Harry Koehnemann, Jeff Shupack, and Dean Leffingwell

Welcome to our talk on Continuous Learning: A Management Approach for the Digital Age. The question mark is a hypothesis of ours. These are some ideas that we've been kicking around for the last few years or so, and we want to formalize them here. We're going to be asking for your input at the end.

We'll also mention that there's a Birds of a Feather session going on later today that we're going to attend, and we hope you all can attend too to provide some more ideas here.

My name is Harry Koehnemann. I'm with Scaled Agile. I'm a practitioner there, and with me also is...

Jeff Shupack

Jeff Shupack. Hi, everybody. Very nice to meet you. I work with the consulting firm that's really helping large, complex organizations shift learning left and focus on learning continuously.

Dean Leffingwell

What I can say about these guys is the bigger the system is, the better they like it. These guys know some of the largest systems in the world. I'm Dean Leffingwell. I'm with Scaled Agile as well.

We've come to this kind of in parallel paths. I've been a student of leadership all my life. I always try to be a better leader and manager. I fail on occasion for sure. I remember those failures specifically. But we're starting to come to a different kind of conclusion about what leadership might look like in the digital age, and that's the purpose of our talk today.

Jeff Shupack

Perfect. Continuous learning is a very intentional word choice. I think we're all familiar with learning, but the moment we add the word continuous, that's a little bit different.

Anybody seen this photo before? Let's see a show of hands. Is anyone familiar with this? What's the date on there? It's 1903. October, the Aerodrome.

Anyone? No? This is a first attempt at a heavier-than-air first flight, and it's from Samuel Langley. Samuel Langley was America's contender to first flight: $50,000, a huge, single-point approach to this. This is attempt number one. You can see in that next photo it didn't go so hot. At the time, he was the head of the Smithsonian, chief of the Smithsonian, and it never flew due to fundamental design flaws. So very point-based.

What about this photo? Is this a little bit more recognizable? One month later, a different group of people. This is the Wright brothers, right? This is Kitty Hawk, a month later, their first successful flight. They took a fundamentally different approach. They optimized for learning itself. They iterated themselves. They started to chase down these barriers of flight. They built a home wind tunnel to experiment with controls and lift and how to propel. They spent a lot of time, and ultimately these rapid experiments created the first flying machine.

So they took a more continuous approach to it. Learning is not really new, is it?

What I want to speak to a little bit, and have Harry chime in here, is: how do we build this into everything we do? How does learning itself become this management framework for the digital age?

Dean Leffingwell

Yeah, we're going to give you several examples that we've seen in our practice, in the places we've worked at. We're going to start with... oops, wrong direction. Apologies for that. Oh, wait a minute. Oh, no, we're not.

So how many people have seen this chart? We want to first talk about the new age that we're in now. Most of us got exposed to Carlota Perez's research through Mik Kersten's book Project to Product. In there she talks about the technological revolutions that had massive economic impact in our world. In every one of those there was an installation period where capital flowed in, and then there was this distillation period where the winners emerged into the deployment period.

In Detroit alone, there were 300 automotive companies in the 1920s. Then, as we know today, not too long after, in the '40s and '50s, there were three remaining. So the question is, where are we today in our digital age? Certainly we probably don't see Amazon being upset anytime soon in digital retail, but in the spaces we play in, like automotive, the jury's still out. We see a lot of capital still being flown in for the winners that are going to emerge in that very complex environment.

Now Carlota Perez didn't just talk about economics. She also talked about the way of working being different too. They also emerged with each of those revolutions a different way of working. We borrowed this slide we saw from Gene Kim and Mik Kersten on a previous summit, where they hypothesized and connected the management approaches over there to each one of the ages, but they left the one on the bottom open as an open question.

So we're going to posit that continuous learning is really the management approach for the digital age, for the new way of working. We have the old ways of working around Taylorism and Fordism. They were trying to optimize a point solution, looking at the steps in a process and trying to optimize those steps, or trying to even automate those steps. In Fordism, the problems we solve today aren't the same as those problems were. They are larger. They're things we haven't done before, like flight, as Jeff just pointed out.

