GA 355 | The Relationship Between Machines and Humans with Prasad Akella

And so the opportunity here is to take this 345 million people and dramatically empower them in a way, in a very fundamental data production style way of empowering people with data so you can move the entire group up and just change how productivity works around the world. And that's really why I'm so fascinated with people.

Today. I'm excited to welcome Prasad Akela to the show. Now, Prasad works for a company called Dristy.com and what we're going to focus on today is kind of this relationship between machines and humans. Now, it's a topic that we've addressed in the past. Industry 4.0, Factory of the Future has many different words and names, but today was an interesting conversation because we talked about that relationship between machine intelligence and human intelligence and how the two can actually work together in harmony. They don't have to be at odds with one another. And so that's where we spend quite a bit of time. And we also. Prasad has quite a bit of knowledge on, on Lean and he's an industrial engineer by training, so he gets what we're all trying to do in Lean. And so towards the end we spent some time talking about what he's seeing out there with their customers as far as Lean and, and why some are more successful than others. So that was also interesting. And then we wrap things up with maybe just a glimpse into the next 12 to 24 months, you know, from Prasad's perspective. On what? On what. On what he sees the future of Lean looking like. So really interesting conversation. Show notes for this episode can over@gembapodcast.com just look for episode 355 again, gembapodcast.com and look for episode 355. Okay, enough for me. Let's get to the show. Prasad, welcome to the show. How are you?

I'm doing well, Ron, how about yourself?

I'm doing great, Doing great. Thank you. Thank you. All right, well, I appreciate you taking some time to come onto the show. I'm really interested in the topic that we're going to be focusing on. But you know, Prasad, towards the beginning all of our episodes, we'd like our guests to share a quote. So do you have one?

I do. The quote is actually from Edward Deming and the quote basically says, without data, you're just another person with an opinion. Now, I'd like to extend that code by saying without an opinion or in TPS parlance, a hypothesis, just another person with data. We'll dig through this as we go along. As you can see data means a lot to me. And we'll talk about it.

Yeah, perfect. I love it, I love it. Can't go wrong with Deming. Not on this podcast at least.

Not at all. In fact, I picked one for that. Exactly why I picked it.

Yeah, yeah, yeah. Good, good, good. All right, well, listen, tell us a little bit about your background, maybe how you got interested in all this kind of continuous improvement type stuff, and then tell us what you're up to these days.

Sure, sure. So I'm Indian by origin. I came to the US to get my PhD in robotics. I've been in robotics for 35 years now. And really, I guess my kids asked me when they were growing up, how do you decide to be an engineer? And I tell them it must be my DNA. My dad's an engineer, my grandfather's an engineer, one of the first engineers in India. And I can't remember a day when I did not want to be an engineer. So I guess I was fortunate enough to stumble into a profession. And I've lived around cars from the pictures of me from 3 years old with cars, little toy cars in those days on my bed, asleep. So I guess cars therefore leads you to building stuff. And when you build stuff, you land up in manufacturing. So robotics, electromechanical systems manufacturing just seemed to all make sense to me. So that's the quick story. But more seriously, I, when I graduated, I went to Japan. I worked at the National Research Labs, the Japanese National Research Labs outside Tokyo, because in 92, I thought about robotics. It was synonymous with Japan. Figured if that's the case, better be there. So spent two years, had a wonderful time, did some good work, and had built some great friends, made some great friendships, and then moved back to gm, to the us. GM actually had an interesting proposition. They had a potential OSHA fine of a billion dollars if they didn't address an ergonomic problem on the plant floor. That became my journey in working with some of the best people in the world, putting together what today you and I refer to as cobots, collaborative robots. We pioneered that space. We sorted out how do you take the cognition of the human being as you assemble a car and the physical power amplification of a robot? How do you safely put them together to actually do things without cages around them on the plant floor? Imagine picking up as simple as a battery one a minute and putting in the car. It's back breaking work. It gets worse when you're picking up 150 pound fully built cockpit and trying to get it through the door opening into the car to assemble it to the A pillar. And essentially, if you could take robotics, you could take virtual haptics and you could build cobots. And that's what we pioneered. 94, 2000, that got me fascinated. Here's a pure R and D guy and suddenly having to commercialize technology coming off the floor. And so that got me curious in the business side. GM put me through business school as well. So I sort of in 2002, landed up in the valley pioneering a second space. This time we looked at social networks, business social networks. How do you figure out who knows who? So you can find good speakers, or I can find a good employee, or I can find a good job, or I can sell better. Eventually, LinkedIn took that market. But what that got me into was data, the importance of data sets and how when you have a data set of who knows who, you can have a Facebook, you can have other social products that Twitter, other social products that let you use your network in different ways. A trillion dollar market today at Drishti, what we're doing is we're taking manufacturing and data and putting it all back together. That is the journey in getting to Drishti, to thinking about manufacturing not just as a physical activity, but the bits that drive the atoms. So my first half of my life was just atoms driving atoms, robots moving stuff and welding stuff and stuff of that sort. And now I'm looking at the data behind it that lets you drive those systems and in particular, understanding people in these systems. And we'll talk more about people, but really Drishti is about not the robotic side of manufacturing, but the people side of manufacturing.

