And I jumped right to AI, and you gave a great answer on the potential concerns there. But just frame that for listeners if you would.
Rodney Zemmel: Yeah, and to be honest, they're all different flavors of the same ice cream. For the average person, the distinctions are not hugely meaningful. I mean, the two categories I would put it in until sort of December of 2022, most of what we were talking about within the broad category of artificial intelligence of AI and what AI means is a system that gets better when you give it more data. So an automatic gearbox in a car was an amazing improvement over a manual gearbox, but you sort of do that once that doesn't get any better the more you use it. The advantage of AI is it learns from how it's being used and it gets better over time. That, in essence, is AI, deep learning, machine learning, and so on. A different flavor of how you do that. The big leap that happened in late ‘22, and early ‘23, with the large language models is the move from sort of regular AI to generative AI. Generative AI is what ChatGPT and all these sorts of language models can make and can create. And in some business terms, the relevance is in the past with the other forms of AI. For everything you were doing, you needed to build a new model. So if you want to look at customer churn, that's a model. You want to look at the supply chain, that's a model. If you want to get pricing, that's a model. What the new generative AI approaches do is basically like a Swiss Army knife for AI. It's one model, one foundation model that not with no tweaking, but with fairly minimal tweaking, you can apply against all kinds of different purposes, like a whole range of different things. So what it does is it massively lowers the cost to deploy AI, and it massively democratizes it. So many more people can use it. So that's sort of the big revolution. But even that two years from now, three years and I'm not sure we're going to be talking about generative AI because there'll be some other flavor of AI, and these things will all sort of work together in a way that the average user probably doesn't even need to understand the differences between.
Willy Walker: So as we jump into your book, just a couple of things on Rewired before I start throwing some questions to you. Number three on the New York Times bestseller list. Congratulations. And when you all started writing it, 90% of companies had a digital plan, but only about 20% were actually creating value by it. And so you also state that your book is not a coffee table book. It's a real roll-up-your-sleeves book. I will tell you that trying to listen to your book on an audiotape is very challenging. Anyone on this webcast listening needs to have the physical book because there are so many charts and diagrams inside of it to follow along with what the team is telling you to think about doing. It's very helpful to see the actual physical book.
But one of the other things, as I approached the book Rodney, that I think is so important for people to keep in mind, is that you state it's still day one. In other words, I think a lot of people look at digital transformation and they think about AI and they say, “Oh my God, we don't even know where to start.” And your book is an incredible manual for how to start. But then second of all, you reiterate the fact that you don't think that the train has left the station and is already halfway to Liverpool. If you're living in London, where you are, it's still in the station, you can hop on.
Rodney Zemmel: So, I live in New York these days.
Willy Walker: You live in New York. London and Liverpool are better examples than New York and Philadelphia, but the same deal.
Rodney Zemmel: It's so good. I'm gonna correct you. You’re going to think I’m being petty. One other thing as well, we're number three on the Wall Street Journal bestseller list.
The New York Times didn't include us in the rankings because they said the book is a manual and they don't review manuals. So I took that as a compliment because we set out to write something that truly is going to be used like a manual. As you say, it's not the easiest audiobook.
Willy Walker: It might be that they have an AI algorithm that doesn't get that right. And in one of the interviews that you did Rodney, you talked about hallucinations, and where if you ask AI who's Tom Cruise's mother, it will tell you the name of Tom Cruise's mother instantaneously. If you then go and say, who is the son of whatever Tom Cruise's mom's name is, it won't tell you. And so it's just interesting about how these models are created and where they go.
Rodney Zemmel: That's exactly right. The newer versions may fix that, but that was an interesting quirk from earlier last year. Back on your day one point. I don't like the term digital transformation because it implies an event. It's like everybody works hard and you do this for 18 months and then you're done. I think the reality is, this is what many executives, maybe even most executives, are going to be doing the rest of their career. This is the new way of doing business. And we're still in the very early innings of it in terms of its real impact and how it's going to change what business leaders do. I actually think if we were having this conversation ten years from now, we wouldn't be talking about business leaders and technology leaders. Every business leader is going to need a service technology leader. Those things are going to be fully integrated. And I think that's the journey rather than a one-time transformation.
Willy Walker: I think that's one of the most fascinating pieces of the book. While you dive deep into technology, on all of your rankings as it relates to what really needs to be done on any of the lists you look at, technology and data are well down. This is a talent and process reorganization inside of corporations that needs executive leadership, which, as CEO of a pretty scale company, is a real eye-opener in the sense that we spend a lot of money on technology. But I thought it all kind of started from technology. It reminds me, Rodney. There's an Instagram piece that's running around of Steve Jobs being grilled. And I think it was 2004, 2005 by somebody at one of the Apple conferences where this guy gets up and says to him, “First of all, what you've been doing for the past couple of years. And second of all, like, you guys aren't creating any good code.” And Steve stops and instead of responding poorly to the gentlemen and telling him they know I'm on the stage and you're in the audience and there's a reason for that. He sat there and he said, “You know, we could go and create a lot of great technology, but if we don't look at the customer and what the customer needs are, all that sort of technology-oriented work is the only technology for technology's sake. And unless you look at what the customer needs and create great products for the customer, all the technology is sort of worthless.” And it's quite something is the greatest developer of the designer of technology in our lifetime that he would say that. And then your book is very much reiterating that it all comes from the top and that creating this digital roadmap is sort of a step one for any company. Why don't you dive into that?