So we want to discuss several patterns that we see emerging that might describe what continuous learning looks like from a leadership perspective. The first one we're going to look at is not one of the CALMS and not one of the revolutionary companies. It's the hundred-year-old company called General Motors.

Back in April of 2019, the CEO of General Motors went to the engineering group and said, we need an electric Hummer. We already were seeing competition from startups. That's where we make a lot of our profit, from our trucks. So we need to have an electronic version of this in three years.

Well, General Motors' vehicle cycle is five to seven years. It takes them three or four years just to get the first prototype out. But she challenged them, and they had to approach it differently. There was a research group that was exploring digital engineering. They brought those folks in. There was some battery research going on, a battery platform. They brought those folks in. They didn't follow the normal process. They did things differently.

Sure enough, in 18 months they had produced their first prototype, and they had the car out within three years. They met the deadline, and, oh by the way, COVID hit in the middle of it. So in addition to this there was massive disruption that went along. Probably a good thing they'd moved to the digital engineering because everybody went home, and they could do the learning digitally instead of the traditional way of manufacturing parts and integrating them and then going back and remanufacturing parts.

So a very great example, even not in a comms company, one of our old standard companies, about using continuous learning to accelerate, get faster business value.

Jeff Shupack

And was it successful?

Dean Leffingwell

It absolutely was. Just a week ago, I read where they're not even taking orders anymore for the electric Hummer. They have such a backlog they've decided, we're just going to annoy everybody by being at the end of the queue, so we're just going to stop taking orders. That's a remarkable success story for a big and old and stodgy, up till then, company.

Jeff Shupack

One of the things that we've done is we've shifted away from these milestones that are traditionally based on phase gates and more into learning-based milestones. You can see that here is the example of SpaceX and NASA. At the top, following the traditional approach: requirements complete, preliminary design complete. What SpaceX has done is they've moved the actual milestones to proof of learning.

The SN stands for serial number, and I want to take a second to focus on SN8 there, which was in December 2020. Let's take a look at that one in detail. About a month before launch, Elon publicly tweeted these remarks here: lots of things need to go right, so maybe a one-in-three chance. But that's why we have SN9 and SN10.

That's a big difference when we tend to think about things going right or successful milestones. We don't always think of it from a learning standpoint. We tend to think of it from a payload delivery standpoint.

Anybody familiar with what happened with SN8? It blew up in a spectacular fashion. The following month it obviously exploded. In most organizations, and I think I can speak to that rather factually, someone would have lost their job on this.

Within a very short period immediately after, Elon gave a quote here: the fuel header tank pressure was low during landing burn, causing touchdown velocity to be high and RUD. That's my favorite acronym. What's RUD? Rapid Unscheduled Disassembly. It happens so often they have an acronym for it. They plan for it.

And look at the line: we got all the data we needed. Congratulations SpaceX. That's a fundamentally different definition of success for most.

Dean Leffingwell

Yeah, at NASA, which we're comparing with, there probably would have been congressional investigations. I've done NASA programs, and they don't fail, or they stop. They stop work.

Jeff Shupack

If we roll back to the pattern we've seen here, they need some telemetry into the system to get the data to validate learning. So when they do have that opportunity of a RUD, they can learn from that.

Further, SpaceX has been very successful at turning their factory into their actual product. They have no interest in producing the same serial-number rocket each time. They want to learn every iteration, and that's what they're defining their actual product as.

Let's take a look at this chart here. You can see the difference over time from SpaceX NASA launches compared to other NASA launches. You can see who's winning the work. I was just reading an article about it earlier today, and I sent it over to Harry and Dean, and in the last two years SpaceX put more folks in orbit than anybody else. It's shocking.

Pulling on the factory as a product, it's not just SpaceX. I want to pick out another Musk-company example. This screenshot I took from the Cyber Rodeo last April, and one of the things that I noted is this is the machine that builds the machine. He's not talking about the Tesla. He's once again talking about the factory that can build and learn based off the next widget, the next Tesla's needs.

