Computing Power and Regenerative AI Constraints

Many companies strive to automate by using more technology and fewer humans. But does their productivity really improve? Does it keep them agile? In this episode, Jacob Stoller and Andrew Stotz share stories of companies that improve productivity because they focus on processes instead of tech alone.

TRANSCRIPT

0:00:02.3 Andrew Stotz: My name is Andrew Stotz, and I'll be your host as we dive deeper into the teachings of Dr. W. Edwards Deming. Today, I continue my conversation with Jacob Stoller, Shingo Prize-winning author of The Lean CEO and Productivity Reimagined, which explores applying Lean and Deming management principles at the enterprise level. The topic for today is myth number five, the Myth of Tech Omnipotence. Jacob, take it away.

 

0:00:29.8 Jacob Stoller: Great, Andrew. Thanks. Great to be here again. Yeah. Tech omnipotence. Well, it's quite a myth. We sort of worship technology. We have for a long time, and we tend to think it can solve all our problems, and sometimes we get a little too optimistic about it. What I wanna talk about is in the context of companies adopting technology and go through some of the stories about that and how that relates to productivity. Really, the myth of tech omnipotence is kind of like a corollary to the the myth of segmented success. In other words, people have believed that you can take a chunk of a company. Now we'll take Dr. Deming's pyramid, and we take a chunk out of that and say, oh, well, that fits so and so in the org chart, let's automate that.

 

0:01:28.1 JS: And they don't consider what happens to the rest of the organization. It's just this idea that you can superimpose automation. So this has a long checkered history. And the way technology gets justified in organizations is generally what it's been, is reducing headcount. And I used to work in a tech firm, and we used to do this. We would do these studies, not really a study, but you do a questionnaire and you figure out if we adopt this, if we automate this workflow, let's just say, I don't know, it's accounts payable. So you automate accounts payable and you say, well, you got so many people involved, we think we could cut this by three people or something like that. So that becomes your business case. Now, they had categories in these little questionnaires where you would try to get other benefits from the technology, but they tended to be what they call soft benefits.

 

0:02:35.4 JS: And you know what that word means. Soft benefits means, well, okay, nice to have, but it's not going to get budget money or it's not gonna get approved. So anyway that's really been the kind of standard way of getting tech projects justified. And that goes through pretty much any industry. So what would happen is people adopt these technologies without looking at the whole system. And guess what? You put the software in, you start to implement it, and you run into problems. Doesn't quite work. Doesn't work the way it was supposed to. And so the tech people tended and still do tend to blame the company. They say, well, they had user problems. Users weren't really adjusting to it. These people are sort of way behind. We're a tech company. We've automated the same process for 50 different companies, we know what's good for them. We have to educate them, but they don't seem to want to be educated. So that was kind of the way it was. And I'll give you an extreme example. I did some freelance work for research firm, and one of the studies I worked on, I'm not making this up, it was called Aligning the Business with IT. So it was trying to get people to smarten up with their business and align it to what the smart people are doing with IT. So that's how extreme that kind of feeling was.

 

0:04:17.3 AS: As opposed to maybe aligning with the customer or something like that.

 

0:04:21.1 JS: Well, yeah, wouldn't that be crazy? Or how about aligning IT with the business? Finding out what the business wants. So anyway, that whole way of thinking has had, it's sort of filtered into manufacturing in the same way. And I found this out really researching Productivity Reimagined as I interviewed Ben Armstrong from MIT Industrial Performance Center. And what I learned from him is the whole history of automation and manufacturing in North America. And really, what he told me is that between 1990 and 2010, there were increases in productivity, but those were always from reducing headcount. They never found ways to actually grow the value of the business by using automation. So around 2010 or leading up to 2010, manufacturing started to change, and we started to transition into what they call a high-mix, low-volume type of markets.

 

0:05:33.3 JS: And I've talked to manufacturers that have said, 10 years ago, I only had to make two or three variations of this part, now I have to make 50 or 60. So you're getting shorter product cycles, larger mix. And the big buzzword now in manufacturing is agility. You've gotta be agile. So there was a study MIT, I think this Performance Center did a study. And they found that when you actually try to grow productivity, and this is really since 2010, you actually lose agility at the same time. You're kind of caught in that situation because you can't... That you lose agility when you let go of people. But that was the only way they could increase productivity. Does that make sense?

