Gene Kim and Dr. Steven J. Spear

Gene Kim: My longtime friend, co-author of the "Wiring the Winning Organization" book, which is Dr. Steven Spear from the MIT Sloan School of Business.

As I have mentioned so many times, it was the most intellectually challenging thing that I have ever worked on, but also one of the most intellectually rewarding. Steve, so good to see you again.

Dr. Steven J. Spear: Hey, Gene. Good morning, good afternoon to wherever people are. It is starting to dawn on me, this "most intellectually challenging" thing you say. I am not sure if I should be taking that as a compliment or an insult. Because if my kids were to say something like that, it would be, "Well, Dad, what you were saying was not clear, and it made no sense anyway." So anyway.

Gene Kim: No, it was good. In fact, I always believe explicit is better than implicit. The amount I learned and the number of times where I realized what I thought I knew but did not actually sufficiently understand was off the charts, but it was so satisfying to work with you to get down on paper things that we think are so important.

Steve, if it is okay with you, I have two questions, unless there is something you want to share first, about things that have been really fun for me over the last year that I would love to ask you about, that I think would be of vast interest to this community.

One of the really rewarding and fun things for me over the last two and a half years was this dawning epiphany about linking option theory, the work of Dr. Bob Merton and Scholes and Black, to quantify the value of modularity and by extension therefore linearity as well, linearization as well.

At one point we had three pages explaining this, which we then reduced to one paragraph and one footnote. So can you take five minutes and teach us about what option theory is, why it is important, and demonstrate how and why modularity is so valuable?

Dr. Steven J. Spear

Dr. Steven J. Spear: Sure. Gene, I was thinking about this. There is value in procrastination. I think that is the key point about optionality.

Let me step back and explain what that means. We are critical of people who procrastinate. The reason we are critical of people who procrastinate is we think they are going to run out of time. But there is the flip side, and what I meant by this as a joke in some ways, is that if we are forced to make a commitment today about something we will do tomorrow, we are making that commitment today based on very imperfect information. Because what we know today about tomorrow is less perfect than what we know about tomorrow, tomorrow.

Obviously, that realization of the value of being able to delay decisions is just something that is inherent to us. Black, Scholes, and Merton gave us a way to put a quantification on the value of procrastination, and a very particular type of value of procrastination: to decide tomorrow and not be forced to decide today.

That was this notion of separating actions and time. It was allowing us to partition where we were today with where we were tomorrow. It allowed us to make that delay.

Along the way, people realized that this whole notion of giving decision rights and allowing that there is value to that is not only value that is temporal, decision rights to procrastinate from today until tomorrow, but there are also spatial decision rights. The folks who really gave us a lot of clarity on that were folks like Steve Eppinger with his design structure matrix, and Kim Clark and Carliss Baldwin with their book about design rules, which is rooted in option theory.

Basically, just as through time there is value in procrastination, in space there is value in decoupling, in partitioning. If you and I have to do everything together, then there is a huge coordinative cost. I always have to say, "Hey, Gene, what do you think of this idea?" "Oh, Steve, what do you think?" "Here is my input, here is my output, here is when I am available, here is when I am not available."

But if we can partition things and say, "Gene, here is your coherent piece of work. You go work on that. Steve, here is your coherent piece of work. You go work on that." So long as we have a well-established standard as to how the work comes back together, then we have all this latitude because you can work at your own pace, in your own way, at your own time, drawing on the resources you need, whether it is Anne or Jeff, and I can work at my own pace, at my own time, in my own way, so long as we just come back together.

This option theory, this ability to decouple, whether it is temporal or spatial, and allow this independence of action, just has extraordinary, extraordinary value. Merton, Black, and Scholes all won Nobel Prizes for showing how to quantify that. Clark and Baldwin and Eppinger have added enormous value, showing organizations the value of partitioning their work spatially and showing them how to do that effectively.