And here's his quote: this is a machine that builds the machine, and it's the latest version of the machine that builds the machine. I've said it before, the factory is the product. Prototypes require imagination, and they're not easy. But relative to production, prototypes are easy and production is hard.

That's a complete shift. Usually, historically, when I thought about cars, I would think about a specific model year, a specific model, and a specific brand. That's a very big departure from what Elon and Tesla and SpaceX are doing here. They're optimizing their factory to learn as they go.

Dean Leffingwell

That's really the story. Everything in Tesla is designed for change. The vehicle is designed to be changed, as Jeff mentioned. Every vehicle that comes on the line might be different, could have a different heat pump, could have a different wiring harness. So they optimize not just the car to be able to change by having component-based and standard interfaces, things that are near and dear to our hearts as software and software architects, but also the manufacturing line and process itself.

Most manufacturing lines are designed for predictability. They don't want any variability in the manufacturing line. Tesla embraces that. They have to if every car is going to be different. They're not trying to prove every model year. They're trying to prove every car.

They have massive amounts of data that comes back in from the operational environment. They're using that to try to improve the next car, not the next model year. So they have to build a lot of quality in. They have to build compliance in. It's all going to be built in because they're not going to certify every model year; they're going to certify every car off the line.

That's what we're showing here. That's the bamboo line, as they call that at Tesla. It's the last line in production where they run every non-destructive test on every car that leaves the manufacturing plant. We would call that some kind of continuous pipeline. That's our pipeline of delivery, and they're doing the same thing in automotive so they can provide the ability to do frequent changes.

Last, we're going to move on to the sixth thing we're seeing, which is the reduction of the learning batch size. So we've got to start shifting things left. Tesla is a great example. I want to learn every vehicle, not every model year. General Motors is an example: I don't want to wait to manufacture parts to learn. I want to learn digitally. I want to learn in the digital environment.

So we're seeing a big shift left and investment in digital environments. There's also a lot of 3D printing, and there are other resources out there: companies where you send them CAD files, they'll send you a board the next day, 3D printing in a day, those kinds of technologies. So there's a lot of shift in this space for us to learn faster and shift that learning left.

Jeff Shupack and Dean Leffingwell

So these are emerging patterns of learning that we have hypothesized, and you can see them up there.

Dean, what do these things have in common?

I'm going to ask you that question.

I think there's a couple things in common, and then I have a question for you guys. The first thing we see in common is leadership from the top, expecting failure or recognizing that failure is part of the learning process and setting the stage for psychological safety, which is a strong elementary leadership model. The second thing we have in common is the fact that it's iterative. Everything is Tesla now. A single Tesla is an iteration of the Tesla before.

Those are two things in common. Now, as we looked at Carlota Perez's slide, there are a few things that weren't mentioned there, like transformational leadership, a very powerful style of leadership, in vogue very much at this conference. Steve Mayner helped deliver that at this conference. Servant leadership absolutely has this role. When I think about release train engineers and scrum masters, that's what I'm looking for. Certainly looking for a leader as expert knowledge leader.

What's different about leadership in the digital age that causes us to think that learning is going to be the paradigm? What's unique?

Jeff Shupack

I think one of the big things that's different today, or that's being recognized as different today, is where learning is occurring. Historically, it's always been done in special projects. These examples could be a Skunk Works type area where we're going to have some R&D, or it could be Google X labs. What we're seeing today is a need for learning to not just occur in these R&D spaces, but actually in the core competency of the business or the enterprise. That's a dramatic shift.

Numerous organizations that we work with will say, well, we're risk-intolerant. We can't take any risk in our crown jewels. We can't allow for experimentation. There is this lack of willingness to learn and grow based off the rapid disruption we're all seeing. So I think it's where it's occurring and where leadership is allowing and encouraging learning to grow.