 

0:06:29.1 AS: Yeah. So I'm thinking about that's interesting because agility means being flexible, being able to accommodate. And when you think about the typical automation, it's about repetitive, repetitive, repetitive.

 

0:06:46.5 JS: Yeah.

 

0:06:47.3 AS: And so I can kind of get that picture about the agility versus, let's say automation or repetitive processes.

 

0:06:56.3 JS: Yeah. And I think that people are longing for this golden age. You go from the 1920s to 1960s, and manufacturers made incredible gains in productivity with automation. You put in these huge welding lines where they just weld. You look at the body welding, say in a plant, and it's at lightning speed. There's no question about that. But they basically ran into a plateau with that. And one of the robotics companies told me, he said, we learned decades ago how to automate these mass production processes, but now we're getting into a different kind of age where as somebody put it, we're moving from the industrial mass production age into what they call the process age, where processes are becoming more and more important. So to...

 

0:07:50.8 AS: And I'm thinking about the automation. I've seen videos on like online about let's say a fulfillment center with all these little robots going around and picking, putting things on them and packaging them, and all of that. So I'm thinking, well, automation has become definitely more maybe, I don't know if the words agile, but it's definitely, it's gone beyond like just automating one little part of the process.

 

0:08:21.4 JS: Yeah. It's gone away from the let's replace people type scenario. And so what the fastest growing segment right now in robotics is collaborative robots, which can work with people. So to put it very simply, instead of a human replacement, they're becoming tools. But these things are amazing. A worker online on the shop floor can programming these, and they have to be able to because things are changing so fast. So a worker, a welder can actually hold the robotic arm and guide it through a weld and thereby program it so it can learn how to do that weld. So then you can get the robot doing all the dangerous parts. If they're welding something large where they might have to get up on scaffolds or something, they might be able to get the robot to do some of the more dangerous types of positions. So that's when you get the real benefit.

 

0:09:27.7 AS: Yeah. I would think like in a paint booth, which we had in factories I worked at, now you can seal it off and have a robot in there, and all of a sudden lung problems and other things like that just go away.

 

0:09:40.8 JS: Interesting. Well, so anyway, we're still in a, I think in a rough spot generally with manufacturing because between 2010 and present day, at least in North America, productivity's gone down. And it's because people haven't been able to... They've depended on those people to keep their agility, but they haven't learned how to add value.

 

0:10:08.3 AS: Can you discuss that just for a second about productivity going down? That's a little bit of an odd thing because I think most people think that productivity's probably going up. What is the measure you're talking about, and how long and why is that happening?

 

0:10:23.5 JS: I think it's basically... At least I'd have to look at the study that they have, but it's basically output in proportion to the number of hours. I think that's pretty well accepted. So they're losing ground as the demands for agility are increasing. And their attempts to automate have been, caused problems. You automate and you lose your people, and then you're gonna have a heck of a time getting them back right now because that's really hard in manufacturing. But yeah, I would have to look at the study in detail to understand how they got that number, but I was taking it on faith that this is from Ben Armstrong, who's the director of the Industrial Performance Center.

 

0:11:11.8 AS: Yeah. You just mentioned something that I was just recently talking with another person about, and that was, one of the downsides of an aging workforce is that you're losing really senior people and you're replacing 'em with people that may not have the skills. Also, US kind of is notorious in America for a declining education. And with education coming down for the last 30 years or so, it's also hard to find, let's say, engineers and people that... There's not a deep market in some of these places where there's need. So that's a real challenge that businesses are facing.

 

0:11:55.2 JS: It is. Yeah.

 

0:11:56.3 JS: Yeah. And now what they're doing is they're looking at manufacturing from that standpoint. They're now acknowledging that the scarce resource is the human. And we have to actually build, if we're gonna automate, we have to build those processes around people. And that's... I'm gonna just read you a description here. There's, I think you heard of Technology 4.0, where they talked about putting sensors all over the place and having smart factories and that kind of thing.