Rodney Zemmel: So it's not not about technology, but it doesn't start with technology and it certainly doesn't end with technology. And what we saw if we looked at the research as we went into the book is we looked at about, our Math is pretty much every large company on the planet now has something called a digital or AI transformation. Only about 30% of those hit their revenue targets, about 25% hit that cost target and not hitting their financial goals. We took about 200 companies that were hitting those financial goals and trying to sort of reverse engineer what the recipe is. And that's what you're referring to. Technology is in that recipe. But it's point four in that recipe. It's not point one in that recipe. And that's for a reason. And we saw as many companies that overspend on technology. Companies said, “But first we're going to go build the amazing perfect data lake and get everything all super well structured. And then, we'll figure out all the amazing things we can do with it down the road.” That can work. But that is a slow and expensive way to get to impact, rather than starting with the business problem and working back.
So I think that the Steve Jobs or the Steve Jobs era insights was it's about the customer. You've got to think about how you design a process around the customer, work back from that, and make the technology serve that end. And that's correct. What we're trying to do in the book, though, is go one further and say actually it's about the business impact overall. And the most important thing that any company leadership team, or organization leadership team needs, if they're trying to drive a digital transformation is to start with a prioritized roadmap of where the value really is. So take any company and break it up into 10 or 20 domains. There's going to be business units, functions, whatever you want, but like 10 or 20 pieces. Which one or two pieces are going to be the ones that are going to drive the most value if you can transform and reimagine them with technology? Start there. Assign a real business value to that. And I'll come back to, what were some of the metrics of that in a moment. But assign a real business value to that. And then think about what's the individual technology use case but think about how you completely transform that domain. It's very interesting if you look at sort of generative AI right now, all the companies that in 2023 said, let's prioritize the bottom-up use cases. I'll go build a copilot, I'll go build whatever. They're still sort of in the pilot phase instead of saying, “Okay, I'm going to take customer service.” I'm not just going to build a copilot, but I'm going to say, how do I improve that by 50%? That's going to take a copilot, that's going to take a scheduling approach, that's going to take three or four other things. And how do we do those things together with a clear financial objective? They're the ones who end up driving real value.
Willy Walker: It was interesting in the book, you point out that those companies that put real financial metrics behind exactly the process you just talked about had much better outcomes. And that doesn't surprise me. And yet, at the same time, I think many people look at tech initiatives and they say, “Hey, don't give me a timeline because it's going to take time.” And then also I can't quantify what the outcome is going to be so to some degree, don't hold my feet to the fire. But it was very clear that those companies that said, we're going to improve EBITDA by 15 to 20% in your study got much better outcomes.
Rodney Zemmel: So that's exactly right. So first of all, if you don't set a financial goal, you're not going to hit a financial goal. You don't need to buy our book to learn that. But the specific finding is exactly as you just said, if you set a goal of 15 to 20% EBITDA in an area, your chance of hitting at least 80% of that goal turned out to be way higher than if you set a goal with like two or 3% in that area. So that was a little odd when we first sort of found that. And we think the reason is, the price needs to be big enough. This stuff is hard, but it needs real cross-functional work and it needs sustained involvement from the CEO. It needs real resource commitments. It needs real talent upskilling or new talent. You're only going to do that if the price is big enough. So companies that set that big price at 15 to 20% EBITDA in the area that we're focusing on were much more likely to hit the price than those who said, “Let's do something small first and then sort of grow to success.” And you're right that I think there was this sort of sense that digital is different. It's hard to measure and it's hard to manage it the same way. And there were so many aspects of it that were different. But we think you can have the same rigor around it as you would around any cost or any sales simulation program.
Willy Walker: I want to dive into pods and all that in two to five. And we'll get into some more specifics in a second. But I've listened to you talk sort of at 30,000 ft, Rodney, about takers, shapers, and makers. And I thought that that framework was extremely good to give. If you're a CEO listening in today, are you going to be a taker? Are you going to be a shaper or are you going to be a maker? Could you dive in and segment the market into that? Because particularly on the taker piece, I know where you're going to go with the answer to this, and I think a lot of people miss that there's someone else who's going to do the taker for them.