Gene Kim and Dr. Steven J. Spear

Gene Kim: This was probably my biggest aha moment, that couch metaphor, which I will be talking about later. In fact, the checkbox project, which the team will be presenting tomorrow, was actually this illustration of what do you do when all you want to do is ship a checkbox that has to transit across 40 teams so that people can opt into a $5-a-month service? If you do not have independence of action, you end up in that situation they describe where it will take $28 million, daily war room meetings, and a 20% chance of success because it did not work the first two times.

That is exactly what we are talking about when you are referring to the absence of independence of action. Is that right?

Dr. Steven J. Spear: One hundred percent, Gene. One of the things, apropos of lessons learned going through this book, as we argue through all of section one, is by the time we hit the end of section one, all the key concepts are there, and we have a list, and we have definitions and all that.

One of the key definitions in there is this notion of coherence. We say something coherent is something which is complete. It has all the resources it needs to perform some function logically and consistently. Then we start comparing something which is coherent in that very minimalist sense, it has everything it needs but nothing else, with things that are overly partitioned, which are undercoherent. They cannot function without extraordinary effort.

We give lots of examples of siloized organizations where you have not created the coherence around a function. But then we also say in the book that you can have things which are underpartitioned and overly coupled, that you have all sorts of things thrown in, in terms of functions that have to be performed, resources that have to be accounted for, et cetera.

In the book, our hotel vignette in chapter two is exactly that in the beginning. It is this underpartitioned, overcoupled system where there is minimal independence of action. In the absence of independence of action, you get one of two things. One, you get calcification because there is so much need for everyone to coordinate with everybody else about every little thing all the time in every place that no one can actually move, because all this conversation about what are you going to do, when are you going to do it. That is one.

The other possibility when you have this overcoupling is just chaos, because people start expressing agency, but they actually do not have independence of action. So you have all these independently moving pieces, and you have chaos.

This whole notion of partitioning, coupling, and optionality is a huge idea. Again, we did not generate the idea. Merton, Scholes, Black, Baldwin, Eppinger. But it is a gigantic idea that we were able to draw upon and put to good effect, I think.

Gene Kim

Gene Kim: I was going to ask you a second question, but I actually want to interrupt and show you what I did two days ago. It was just so mind-blowing to me.

Many of you have seen that I have been using ChatGPT and LLMs like Claude to use the mechanisms of simplification, amplification, as a lens to view these case studies. I have been posting case studies from the DevOps Handbook. I have used it to interpret the Google "We have no moat, and neither does OpenAI" paper, the Cloudflare outage.

There is something I did two days ago that delighted me, and I want to just share my screen and show it to you because it is exactly what Steve was talking about.

This is a ChatGPT session, and I prompted to say, "Steve and Gene wrote a book," and I gave it the couch vignette, saying that Steve and Gene moving a couch requires communication and coordination, and if you turn off all the lights or prevent them from talking directly with each other, all the right things that you would want stop happening.

Then I said, "Can you explain and link how this metaphor explains the Amazon API, the $1 billion API rearchitecture in the early 2000s?" I was so delighted that it said, "Oh, yes. Absolutely." First it rated the applicability, but then it said, "Oh, absolutely. I understand how in the before case, it was one couch that 3,500 software engineers were coupled to. None of them had independence of action."

Then I told it about the Amazon Video case study. This was the example where earlier this year, Amazon Prime Video put out a blog post that said they moved from a microservice architecture, which Amazon e-commerce eventually turned into, and moved it back to a monolith. This set the world on fire because everyone is thinking microservices are dead, everyone should move back to monoliths.

What I am so delighted by is that the couch metaphor also explains this. I fed ChatGPT the Amazon Video case study and asserted, and asked it to confirm, that this is as if you had chopped up the couch into two small pieces, where you are now dominated by coordination and transport costs. Absolutely, it confirmed: yes, you had high independence but low coupling, coordination complexity, et cetera, and moving back to gluing the couches together brought back the right level of coupling and cohesion.

Steve, how does that resonate with you?

Dr. Steven J. Spear

Dr. Steven J. Spear: I will tell you, it resonates on many levels. One is I am glad the AI bot whatnot saw the logic.

