Matt Beane on the 3 Cs of skill development, AI augmentation design templates, inverted apprenticeships, and AI for skill enhancement

“The primary source of our reliable ability to produce results under pressure—i.e., skill—is attempting to solve complicated problems with an expert nearby.”

–Matt Beane

About Matt Beane

Matt Beane is Assistant Professor at University of California Santa Barbara, and a Digital Fellow with both Stanford’s Digital Economy Lab and MIT’s Institute for the Digital Economy. He was employee number two at the Internet of Things startup Humatics, where he played a key role in helping to found and fund the company, and is the author of the highly influential book The Skill Code: How to Save Human Ability in an Age of Intelligent Machines.

Website:

mattbeane.com

LinkedIn Profile:

Matt Beane

University Profile:

Matt Beane

Book:

The Skill Code

What you will learn

• Redefining skill development in the age of AI
• Why training alone doesn’t build true expertise
• The three Cs of optimal learning: challenge, complexity, connection
• How AI disrupts traditional apprenticeship models
• Inverted apprenticeships and bi-directional learning
• Designing workflows that upskill while delivering results
• The hidden cost of ignoring junior talent development

Episode Resources

Transcript

Ross Dawson: Matt, it is awesome to have you on the show.

Matt Beane: I’m delighted to be here. Really glad that you reached out.

Ross: So you are the author of The Skill Code. This builds on, I think, research for well over a decade. It came out over a year ago, and now this is very much of the moment, as people are saying all over the place that entry-level jobs are disappearing, and we’re talking about inverted pyramids and so on. So, what is The Skill Code?

Matt: Right. The first third of the book is devoted to the working conditions that humans need in order to build skill optimally.

The myth that is supported by billions of dollars of misdirected investment is that skill comes out of training. And that is—we just have a mountain of evidence that that’s not so. It can help, it can also hurt. But the primary source of our reliable ability to produce results under pressure, IE, skill, is attempting to solve complicated problems with an expert nearby.

Basically, we can learn, of course, without these conditions—sort of idealized conditions—but it can be great. And the first third of the book is devoted to what does it take for it to be great?

I got there sort of backwards by studying how people were trying to learn in the midst of trying to deal with new and intelligent technologies at work—and mostly failing. But a few succeeded. And so I just looked at those success cases and saw what they had in common across many industries and so on.

So, I break that out in the beginning of the book into three C’s—thankfully, in English, this broke out that way: Challenge, Complexity, and Connection. And those roughly equate—well, pretty precisely, actually, I should own the value of the book—they equate to four chunks of characteristics of the work that you’re embedded in that need to be in place in order for you to learn.

Challenge basically is: are you working close to, but not at, the edge of your capacity?

And complexity is: in addition to focusing on getting good at a thing that you’re trying to improve at, are you also sort of looking left and looking right in your environment to digest the full system you’re embedded in? That’s complexity.

And connection is building warm bonds of trust and respect between human beings. All three of those things—I could go into each—but basically, in concert, in no particular sequence—each workplace, each situation is different—but these are the base ingredients.

I used a DNA metaphor in the book. These are sort of the basic alphabet of what it takes to build skill, and your particular process or approach or situation is going to vary in terms of how those show up.

Ross: So, for getting to solutions or prescriptions, I mean, it’s probably worth laying out the problem.

AI or various technologies are making those who are entering the workforce—or entering particular careers—be able to readily do what they do. And essentially, a lot of the classic apprenticeship-style model has been that you learn by making mistakes and, as you say, alongside the masters.

And if people, if organizations, are saying, “Well, we no longer need so many entry-level people to do the dirty, dull work,” then we don’t have this pathway for people to develop those skills in the way you described.

Matt: Yes, and it’s even worse than that.

So, for those that remain—because, of course, organizations are going to hire some junior people—the problems that I document in my research, starting in 2012… Robotic surgery was one early example, but I’ve since moved on to investment banking and bomb disposal—I mean, very diverse examples.

When you introduce a new form of intelligent automation into the work, the primary way that you extract gains from that is that the expert in the work takes that tool and uses it to solve more of the problem per unit time, independently.

That word independently—I saw in stark relief in the operating room. When I saw traditional surgery—I watched many of these—there’s basically two people, shoulder to shoulder, four hands inside of a body, working together to get a job done. And that’s very intensive for that junior person, the medical resident in that case, and they’re learning a lot.

By contrast, in robotic surgery, there are two control consoles for this one robot that is attached through keyhole incisions into the patient. One person can control that robot and do the entire procedure themselves. And so, it is strictly optional then for that senior surgeon to decide that it’s time to give the controls to the junior person.

And when’s the right time to do that, given that that junior person will be slower and make more mistakes? This is true in law, in online education, in high finance, professional services—you name it. The answer is: never.

It is never a good time. Your CFO will be happy with you for not turning the controls over to the junior practitioner. And you yourself, as an expert, are going to be delighted.

People these days, using LLMs to solve coding problems, report lots more dopamine because they can finally get rid of all this grunt work and get to the interesting bits. And that’s marvelous for them. It’s marvelous for the organization—even if it’s uncertain there’s a little ROI.

But the primary, the net, nasty effect of that is that the novice—the junior person trying to learn—is no longer involved in the action. Because why would you?

And that breaks the primary ladder to skill for that person. And so, that, I think, is happening at great scale across…

Let’s put it this way: the evidence I have in hand indicates to me that there will be very rare and rare exceptions to the rule that junior people will be cut out of the action. Even when they’re hired and in the organization and are supposed to be involved, they will just be less involved—because they’re less necessary to support the work.

So even if you get a job as a junior person, you’re not necessarily guaranteed to be learning a dang thing. It’ll be harder these days by default.

Some interesting exceptions—and that’s what I focus on in the book. But that is the—in my view—I’ve done some arithmetic around this, and it’s all estimation of course. I published a piece in The Wall Street Journal on this about eight months ago.

This is a trillion-dollar problem for the economy, in my view.

Ross: Obviously, this is not destiny. These are challenges which we can understand, acknowledge, and address.

So, let’s say—obviously, part of it is, of course, the attitudes of the senior people and how it is they’ll be on frame. A lot can be organizational structures and how work is allocated. There’s a whole array of different things that can be done to at the very least mitigate the problem—or, I think, as you lay out in your book, move to an even better state for the ability to learn and grow and develop in conjunction, not just using learning tools.

But why don’t we go straight to Nirvana? Or what an ideal organization might do. What are some of the things they might do to be able to give these pathways where people can contribute and add value immediately, as well as rapidly grow and develop their capabilities?

Matt: Right. So, I’ll give you a few examples, one of which was evident in my book—and a couple examples, one of which was in the book, and one of which is new since the book’s publication.

So, the one that’s in the book—and that has always occurred, I think, and is more intensely available now and is a real cool and valuable opportunity for organizations—is what I called inverted apprenticeships.

This comes out of a study that I did with a colleague at NYU named Callan Anthony, where we contrasted our surgical and high finance data. We both have sort of “who said what to who every five seconds” kind of transcript data on thousands of hours of work in both contexts.

What was very clea