Rodney Zemmel: So, I'll illustrate that maybe with generative AI again. So around this time last year, lots of people were launching their first pilots with generative AI. And a lot of companies said, look, we don't want to do anything that touches a customer because that's risky. Let's find some safer places to start. So as an example, people said, okay, let's go to HR. We'll use this to generate job descriptions. We'll use this to go scrape LinkedIn. Or maybe we'll do finance. We'll go generate summaries of financial reports and so on. And we ended up, what we found was that's fine. You can do that. It's not that hard. The reality is it's not a great use of most companies' innovation brain cells because there's a very effective software market that's going to do that for you. That's going to get built into the workday or SAP or whoever it is that you use. So anything in your company that is an essential function, but it's not actually driving differentiation. So an enabling function, you're both companies, like finance, like HR, and so on that we would say look for off the shelf tools. Just go wait for the vendors that cut across many industries. Just go wait for them to develop something and then go bring it in that all the way. All the way at the other end of the spectrum, there are some companies we're going to go and build their own large language model. We're really going to go and create their own very specialized applications. That's expensive, that's hard. That's going to be hard to maintain. We think there's going to be only a very small number of companies that we're going to want to do that and it's going to make economic sense. Maybe you've got very significant concerns around privacy or very special needs. For an average company in areas that are competitively differentiating. You're going to want to be a shaper. So pick an area that really drives competitive advantage for you. So for most companies that might be something around customer, or on supply chain or depending on what you do, it could be anything. But what part of your business is you think you're going to derive some competitive advantage from? And if all you're going to do in those areas is go and buy some industry solution, you're only going to get industry average results. If it's going to be a competitive advantage, you need to make something that's more tailored for you, but you don't need to make it from the ground up. You can make it from pre-existing components That's where the shaper concept comes in. So it's not differentiating but you need it, go be a taker. If it's going to drive differentiation, be a shaper. And in some rare cases for some companies, you may want to go all the way to being a maker.
Willy Walker: You talk about two makers of being Bloomberg, and Saudi Aramco, who are both in the process of building their own models and their databases. But I've heard you talk about the six large language models and the fact that they all are going back and deriving from the same Google data to start with. And therefore don't get too caught up in picking whether you're going to work with ChatGPT or whether you're going to work with Google or any of the other ones out there, that they're all essentially the same. I thought that was super interesting, Rodney. But you then add to that what people ought to think about, about who they're going to choose to partner with as it relates to architecture. Why don't you talk about that?
Rodney Zemmel: So the history of how these large language models came about is fascinating. Or maybe it's only fascinating to a nerd like me, but I find it fascinating. So if you look at it like there's a several hundred-year history of people trying to imitate human language with machines. And there were broadly two schools of thought. There was a structuralist school of thought that says there are hidden patterns in language. We've got to figure out what the hidden patterns are, and then we can imitate them. And then there was a more probability or statistics-oriented school that said, “All you need to know is just understand the probability that one word will follow another word.” And for like 100 years, there's been this debate in linguistics about how language works. What these large language models show is that you can get a very long way with the probabilistic version. It's just a prediction machine for what word will come after the word before. But the breakthrough was the computing power that said, you're not just taking word one and trying to predict word two, but you're predicting word two based on what was, word minus three, minus four, much further back in the sentence. And it was only through the availability of massive computing power that we were able to do these giant sort of statistical correlations and come up with these prediction machines for what word was going to come next. It's a token. It's a fraction of a word. But the same concept. This was laid out in a paper from some Google DeepMind Deep Brain called the Google Researchers in 2017. And they were trying to solve a translation problem. They were trying to show, How do you predict word order between different languages? And I think at the time they published that paper, they thought they were doing it just to service the language translation, like solving this linguistics puzzle. And I don't think they quite realized at the time just how relevant this was going to be to predicting everything like math, image, music, whatever, and the whole sort of revolution that we're now in. But because that piece of work was published and open source and so on, lots of other people have taken that and done very similar things. And what we see, if you take an institute like Stanford, artificial intelligence, human artificial intelligence engine, they benchmark these models and what they see is all the models essentially perform incredibly similar. And then someone will come out with a new version, and then they'll pretty quickly, asymptotically all sort of converge on that again. So we're seeing this sort of model arms race. But they're all pretty similar. I don't want to diminish the differences. There are certain performance differences for certain applications. But there's a cluster of them that all behave similarly. So if it's not the prediction power that's different between the models, you're then left with a set of choices and you have different kinds of choices on which one you want to use. Some people are betting on open source and saying, I don't want to have one company Black Box. I'd rather have a more open approach. Other people say open source is risky. And we'll never perform as well. Some people are saying, I want to use a model that exists in the Cloud, I'll send my data into the Cloud, and so on, because that's going to be better performance and lower cost and so on. Other people are saying, no, I'd rather keep my data where I have it, and therefore I want a model that will run on my environment. So I think it's those sorts of architecture choices and ultimately cost choices as well that are going to drive choices of the model. As much as, which one can perform better.
Willy Walker: When you talk about that architecture, you sort of bury in the middle of the book a memo from Jeff Bezos to everyone at Amazon as it relates to APIs. Will you dive into that for a second, because I found that memo to be eye-opening and kind of mind-spinning as it relates to that seminal memo that he sent out to the team at Amazon.