I will say something snide and sarcastic, but this artificial intelligence, it can be more and more intelligent the simpler it is, the rules we give it to apply. I talk to my students a lot about the huge power that Newtonian mechanics gives us in the design of anything. Really, you just have to ask the question constantly: if I want something to be stable, do the forces and the moments balance? I just have to ask that question over and over and over again, but I only have those questions.

When we go to the chat AI, this chatbot, what we have been able to do is give a very simple rule, which is if we are trying to get work done around a function, we have to make sure that we have designed the work around that function so that it is coherent. All the resources that are needed are there, and no extra resources.

I bet that had we not supplied this notion of coherence, coupling, partitioning, et cetera, to the bot, it would have no flipping clue. But what we were able to do is give it a very simple rule by which it could reach a fairly sophisticated or complicated conclusion, but only through the iteration.

One other thing I want to say, Gene. This is great. Here is the thing about coming back to basic theory. Whether you want to call it first principles, basic thing, or whatever else, it is equivalent to Newtonian mechanics in mechanics, or whatever else there is. This whole thing about Amazon and Video and they had microservices, not microservices: you could imagine the stampede, which is everyone saying, "Oh, microservices, because Amazon is the best. We should do what the best does. Boom, boom, boom, boom, boom." Go that way.

Then when Amazon shifts course and they are not doing microservices, then there is the stampede to say, "Well, Amazon is still the best. Boom, boom, boom, boom, boom." We talk about this at the end of the book, where we take some real potshots at how certain schools of management teach. It is the teaching by analogy.

If you are teaching the Amazon case by analogy, microservices, ba-ba-boom, or not, you have to have this stampede because people cannot reason through why Amazon is doing what it is doing. It is just that they are doing it, so, "Oh gosh, we have to copy the same thing."

But then you take what you did with the chatbot as a tool. It is not really intelligent, as a tool, and you said, "Chatbot, here are the first principles, the basic thinking, the core theory of how we think about organization. Part of it is this idea of coherence. Can we understand what Amazon did in terms of coherence?" It is like, absolutely, yeah. Of course. They had over-partitioned this way, under-coupled that way. They had gotten incoherent in that direction, and they are course-correcting.

What is the lesson to learn from Amazon then? It is not about microservices versus not microservices. It is about using the concept of coherence and making sure that you are coupled up to the level you need to be, but not over, and you are partitioned down to the level you need to be, but not over-partitioned. I think that is a great example. I am so thrilled. I love it.

Gene Kim and Dr. Steven J. Spear

Gene Kim: I love it that the before and after states of both cases are actually the inverse of each other, so there is a beautiful symmetry there.

Steve, I will share the ChatGPT session right after this conference is over today.

One of the things I wanted to ask you was about one of my favorite parts of the book: your depiction of the San Antonio truck plant that is in the Amplification chapter. It really struck me because I have read all your work, and yet the way you describe the plant is so different than in your previous writings, in "High Velocity Edge," in your "Decoding the DNA of the Toyota Production System."

It is something I have heard you talk about and present on, but the way you describe just the intense choreography and the level of dynamism that is happening within the manufacturing plant was totally new to me. Can you talk about why your depiction of it is so different than in your previous writings, and what it was like to write that section?

Dr. Steven J. Spear

Dr. Steven J. Spear: First, let me just make a point here for people who say, "Oh, well, we do not make cars." Looking at Toyota, which has had a 50-year lead in arguably one of the most competitive sectors in the world economy, we lay out in the beginning of the book the paradox. Why is it that in level playing field situations, we have these exemplars?

For anyone who says, "Oh, well, we do not make cars," you might be curious, or you should probably be curious, about a company that has had a 50-year lead in a very competitive level-playing-field situation.

The other part is making cars is wildly complex. Lines of code is bigger than most software products by orders of magnitude. Setting up a system that can generate a defect-free Tundra or Sequoia every minute with all the touchpoints of people and machines is wildly complex. The fact that one car comes out defect-free is, I think, miraculous; that it comes out one a minute.