 

0:12:27.7 AS: Yeah.

 

0:12:28.3 JS: Well, we now have something called Industry 5.0, and I'm just trying to get the wording here 'cause this has been around for a couple years, but it's on the EU website. It says it's "a vision that places the wellbeing of the worker at the center of the production process and uses new technologies to provide prosperity beyond jobs and growth while respecting the production limits of the planet." So they're really trying to center technology around that so you're not doing your sort of environmental and your DEI and all that independently of your production, it's all integrated part of it, which is I think something I'm sure Dr. Deming would have advocated.

 

0:13:17.8 AS: I'm still kind of fascinated by the productivity, and I just look at here in Asia, productivity is just rising. Education levels are rising. Engineering skills are rising. Competency in certain areas, specialties is just rising. And I oftentimes, I think that one of the things why this... One of the reasons why this is a good discussion that we're having is because in the West, in particular in the US, there's a new challenge. And that is how do you bring business... How do you bring jobs back to the economy when you're facing a very, very different workforce from when, let's say I left Ohio in 1985, roughly. It's a very different workforce nowadays.

 

0:14:07.1 JS: Well, yeah. And I think a lot of the offshoring arguments were about, well, we'll keep the smart jobs here 'cause we're all well educated and we'll export the low paying, less skilled jobs abroad, and we'll all win. But now, of course, we're finding that people overseas are getting darn well educated, so you can't have a more expensive labor force and have people that maybe aren't even as well educated.

 

0:14:40.0 AS: Yeah.

 

0:14:40.2 JS: So it's... Yeah, I think the West is in a very tight spot right now.

 

0:14:45.3 AS: Yeah. So speaking of automation and technology, I was just typing as you were speaking, and looking at productivity, it says... I was using ChatGPT and that says, US productivity growth average 2.7 annually from 2000 to 2007, but slowed to 1.4% from 2007 to 2019. There was a brief pickup in 2020, and then it's been slow since then. And they talked about this productivity paradox that I think is what you're referencing what Ben is saying.

 

0:15:21.3 JS: Solow's paradox? Yeah.

 

0:15:22.6 AS: Yeah. So that's interesting. Yep.

 

0:15:25.8 JS: Yeah. Solow's paradox, what does it say, that you can see the impact of technology everywhere except in the productivity numbers. I think that's what he said.

 

0:15:36.8 AS: Yeah, so he said that...

 

0:15:37.2 JS: He said that by the way in 1987. So anyway, yeah, maybe we're slow learners or something like that. But no, that's really fascinating. But I think that there's a difference between GDP growth and the growth of productivity in manufacturing. I think probably the ones that Ben Armstrong quoted were a little closer to actual manufacturing. But right now, GDP includes financial intermediation, it includes... If you own a home in North America, they include imputed rent, the rent you would have been paying as part of the GDP. So I think there's a bit of inflation, I guess, in the GDP over the years. So I think we have to take that sometimes with a little bit of a grain of salt and look a little more carefully at what the numbers are telling us.

 

0:16:32.8 AS: Yeah. The main ways that we typically look at it outside of GDP is like non-farm productivity, like non-farm worker, what's the output? And the other one is total factor of productivity. So yeah, GDP can be quite distorted for sure.

 

0:16:50.4 JS: Yeah, for sure. And anyway, and also just taking GDP per worker can be a very misleading number.

 

0:17:00.5 AS: Yeah.

 

0:17:01.3 JS: But anyway, yeah, it's fascinating. But again, the myth is... This myth that technology will solve everything is all over the place. I think with autonomous vehicles, the idea of being able to replace drivers is a just enormous economic cherry, I guess, that everybody wants to pick. You think about it what that would mean if you could... If you bought a car and then you could rent it out as a taxi at night, or what it would do to Uber if they didn't have to have people driving the cars. It's just enormous. But it's been very, very frustrating to get to that point. And when you look at a lot of the forecasts, it's still a long way away. So I think we have to be more conservative about that and talk about more the benefits really of technology and people working together. And I think the automatic driving features they have on cars now are fantastic. You can make a car a lot safer. You can slow down if you're tailgating somebody, it alerts you of just even the simple things that if there's a car to your left passing on the freeway, you get an alert, and that's... This is all really, really good stuff, but I still think that the self-driving part is maybe longer off than people think.