Rodney Zemmel: So thanks for spotting that. And it's a pretty famous story in the history of Amazon, in the history of technology. So one of the hard problems in any big organization is how do you get teams to work autonomously, without drowning in all the interdependencies. And that's a problem in any kind of business problem. Anyone who's done a big integration or anyone who is rolling out a big new org and change problem, how does the work of one team depend on the work of another team? Like in technology, that problem is really acute because everybody's interacting with the same core systems. And how do you actually move ahead in a way that you don't create horrible dependencies, what everybody else wants to do to come after you? So the technical solution to that is called an API. An Application Programming Interface that just allows different pieces of technology to work with each other. And what Jeff famously did at Amazon, I only know this not even second hand, probably third hand, is there was a moment in time where they were worried that all these interdependencies were going to slow the place down. So they put down hard edit, saying everything needs to have an API. Like every team, your output that you are ultimately responsible for is an API to interact with all the other teams. And if it can't be turned into an API, don't do it, because that's the only way we're going to scale. And that drove a big culture change that was hugely effective for them. And I think in many other industries that are maybe a little bit less tech intense, they're still sort of learning that lesson. But it's going to be incredibly important how you build real scalability and what you're trying to do in technology.
Willy Walker: You state in the book that this Roadmap and the transformation roadmap has to start from the very top that any CEO who thinks that he or she can delegate this to somebody else, whether it's the Chief Information Officer or somebody else on their executive team, is missing the boat. You also talk about the 20 hours that are required for any member of the team that is going out to try and come up with this digital roadmap needs to invest. And I literally talked about it yesterday with my executive team, Rodney. And it was fun because we sat there and someone's like 20 hours isn't nearly enough. And someone else says I don't have 20 hours. I think that the 20-hour framework is extremely helpful because it does make people sort of say, “Well, how do I do it?” A: I read your book, and B: after doing that, you put forth go out and look at who's well ahead of you on the digital transformation, try and benchmark against others, look at business cases, etc. One of the things that you talk about and a term you use is exothermal and I thought that was such an interesting way of putting it. So talk for a moment about the exothermal digital transformation of companies.
Rodney Zemmel: You're right on the 20 hours. The first four hours reading the book.
Willy Walker: I wish it was only four hours. And you read faster than I do.
Rodney Zemmel: So there's a few things in that. Maybe, first of all, I'll talk about the learning part, and then I'll talk about the exothermic. So, one of the fun findings, one of the positive findings when we did our research is, that it's actually not about going out and sort of hiring all the Silicon Valley cool kids. Yes, you will need some external talent. But the companies that went out and got the digital natives, whether it was Silicon Valley or London or Berlin or whatever, that turned out to be a really good way to change the company dress code, but not a great way to actually drive sustained business impact. It's just hard to make that stick unless you also really invest in reskilling and upskilling the leadership team and the front line. So that's the 20 hours the senior leaders can be more in other parts of the company. But how people really learn about the power of what digital and AI can do and what we find is a couple of things. First of all, it's hugely more valuable to actually start with the business context and then learn technology than the other way around. There's a great quote from a steel company that we work with, the CEO likes to say, “It's easier to teach a metallurgist how to be a data scientist than to teach a data scientist how to be a metallurgist.” So it's about taking people who know something in your company context and upskilling them. How they learn then, can totally depend on the individual in that role and so on. We actually by going to see what other companies are doing, I like going on to what we call a go and see is a really helpful way to do them, and not just the tech companies, because if you go and see what's Microsoft or Google or Amazon or doing, you'll be blown away by that. But it can be quite hard to apply that to what an average company can do. So instead, go to someone who's like a traditional company that maybe is a few years ahead of you on the digital journey, maybe in a different industry. That tends to be a really good way to learn.
That gets a bit to the exothermic idea. I'm from a nerdy science background, but the idea is for the reaction to keep going. It's got to generate heat, not absorb heat. So what does that mean? First of all, you need to do things that are actually exciting and fun for the people who are participating in it, like going to see what other companies are doing and so on. But also then in terms of where you start, if you're trying to say, “We're going to work really hard and that's the two years and so on, the amazing things will come out at the end.” Very few companies have the patience to sustain that. Instead, it's about breaking it up into three-month or six-month chunks and saying, “What's the amazing success that we're going to see at the end of that period that we can see in the P&L? That's going to give us the energy to get more people excited,” and then go to the next thing and go to the next thing.
Willy Walker: So it's so interesting to hear you talk about the recruiting strategy. In the book, you talk about both how McKinsey has changed its recruiting strategy and how you all believe that AI is going to help your associates, and consultants inside of McKinsey on real creativity, technical skills, and then overall leadership. I think that the concept that AI is going to help leadership is fascinating to me. It's like you wouldn't think that AI is going to be a big enabler of leadership. And then at the same time, anyone who wants to be in a leadership role, as you said, ten years from now, you're not going to look at a CEO and say, “Is he or she good at technology.” It's all going to be integrated. And those people who understand the technological revolution that we're going through right now are going to be the leaders. But you also, when you look at the banking sector, Rodney, those companies that have really been able to digitally transform in the banking sector, we're really good at the soft skills. They were good at recruiting talent, finding career paths, and making those teams agile. Talk for a moment about how important those soft skills are. Because I think, if someone rewound the tape on the last 40 minutes that we're talking, they'd sit there and say, “This is all about X's and O's. It's all about big databases. It's all about investment in technology.” And what you really point out in the book is, unless you get the people-side of it and the soft side of it really nailed, all the rest, doesn't really matter.