Anyway, that is just the setup for those who are dismissive of, "Oh, we do not make cars."

In my first book, "The High Velocity Edge," I talk about the four S's of design work so you can see problems. When you see a problem, swarm on the problem to contain it, and then solve the problem so the local problem does not have a chance to exponentially snowball. Then when you have discovered something, make sure that you share, you systematize, you spread what has been discovered, where the fourth S is sustaining through leadership.

When I wrote that book, I was focused on things that occurred with some rapidity and some repetitiveness, but not intense, like a gerbil on amphetamines. I was talking about providing medical care and how you have to see little problems before they become big problems: confusion around medication, confusion around devices, confusion around patients, et cetera.

What changed? I would say from "The High Velocity Edge" to the Amplification chapter in "Wiring the Winning Organization," it is not so much a change, but, pun deliberately in here, it is amplified. The running of a plant with thousands of people who have some touch on a product, and that product is one that pops out and launches every minute, and then there is another one and another one and another one, is an unbelievable ballet of interconnected motions. It is just remarkable.

The people, the machines, the people working with machines, like we said, the sociotechnical system of many, many people using lots and lots of technology to accomplish something through collective action toward common purpose, it is amazing.

I think this is the first time, Toyota Texas, where I had a chance to spend a lot of time in a plant and then revisit the plant and really just be blown away, like that old Memorex commercial, just blown away by the kineticism and the beautifully choreographed, coordinated kineticism of that environment.

This is where it really started to come through: one person executing on a routine is going to glitch. One machine executing on a routine is going to glitch, and yet you have thousands of people and thousands of machines executing on those thousands and thousands of routines every minute, and the factory does not seem to glitch. Why is that?

That is where we talk so enthusiastically, and with, I would say, adoration, about the amplification mechanisms in those plants. As the work is unfolding, the work is so carefully scripted to become immediately obvious when a problem is at its earliest gestation. An associate can recognize it, a team leader can be responsive to it, a group leader can be responsive to it, an area manager can be responsive to it, all the way up to the site president, who can see like this that there is something that needs to be seen and solved, or recognized and stabilized in the run phase and, in between, rapidly improved.

The dynamicism of the kineticism, but also the intellectual effort to address problems as they first occur and then resolve them before they have a chance to become big problems, was just extraordinary.

Gene Kim and Dr. Steven J. Spear

Gene Kim: Awesome. By the way, the way the chapter opens in terms of the countercase being the Southwest Airlines holiday failure at the end of 2022, that too was one of the most intellectually challenging things.

I believe that we are wrapping up on time, but could you give technology leadership one piece of advice before we adjourn this session?

Dr. Steven J. Spear: When we had the conference back in Vegas some weeks ago, the question was, what did I take away out of this? There is a lot I took away out of the writing experience, but here is the shorthand and the shortcut: focus on the individual.

There are all sorts of ways to monitor and metricize and do all sorts of other things about these large systems, technical, social, sociotechnical, et cetera. But here is the thing. You look at the individual, and you see that individual struggling to do something that is valuable to somebody else in a way that they can easily be successful and in a way that they feel they are creating value for themselves also. If you see struggle, boom, there is a flaw in your system. End of story, full stop.

This was the Toyota thing to which you were referring. They have built this elaborate, miraculous overlay of support systems just so when something glitches, it is seen right away before it has a chance to become a big thing. That would be the advice to the tech leaders too.

Look at the individual. Do not worry about the metrics. Do not worry about the reports, all of this thing, that thing. Look at the individual, and if you see an individual having a glitch experience, see it, swarm it, solve it, and systematize what you have learned.

Gene Kim: Phenomenal. Thank you so much, Steve. It has been so exciting to see how this is being received by this community and broader. For those of you who do not know it, the book is really, in some way, an homage to the Toyota Production System and DevOps as a pinnacle of slowification, simplification, and amplification.

All right. To be continued, Steve. Thank you so much.

Dr. Steven J. Spear: Thank you, Gene. See you all. Bye-bye.