 

0:18:39.4 AS: Yeah. I think regulators too get panicked and then people want action when there's an accident or something like that. You also mentioned something about the computing power that's required for some of what this is doing, and that's a fascinating topic because it's funny, it's just amazing how much computing power is really going to be required over the next 10, 20 years.

 

0:19:05.0 JS: Yeah. I think there's a bell curve around some of this stuff, and I'm just gonna talk and I'm gonna jump to regenerative AI, which everybody is talking about. And they're saying, how long before I can have regenerative AI write a document that we could actually be held liable for? It can write documents, but you can't trust it. So they keep trying to improve it, but it's a kind of an exponential problem here where the wider you make your bell curve, the exponentially more power you need to do that. To the point where Microsoft is talking about buying Three Mile Island nuclear plant and rebuilding it to power all this AI stuff. So it's just phenomenal amount of power. I think that's somewhat... I don't know, relying purely on more computer power seems like it might not be a winning strategy.

 

0:20:13.3 AS: Yeah. It's the regenerative AI and all that's going on is also... I like to say when proponents talk about it and its strengths, which it definitely has strengths, I'm not arguing against that, I use ChatGPT almost every day. And I can say I used to have an editor sit next to me a lot of times and now I don't need that because I can go back and forth. But what I can say is that when a proponent of AI gets accused of murder and they're innocent and they're gonna go before a judge, is that proponent of AI gonna use purely AI to build their defense or would they prefer to have a lawyer who's using AI as a tool. I think I would argue we're far away from the trust level of being able to walk in there and say, I trust AI to get me out of this situation that I've been accused of murder and I'm innocent and it can get me out. There's no way any of the proponents of AI would take on that I would argue.

 

0:21:23.3 JS: Yeah. Well, it's interesting. I very recently had to write an affidavit and my lawyer was being a little slow on it, so I tried ChatGPT just for the heck of it and I created what I thought was pretty convincing. I gave it the facts and it gave a pretty convincing sounding affidavit, but then the lawyer did it and I saw what she did and it was so much... She had it... It was almost a human touch to it. It almost looked a little less like an affidavit. It was more of a sort of a document that had some meaning to it. That was an eyeopener for me.

 

0:22:10.8 AS: Yeah. Yeah. Interesting.

 

0:22:13.6 JS: But anyway, yeah, I'm wondering if we could jump back to automation and manufacturing because there's a story I wanted to share with you about some of the followers here of Toyota and, of course, company that's strongly dedicated to Deming's principles as well. And this is a company called Parker Hannifin. And what they do, and this is in the Lean tradition, is they're very conservative about adopting robots or any kind of automation. And they realize, when you bring in robots, you're bringing in software, you have to upgrade the software, you have to maintain it, you gotta train people, there's a risk of obsolescence or whatever, there's all that risk. So you really wanna be very, very careful. So what they do at Parker is you have to, but if you're gonna present a business case for a robot, you gotta be able to show that that's the only way that you can get the improvements you want.

 

0:23:22.3 JS: And by the way, you gotta have a target. You don't just say I wanna automate this, you say I wanna make this process better, here's how. So I got an example from Stephen Moore who's... He's retired now, but he was the VP I think of operations. So he was certainly the top person in terms of all the Lean initiatives that they did. But he told me and gave me an example. He said that somebody came to them, they had a cell with three people and they wanted to use the robot, one, so that they could reduce from three to two because they needed another person in another area. And secondly, there was a safety problem with that cell with loading and unloading the machines. So they came to Stephen and Stephen said, okay, let's divide our team into two groups. One group can sort out, plan the robotic implementation, how it's going to be done. The other group is gonna see if they can achieve the same objectives without a robot. So by the end of the week, the team that was without the robot team was able to achieve both objectives. They were able to reduce it down to two people and they solved the safety problem over the loading. So just by thinking it out by really going deeply into the process, they were able to do everything that people expected the automation to do.