Rodney Zemmel: I'll tell the banking story that you're referring to. It's a little bit of what gave us the title for the book. And banking is an industry that was among the first to really digitize. And it's further along than most. And I was invited to a consumer banking roundtable that a few of my colleagues were holding. And we have the CEOs of a number of the big banks with consumer divisions were there. I was a couple of minutes into sort of talking about how AI was going to change the world and the bold AI future and so on. And one of them interrupts me and says, “You know what? This sounds great, but my consumer app is the same as his consumer app. My private wealth app is the same as her private wealth app, and we all now have thousands of people doing digital that we didn't use them. Like what's going on here? Is this like some trick invented by the consultants? That's just adding a cost for all of us. And we're all having to run faster to stay in the same place?” So, as you can imagine, this became a pretty lively back-and-forth in this roundtable. Yea, we're just adding cost, but what's going on? And we said was, “Hold on. Let us actually look at the data? Let us benchmark what you're really doing in digital. And then who is and who isn't making money on it?” And we looked at about 50 banks, I think about a dozen in the room. But we got data from the 50 more through a benchmarking service that we had. And what we saw was at a superficial level, what he was saying was right. If you looked at the actual numbers, they were the same. You could not meaningfully tell the difference. Maybe somebody had a feature a few months ahead of someone else, or the design was different, but they're all doing the same stuff. But if you looked and said were they making money, you could see that about the same ratio held as 25% of them were really making money. From what you saw, you could see that in terms of the growth of digital channels, the value per customer, how they were actually increasing labor productivity, in the branches, and physical channels. And ultimately, you could trace it all the way back to the return on equity. About a quarter of them were making money. So then if you look at the ones who are making money and you say that their apps are the same, what are they doing differently? And what we saw in our benchmarking in our survey work was exactly as you just said, Willy, it's the soft stuff. And what we saw actually, there were three questions in our survey that were most predictive of who was making money on that transformation journey. First of all was, how well business and technology work together. And the ones that really were working well together and really working in agile, that made a huge difference. The second one was, how good is your tech talent career model? And the companies that have created real career ladders for tech talent, where the tech people did not feel that they were sort of glorified IT helpdesk, but actually they were sort of co-equal with the businesspeople. That was a huge driver of success. And then the third one was the funding mechanism. And if you were following a typical annual funding plan that was not great, and if instead, you were taking a sort of a more VC, maybe three year funding with regular stage gates approach. That ended up being the path to success. Maybe banking is further along than others because they were all pretty similar in the technical components. But the importance of those softer components and the extent to which, is not about like we have a digital veneer, but actually have you rewired under the hood. It’s what ultimately actually gave us the title for the book.
Willy Walker: It's absolutely fascinating. And it's so insightful as it relates to… anyone who can build an app. And what you're basically saying is here through to what the app actually does and how it's integrated into your core business. And also the focus on the front end. That was one of the other things that I totally took out of your book, Rodney was just that those people who focus on the HR function or the finance function, as much as those are important, are missing the real capability here of interfacing with the client. It's really all front-end. And obviously AI and these tools are being used in customer service with chatbots and quick response times to the frequently asked questions and all that great stuff. But it really is taking the risk of rethinking your business in a digital world. If you sat around the table yesterday with my executive team and sort of said, okay, how is someone going to borrow from Walker & Dunlop five years from now or ten years from now? It takes a lot of dreaming. One of the quotes I mentioned at the beginning, your Gloria Steinem quote that you put in the book, says, “Dreaming is part of planning. “
And then unless you're willing to sit around at the table and throw some stuff against the wall that just sort of says, well they're going to borrow seamlessly with one click on their iPhone. Given the amount of time and effort that goes into making a $25 million mortgage on an apartment building, that's truly dreaming. But it's going to be at a theater near you probably sooner than we would expect. And so I thought it was so helpful that you are really in the book, there's so much technical data in there, you also underscore the need for the human resource, human allocation, and career pathing. And that point about tech resources, I think anyone who's listening in on this, to understand the importance of those tech resources and tech professionals to your company. I know the names of every senior banker at Walker & Dunlop. I engage with him or her consistently to go out and meet with clients. I don’t, to my own criticism, know the names of our top tech talent inside of work or not. I know some, but not as many as I need to. And to me, in reading your book, I sort of said, you got to rewire the way you're thinking about talent inside of W&D. And if we're going to make this transformation, I need to be spending not only time with them but rewarding them and having them grow in a career path that is commensurate with where our bankers and brokers are.