Dean Leffingwell

I think there's a sentiment of leadership itself. I grew up as a software engineer and software developer, and I spent most of my time in engineering companies. I hear the challenge of connecting business with IT. As we move to Agile at scale, I didn't experience a lot of those problems in the aerospace and defense and automotive industries. I credited it, frankly, to: well, the leaders there came up from engineering, they're just engineers. They understand the problems.

Instead, as we put these slides together, I think it's that they come from a different background. The scientific method and research is important and validated, as opposed to business, which is more predictable and accounting- and number-driven. We don't quantify the value necessarily of research that's going to be done. So I think that we need leaders that understand the value of the research. It's hard to put numbers on that, but engineering people understand the scientific method and the value of hypothesis and doing research and validating things. I think other segments of the market may or may not do that.

We saw this morning where I like your point about where learning occurs. We saw the point where the feedback from Ops was never even fed back into the R&D cycle. So I wonder if there's a parallel with this Dev and Ops, or R&D versus manufacturing. Is this a continuous thread of understanding where the learning has to occur and getting feedback?

Harry Koehnemann

Yeah, that's a great point. I'll take that. So all the new parts, the engineering of the new wiring harness, the engineering of that new heat pump, that's not done in a research group and then thrown over to manufacturing. The engineering for those new pumps is really done in the line. It's done in the manufacturing plant.

We're trying to connect Dev and Ops. They're connecting engineering and production. They're all done in the same place. That's by the same people. That's why they have the cross-functional team. The people that are designing the part and the people that know how to manufacture and install those parts are right there working together side by side.

Jeff Shupack

I think there is also a shift in what value is that we're optimizing for. Historically, if you took a very traditional lean approach, you'd be optimizing for maybe unit production or the number of widgets through the system. That is still important. But now I'm really focusing not just on speed of delivery but speed of learning in that small batch size, which is one of the patterns we saw: how fast can I learn, which then will drive faster accelerated value delivery at the same quality.

But it's those rapid experimentation, those rapid learning loops. To play on the leadership piece, there's a psychological safety component needed. When one has a very public tweet saying the value is on the learning itself, it does create an environment, at least in this little microcosm, for learning to occur. I didn't have to worry so much about success being defined as payload delivery. Of course they wanted that, but if we learned and we had full-stack telemetry and we had a way to leverage this uncertainty, the growth in that uncertainty over time, that's a big shift. It's the value definition change.

Dean Leffingwell

I grew up in a trucking company, and to make my way through college I drove a truck. There's a little learning involved in that, but to be specific, I was told what to do and when to be there, how big a load to do, how to run the compressor, and how to unload that system. There were people who knew more about that than me, so that was kind of Taylorism.

But I wonder if, in the age of knowledge work, we don't know more about that than people that are doing the work. I wonder if that's a change that caused us to think about learning differently, too.

Jeff Shupack

Absolutely. That goes back to Daniel Pink in the book Drive: what motivates us, purpose, autonomy, mastery. If I am a knowledge worker, and by definition I know more about my job as a subject matter expert, as an engineer, than my leader, the encouragement to have learning-based milestones where I can express autonomy and I can express mastery in that growth over a vision that feels purposeful is huge. I think that's a huge motivator. That's what's drawing a lot of folks to this type of enterprise, and it's on leadership to unlock that environment for those individuals to thrive.

Harry Koehnemann

That's great. And I notice we're out of time, like 12 seconds left. In our last 12 seconds, I want to remind everybody there is a Birds of a Feather today. We're going to talk about culture and leadership. We're going to be there, and we would love to hear additional insight you have at your organizations on the importance of culture and leadership in a learning environment. So with that, we'll end and thank you. We'll see you next session at one.

Dean Leffingwell and Jeff Shupack

But is learning a need of the knowledge worker? So as leaders and managers, if we don't address that need, we're not going to be effective in our role.

I would agree fully.

Great. We'll see you at the breakout, and we want to hear from you how you've seen learning as a leadership paradigm. We're up for any tips and tricks you have to help us promulgate that message.

Thank you, everybody.