✨ AI and business: A Quick Q&A with … Andrew McAfee, author of "The Geek Way"
'Regardless of the shape of the crest of the AI wave, we are in the middle of a huge transformation.'
For people like myself and Andrew McAfee, there’s little doubt that AI is a transformational technology. How industry leaders choose to respond to the tech, however, might well determine the future of their companies, as well as the winners and losers of corporate competition in the 21st century. To talk about what AI means for the labor market, corporate strategy, and American business leadership, I asked McAfee a few quick questions.
McAfee is a principal research scientist at the MIT Sloan School of Management, as well as the co-founder and co-director of MIT’s Initiative on the Digital Economy. He is the author of The Geek Way: The Radical Mindset that Drives Extraordinary Results (you can listen to our discussion about the book here) and publishes a Substack newsletter of the same name. He’s also the co-author of The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies with Erik Brynjolfsson.
1/ I would guess that, especially since the publication of your book, and having spoken to plenty of folks in the business world — founders, CEOs, CIOs, consultants — what have they been telling you about AI, large language models? Are they using them? Has their enthusiasm changed over the past 18 months? What have they been telling you, if anything?
The definition of a great question, Jim, is one that came up over breakfast today. I literally today had breakfast with the CEO of one of the world's largest insurance companies, and he's a guy who I've learned a ton from. We have great conversations and so I said, “Mr. X, over the course of your career, where does AI rank?” And he said, “Number one, two, and three.” He said this is the biggest thing by far that he has come across in his career.
Is there a wave going on right now with generative AI? Is that part of what's going on as we're all so excited at what these language models can do? Yeah, absolutely. Regardless of the shape of the crest of that wave, we are in the middle of a huge transformation. Erik [Brynjolfsson] and I talked about the transformation of cognitive work being at the same level as the transformation of physical work in the First Industrial Revolution. That transformation, it ain't half over, it ain't peaking, it's not yesterday's news — It's a thing that's going to go on, to the point that Erik and I and two co-founders have actually started a company to help enterprises go through this, trying to help them scope out their priorities, get systematic about where their priorities with AI are, and then monitor progress toward them, because we've found that even accurately assessing ROI is really hard for these projects. I think the big story of economics in the 21st century is this credibility revolution, is this huge improvement in research designs, so that we can actually get at causality. How big a difference does this make?
I'll rattle off three of my favorite Erik Brynjolfsson papers for you: He did a great study where he looked at whether or not your cab driver is ripping you off on the way to JFK compared to Uber drivers, because you finally have enough data and you’ve got the right methods to answer that question: Yes, your cab driver is ripping you off in a way that Uber drivers can't and don't.
Number two: Do you sell more stuff when you offer automatic translation on eBay? Yes, you do. Again, that's kind of a subtle research design.
And then, most recently, you've seen his work on generative AI. He looked inside a call center with the team and found out that average productivity improved by 15 percent very quickly, but the really interesting part was that productivity went up by 30 percent for the newest and the least-skilled workers.
These are the kind of things that we can tease out with enough data and the proper research designs. So our feeling with founding Workhelix is: It's time to bring that to the enterprise.
2/ Let's assume that the insurance CEO you're talking to is correct, that this is an important technology, maybe the most important that any executive has seen in their career. For the companies that are unable to incorporate it, what will likely go wrong? The ones that find themselves falling behind, what won’t they be able to do?
We talked about this over at breakfast. There are a bunch of different failure modes:
One is that the CEOs of a lot of organizations are going to consciously — or, very often, I think subconsciously say to themselves, “This too will pass. I know my board is bugging me about this right now, so I've got to invest in a couple generative AI projects and I've got to report back and say how great they are and how we're all enthusiastic about it, but this is not where my heart is, it's not where my expertise is, and I've got X number of years left on my clock at the helm of this place, and this thing will pass and I've got other ways to finish out my time at the top of this organization.” I think that's one actual failure mode.
Another one is going to be nonbelievers: People who aren't saying, “Look, I don't understand this,” they’re saying, “I kind of do understand it and I don't think it's for us and I'm not enthusiastic about it. Our company has plenty of other things to do.”