Rodney Zemmel: That's exactly right. And what we've seen is that companies who do that right, who really invest in that tech talent career path can attract and retain amazing tech talent no matter what their industry or their location is. There's a company we talk about in the book, Freeport-McMoRan.
Willy Walker: That's a good story.
Rodney Zemmel: So in Baghdad, Arizona. It's probably only a two-hour flight from Silicon Valley, but it's probably spiritually pretty far from Silicon Valley. They built an amazing tech and data team right in Baghdad, Arizona, and at their other mining sites, really through doing exciting things, from getting people excited about the mission and the impact that they can have, and being very thoughtful about creating this tech talent ladder that lets them bring in tech people who really have growth opportunities, leadership visibility, and real sort of equality with the top business leaders in the company.
Willy Walker: But double-click on that for a second, Rodney, because I think it's so important. That’s a mining company. It was faced with a reinvestment of $200 million in the mine. And most of you would sit there and say technology, hold on a second. You either write the check to go in and dig a little bit deeper or you pull out and close up the mine and go home. Talk for a moment about how they transform those 40 decision points to basically not only change the yield of that plant but change the way that the entire mining industry looks at plant utilization.
Rodney Zemmel: It's a great story. So by the way, when I'm mentioning a company name, that's because they've talked about it publicly. Normally I wouldn’t talk about our clients. So I'll tell the story as they tell it. So, you take that Baghdad mine. Their business is relatively simple. You take all out of the ground at about 1%. You put it through a copper concentrator, it comes out of the other end at about 20%. I live in New York City. There's only so much I know about mining, but I think I'm roughly right. Because they've been mining in Baghdad for nearly 100 years, the quality of that all seems to have gone down, and it's down to about half a percent. So they were looking at that and saying, “Okay, so to get the same output we need to go build another copper concentrator.” That's a $200 million CapEx piece of equipment. And then we have a picture of it in the book. It's an enormous thing. And they were looking at doing that. And at the time they were looking at doing that. This was pre-COVID. The global copper markets were wobbling a little bit. And they were thinking, is there really going to be enough ROI on a $200 million CapEx investment? So instead, the head of the site and his technology lead said, give us a year. Let us actually figure out there's another way to get there. Instead of just going, putting this massive CapEx investment. And what they did was they got the mining engineers at the site and they got some external help, but really pretty focused amounts. And they really said, let's actually try and build a mathematical model, a digital twin of that copper concentrator. And let's look at the 42 different decisions that go into how a copper concentrator works. And let's see if we've optimized them. I'd like to model it out. It was everything from the temperature, the throughput, to how many trucks wait in line, to how big the safety stock should be, and so on. And sure enough, when they model that out and through a series of, I think they were doing six-week or eight-week sprints where they would try a model test it in real life, keep trying and testing, and so on. What they saw is the way they were running it, which was based on the dozens of years of experience of senior operators, wasn't actually the best way to run it. So for example, the safety stocks had a rule of thumb. They said the safety stock couldn't be lower than a certain level or it would risk the plant running out. Turns out that level was way too high. You could run the safety stock much lower. On the number of trucks waiting in line, they didn't want more than one or two trucks waiting in line. They thought it was wasteful. The optimal number turned out to be three or four. Temperature, throughput, all these decisions, but no one decision any human could have spotted. But when you look at the system together, there's a different mathematical way to optimize it. That for them, got to aggregate an 11% output improvement or yield improvement. By the way, a better safety record as well, because they had, I think, fewer changes in the handoff during the process. Over time, 11% I think grew to 15%. And that was actually way more than enough to offset the need to go and build giant CapEx. They did this in enough places over time that the company eventually publicly said, “We’re actually going to tilt our strategy away from giant holes in the ground and new CapEx and towards trying to get more value out of these existing assets with massive success.” The rest of the industry saw that and then tried to follow. But with the 18-month lead that they had, they've been very successful at continuing to sort of move them that learning curve and stay ahead and really use digital twins because AI optimizers to really just help transform the delivery of copper in this environment, as far from Silicon Valley as you can imagine.
Willy Walker: So you talk about the impact that had on Freeport. Talk for a moment about the database that McKinsey has built that allows all of the consultants worldwide to go in and do a query into your own database. And I think one of the most interesting things about how you all are using it is not that it's a great resource library if you will, that says that Rodney Zemmel used to talk about mid-stage leadership for CEOs and not the beginning at the end of tenure, which you can go and find pretty quickly. And I appreciate that you're smiling that I went and did that research on you. But it also defines who inside the enterprise is the sort of current expert.