 

0:24:58.3 JS: So that is a philosophy, I think is a lesson I think to anybody that's automating. 'Cause remember, we've got lots of companies that are just thinking about replacing people, whereas Parker Hannifin is talking about increasing the value of processes. They're concerned about safety here as well as headcount. And very often, they're looking at processes to improve the quality. So we've gotta look with a broader lens.

 

0:25:29.1 AS: That's fascinating. And for those people that don't know Parker Hannifin, I had mentioned before that was one of my father's big accounts when he was working in DuPont in the old days.

 

0:25:37.4 JS: Oh yeah.

 

0:25:38.4 AS: He was living in Cleveland. We were living... I grew up near Cleveland. But Parker Hannifin is about a $77 billion company. It's got a net profit margin of 14% versus the industry average of about 11%, which is already pretty high. And that's pretty impressive. But what's really impressive about Parker Hannifin is that it is the 11th most... If you look at all companies in America and you ask them which has been consecutively producing dividends since 1957, so about 66 years, Parker Hannifin has been producing an annual dividend. And in fact, they've been increasing that dividend ever so slightly every single year for 66 years. That is a very, very impressive feat. And very few companies are out there. In fact, only 10 companies are better than that, that are listed in the stock market. So there's some fun information from a finance guy.

 

0:26:35.4 JS: Well, of course, and the fact they've... We talked about some of the productivity challenges in the last while and the fact that they've sustained this. We're talking post 2010 when the productivity has been slowing down, and they've clearly kept things going, which is... We've seen that with Toyota and a lot of companies that follow these principles. It's a way of sustainable growth.

 

0:27:03.3 AS: Yeah. One of the things about Toyota is it's so fascinating is that they're not sold on automation, they're sold on improving processes. And if automation can help that, that's impressive. That do it, but otherwise, fix the process before you automate.

 

0:27:21.5 JS: Absolutely. And that's again I think this isolation of operations is a sort of a black box of the corporation where people sit in the boardroom and they just say to the operations person, well, that's your problem, solve it. We don't wanna know about it. So they see things outside the box in a sort of a financial lens. I think we talked about that in myth two.

 

0:27:45.2 AS: Yeah.

 

0:27:45.8 JS: Whereas the things that go on with process actually defy financial logic. We're improving quality and productivity and timeline very often too, delivery at the same time.

 

0:28:03.3 AS: Yeah.

 

0:28:04.2 JS: 'Cause it's a better process. It's simpler, it's better and it's a powerful concept. But I think a lot of people that are not inside process or not inside operations, aren't aware of that.

 

0:28:17.8 AS: Yeah. So how would you sum up what you want people to take away from this discussion?

 

0:28:25.3 JS: Okay. Well, I think there are a few, I guess, bullet points I would emphasise. First of all, there's no question that technology has potential to help companies get significant productivity gains. But you shouldn't see it as a technology-only solution, I think again like we were saying, you have to look at it as a way of improving processes and that's where the power of it really is. I think it shouldn't be about replacing people, but it should be combining the strengths of people and the strengths of technology. I think that's where a lot of the high potential is right now. But that means you've got to know how to optimize your process. And that's what Dr. Deming, what the Lean folks all work very hard on. And I kind of think this is a time when companies maybe need to think more seriously about that. And finally, last but not least, I think one of the wonderful things about technology is you can use it to remove the dull, dangerous aspects of work and you can make the jobs more, you know, safer and more human, I guess, more friendly for human workers by using technology. So I think that's a big hope there.

 

0:29:55.5 AS: Well, that's a great discussion of myth number five, The Myth of Tech Omnipotence. Jacob, on behalf of everyone at the Deming Institute, I wanna thank you again for this discussion. And for listeners, remember to go to deming.org to continue your journey. You can find Jacob's book Productivity Reimagined at jacobstoller.com. This is your host, Andrew Stotz, and I'll leave you with one of my favorite quotes from Dr. Deming and I hope you're living it right now. "People are entitled to joy in work."