There’s another category: People who just flat don't get it. They just don't understand. They don't have the view that you and I, and Eric, and this insurance company CEO have about the magnitude of this transaction. And so they're just not even going to make a serious attempt at it.
The last category to me is maybe the most interesting. In one of my Substack posts, I juxtapose a bunch of quotes from Steve Jurvetson, who deeply gets it, and I learned a ton from him, with Jim Farley who's the current CEO of Ford, who's given these really just incredibly admirably frank interviews about how car makers don't know how to write software. And the reason I love that juxtaposition is it's right next to Jurvetson saying, what we geeks have learned to do is build everything the way we build software: this fast-cycle, iterative, maximize-your-learning approach to building stuff. That's how we do everything! That's why Elon is rescuing the astronauts that Boeing stranded up at the Space station.
3/ Let me ask you about that one, because you've written about that in your great Substack, and that to me seems like a real-time example of two companies, one of which is doing exactly the kinds of things that you talk about in the book. They're agile, rapid-iteration — an approach that most journalists still seem to not understand about SpaceX — versus Boeing, which I grew up thinking, “That is a first-class company, American technology leader.”
This is an iconic American company. This is a jewel in the crown of the American century, without question.
They can't get their astronauts home from the space station.
What on earth is going on here? This is wild. There were a bunch of different threads to the story, and I don't know which of them are the most important. You hear that Boeing lost its way, that it got run by bean counters as opposed to engineers — I don't know. It's absolutely possible. The former CEO of Boeing was Jack Welch, an ex-GE [General Electric] guy, and that style of running a company —for good reason, I think — has really fallen out of style.
Another quote that I keep coming back to from The Geek Way, in addition to the Jurvetson quote, which I adore, there's a quote from Joe Henrich's amazing book called The Secret of Our Success, where he talks about, why are we the only species that launches spaceships? That's kind of weird. And he gives the best answer ever heard: We're culture-producing, rapidly learning animals, like nothing else on the planet. However, we're also really self-interested, and so we human beings, we have this really groupy nature and we have a really self-interested nature, like every other living thing.
The challenge of running an organization is that we humans do what we do, and we form cliques, and we form political bands to our own ends, which might not be lined up with the goal of the organization, and he says, basically, history teaches us that cooperative enterprises fall apart over time because of good old-fashioned self-interest, and coalitions, and stuff like that. And that helps me understand why the natural trajectory for companies — great ones — is to slow down, is to get sclerotic, is to get bureaucratic, get really jammed up over time. It's not because the CEO designs it that way, it’s because we humans do what we do: We politicize, we form cliques, we get involved in work, we want to be involved, and you wind up with an overly complicated, bureaucratic organization. I'm almost sure that's a big part of the Boeing story.
And then to fold in the Jurvetson thing, the art of building complicated things in the 21st century has become massively more difficult, because the things have become massively more complicated, because just about every important thing that I can think of now has a software component that's at least as important as the hardware component. And so what the geeks have learned to do is deal with that ridiculous complexity, and they do it by modularizing both the technology development and the organization. They do it by iterating, as opposed to planning. They do it by learning as quickly as possible. They do it by being willing to fail, as opposed to having that be a black mark on your career. They do it by building inherently more mouthy, egalitarian, evidence-driven organizations than the ones that you and I are used to when we were all admiring the Boeings and the GEs of the world. Those are the design decisions that they've made, and I think they've been forced into, because of the silly complexity of the stuff that they build and the fact that, in high-tech sectors, competition, I think it's an Eric Schmidt quote, the competition really is just a mouse click away.
And so you failing at any of these things is a terrible, terrible survival strategy. So I think it's kind of an unfair contest between organizations that run themselves the Geek Way and legacy organizations. And I think those gaps are already appearing to be quite big and I think they're going to get bigger.
So to go back to your question, Farley, to me, is the most interesting category of CEO because I think he honestly realizes the magnitude and the urgency of the challenge facing another great, big, iconic, successful American 20th-century company. And then the question is, can people like him pull it off? Because as I've listened to him, I believe he's got the awareness and he's got the drive. This is what he wants to do. Let's watch that. That's going to be interesting.
Keep reading with a 7-day free trial
Subscribe to Faster, Please! to keep reading this post and get 7 days of free access to the full post archives.