I'll just give you one quick example, Rodney, that we are just dealing with. We’re going after a piece of business. We were bidding against eight competitor firms. I was pulled into the pitch call, we did the pitch call, and I was the only CEO of the eight firms that showed up on the call. So that helped differentiate a little bit. And then after the call, I put an email out to my executive team saying, “Hey, we just had this call and this company is selling a property and we're on this and that.” And then back from that came a note from one of my senior executives saying, “This person inside of Walker & Dunlop knows this person inside of that big, large client really well. And you ought to connect this team pitching to go in and make sure that they know that we have great connectivity.” And what came out of that was that if you'd gone into Salesforce at Walker & Dunlop, Salesforce wouldn't tell you that. Salesforce wouldn't say, this person knows that person well. You could set up something that says, who has the strongest relationship and it would come back, relationship lead, or whatever else. But that type of anecdotal feedback from a senior executive only came from that senior executive knowing the relationships. And so we sat around yesterday and talked about how you create a database that allows for that connecting the dots that doesn't require a human to do it. An executive who gets an email from me, which would be the only way the team would get that kind of feedback, but actually instill it into the database. And I was interested that you all have created this feature that says, if you want to know the answer to this question, call Rodney.
Rodney Zemmel: That's right. So what you're talking about is that we have this thing we built called Lily inside McKinsey, but our original working name for it was Jarvis. But it turned out Marvel owned the copyright to that, so we couldn't use it. So, we call it Lily. And what it is, it's basically generative AI for concision. And concision is concise plus precision. And it lets you ask and you can already say, what's the global market for apples or something? You can get the answer in milliseconds on your phone. But that's not that useful. It only becomes useful when you can combine that with your own company's proprietary data in a way that has the right one-way valves on it so it doesn't leak to the outside world, but also in a way that understands context. It knows the difference between something from a real thought leader inside your organization as part of a thoughtful discussion or presentation versus something it's pulling from some random email. So that's what we set out to build. And we went actually searches across 42 different internal databases at McKinsey, and about 150,000 hours sanitized external interviews. A sanitized staffing log, who's worked on what topic, and so on won't necessarily know what client. So you can ask a question like, How do you do a diagnostic of retail store operations? Something that many of our teams might be working on at any one day. And you put that question in and out will come, here's the recommended approach. My best internal knowledge documents. Here are four or five names of internal experts who know this topic well based on the work they've done. Click here and sort of slide across, here's what the external world has published on this topic. So you can see what Harvard Business Review says about it, or what a Google search would say or something. But then, hopefully, people would draw on the McKinsey knowledge. And then we're tracking to see if people actually call the experts. Because the written text version is never going to give you the whole story. So we want teams to use this as an accelerator, but not such an accelerator that it's a shortcut. And then they don't bother actually going and checking with the real expert when they need to. And it's working incredibly well for us. And I'm thrilled that for all the talk, are we going to see white-collar labor decimated and massive replacement of people by machines and so on. The very first effect is going to be the most boring and tedious parts of people's jobs going away and being automated. And what we're finding is teams are loving this and using it to really accelerate their work and get further and help clients better sort of further faster, rather than having to go through some of the more mundane, search and phone a friend and so on, it's making a huge difference.
Willy Walker: I think it's so fascinating. Do you think about Freeport a mining Company and how they implemented it? You think about a consulting firm like McKinsey and how you all have implemented, I just think.
Rodney Zemmel: I'll give you another. The few companies I've heard of doing this work or building versions of this in a few other places. And there's a Taiwanese electronics manufacturer that I work with and they have some amazing sort of product leaders and electrical engineers and some they think of they're a little bit more average. So they wanted us to build one of these for them to help upskill our electrical engineers. So how to make the average electrical engineer as good as their most expert electrical engineer. And it's super exciting. It's really cool to see how it's helped them in a short period of time.
Willy Walker: It's fascinating. One thing as we try and wrap this up, the book has so much on getting the roadmap established, getting senior leadership, hiring the right people, having a hiring strategy with these pods that go in, make sure that you're only picking two to five projects out of the bag, and make sure that they're going to have a significant impact on your business. And then after all of that, that helps you kind of set the stage you come back to. But don't forget that establishing really good OKRs is fundamental. Creating sprints is fundamental and doing QBRs is fundamental. Talk for a moment about how we think about this whole new realm. And yet at the same time, you all bring it back. And one other quick thing before you answer that question, which I thought was fascinating. In the book, after every chapter, you set up a series of questions that allow the reader to reflect back on what he or she read in the chapter to make sure that you're kind of grasping the overall impact of what you're trying to do with that chapter. And I don't know whether McKinsey does that in your presentations to clients, but I found it to be an incredibly helpful tool, as they all say, tell them what you're going to tell them, then tell them again, and then tell them again at the very end. But the way that you created those questions at the end of every chapter to reflect back on what you just learned was an incredibly helpful way to sort of, if you will, sort what it was, but specifically to OKRs, sprints, and QBRs.
Rodney Zemmel: I'm glad you like that and how we set it up. Some people get a bit put off by like, okay, it's all about scrums and agile, like this is just some weird fancy religion, and so on. It works. It doesn't actually matter what you call it. But what we saw is, if you look at the 25 to 30% of companies that were making money in digital, they were all doing some version of agile. Many of them didn't call it agile they call it something else, and they wouldn't call it anything. But pretty much everybody within that money making quadrant was doing this. And this has a few characteristics.
First of all, it is business and technology working together. If you've ever got a business team that is writing requirements and then tapping those over to technology to go and execute, that's a failure mode. That's the old IT world, not the digital world. You need business and technology co-owning a problem and working together on it rather than shipping requirements back and forth. Next, you actually need the control functions also in those agile pods. If you have business and technology but you don't also have finance, quality, HR, legal, whatever it is that's relevant in your industry, then all you're doing is moving the bottlenecks further down the road. Then in terms of how you manage those teams, the sprint cycle, whether it's six weeks, eight weeks, ten weeks, doesn't matter, but something that is short enough to get to a clear deliverable that people can see. And then you would actually involve the senior leadership, often the CEO in those sprint reviews to see what comes out on those sprints, to keep things moving on the right cycle, and to make sure people are talking about the real product and not some sort of PowerPoint distillation of what the product is.
So, back right near the beginning of the conversation we're talking about, can you actually measure the business impact of digital, and how do you do that? It can be quite hard at an individual application level to say how much money we're really going to make through feature X versus feature Y and so on. So our general advice is don't try and do that. Set the financial goal at the top and then for every piece, for every individual use case there might be underneath that financial goal, you manage that through OKRs (Objectives and Key Results). So just very focused. What are the objectives that you're trying to achieve with that particular use case or that particular piece of the project and manage to that. And I have confidence if you keep managing to that, it will then ladder up to the overall financial objective. But don't just try and measure everything through a typical financial planning application.
Willy Walker: So I want to end with two quotes and then have you sort of leave us with your thought as it relates to something that we may not have either discussed or what people ought to keep their mind on.
This is backing up to you talked in the book a lot about security and security of data and people who want to have their own databases controlled and putting things into a large language model and what the security risks are there. But you put a quote in the book, Rodney, that I think is just great, which is just talking about basic passwords to log into your computer this morning. I've never heard this before, and I wanted to make sure I got it in here, which is to treat your password like a toothbrush. Don't let anyone else use it and get a new one every six months. That makes so much sense. If you're talking about password security and changing it on time, I say, yeah, don't let anyone else use it, and change it every six months. I thought that was just great, even though the security you're talking about is on such a different level. But to those of us who log into a computer every day, I thought that was a great reminder.
Rodney Zemmel: Yes. And unfortunately, I didn't like changing my toothbrush. I don't quite do it that often, but we should.
Willy Walker: And then the other one that I thought was so great was you have a great quote from The Princess Bride of Miracle Max, which to anyone who, I'm a huge fan of The Princess Bride, one of the great movies ever. But he talks about rushing the miracle. And if you rush a miracle, you get rotten miracles.”
And I think that's super helpful as people think about diving into this, that there's so much talk about it. When I listen to all of your interviews, you're like, I've never been asked to talk more in my entire career. In the last six months, I've had everyone on the face of the planet asked me to come talk about generative AI. There's a little bit of hype there today, but this digital transformation is real. It's been going on for 30 years, and it's going to continue to go forward. And anyone who doesn't get on this train is really going to miss it. But there's this sense of everybody, it’s sort of, “I got to get to it.” And I thought that miracle Max quote was so good in the sense that there are no miracles here. There is no technological solution that you can just snap your fingers and implement. And that really, if you think about it, while we all talk about digital transformation technology, as you so effectively underscore in the book, it's really more of a management-corporate transformation and journey than it is. Hey, we got to get there and tomorrow we're going to have some whiz-bang technology.
Rodney Zemmel: And actually, that's a perfect thought to leave people with. Because this is just going to be the way sophisticated modern companies do business. And it's interesting. We have a survey that we operate on digital quotient, and how digitized different companies are. And if you looked at the data from three or four years ago, everything was pretty much erased by industry, right? So if you are in banking or in high tech, you are super digitized. And if you were in industrials as a public sector, you were super non-digitized real estate generally more towards the non-digitized end. And so the industry was really destiny. What we've seen in the last maybe 18 months in particular, is companies have really broken out of their industry averages. And there's now more difference within an industry than there is between industries. So the best retailer is more digital than the average high-tech company. And the lagging retailers are less digital than the worst public sector institution. And it's really sort of spread out. And once companies get ahead, they then stay ahead and move further. We see the spread increasing over time.
So, maybe ending with your sort of miracle Max quote. And by the way, our editor cites a letter to help us find all these quotes. I'll pass your appreciation on to him. It is that this is not your transformation, that’s the wrong word. This is a journey. And if you've not already started on the journey, now is really the time to move on it. Because these new tools, generative AI and so on, massively actually lower the cost and accelerate the speed of moving. That's a challenge as well as a blessing because that could lead to death by a thousand pilots and proliferation and so on. So that's why you really have to start where chapter one is in the book with this top-down business like digital.
Willy Walker: I am deeply thankful for you spending an hour talking about it. It's been fascinating. It's a great book as I hope everyone who's listening today can tell. Go buy it. And Rodney, thank you so much for your time.
Rodney Zemmel: Thank you and great to see you.
Willy Walker: Take care.
Rodney Zemmel: Bye.