So. Okay, well, those seem to be, you know, a skeptic might say they conflict, right? Like, why do you need people if you have robots? Obviously you need someone to program the robots, right?

I guess.

And to maintain the robots.

But.

But do we need to think still with robots and so maybe talk about that relationship between automation and, and the human being and our purpose and that system?

Yeah, that's an awesome question. There's a question I asked myself five years ago sitting in the same room I'm in right now. So let me start with a few facts. There are about 345 million people, 3, 4, 5 million people around the world working on production lines. The sum total of robots is 2 million. And we are producing about 300,000 robots globally today. And even with a displacement rate of 5.6 people per robot, so every robot essentially takes 5.6 people out. You're Decades, almost a century away from replacing the 345 million people. It's just a sheer technology problem. Right? So the second thing, as you correctly pointed out is we don't, you know, you need people to program these robots. And every time you change the line to go from product A to product B, you're going to reprogram them and you need tooling. And it's not as simple as people think it to be. In fact, if you are like me, you have an iPhone in your pocket, the chances are 141 people, 141 people on a line manually assembled the product in its final assembly, took the glass, took the battery, took the board, the buttons, all of that stuff, and manually put it together on a final assembly line. So while automation is here to stay, don't get me wrong on that at all, the 2 million robots will creep up over time. The fact is that there's a bulk of manufacturing is done by human beings today. And in some senses, the thing that shook me when I looked at the space was nothing has really changed since the time of Henry Ford. Right. It's a little staggering to make that kind of comment because just think about simple thing. You had Frederick Taylor and Frank and Lillian Gilbert doing time in motion studies. What did they do? They had a stopwatch and a person trying to figure out the elemental steps and trying to figure out what was going on in the production line so they could standardize the line a good 100, what, 1908. So 112 years later, nothing has changed. You still have an industrial engineer, gets a fancy industrial engineering degree from some of the best schools in the world. What does he or she land up doing? Spending 60% of the time measuring things on the floor. The opportunity here is to take this 345 million people and dramatically empower them in a way, in a very fundamental trader production style way of empowering people with data so you can move the entire group up and just change how productivity works around the world. And that's really why I'm so fascinated with people. I saw that in the plant flow when you were building cobots. It appears like putting that door, you know, that instrument cockpit into the door opening is an easy problem. It is very hard. These people are almost like artists. They make it look simple. It is not that simple. And that's really what fascinates me about this whole human and machine and robot story that I'm talking about.

Yeah, yeah. So I guess talk a little bit more. I'm just, I'm, I'm genuinely interested in hearing about how you think the next, I don't know, 10 years are going to look in the modern day factory, you know. Yeah, what is that? I mean, obviously robots exist that have existed for decades, right. And they're not going anywhere. But when we think about Lean and we think about the power of Lean is really in our, in, in the human person's ability to think, to think critically, to think scientifically in order to, you know, move the needle. Are we going to get, we already are at a situation where machines can think. It's called artificial intelligence. Right. And they can make decisions, if you will. They can program, machines can program. So where do you see the, the relationship going? I mean, you see the movies where the machines take over the world, right? And you know, I mean, I don't know if that's ever going to be, you know, accurate, but who knows, right? In 20, 30 years, what would it look like? Will we have, will we have machines managing each other or is there, is there always going to be people, you know, at the top?

Yeah, it's a fascinating question. I think what you're going to see is a blend of man and machine. Who takes the leadership role? Depends on the situation. I'll give you a little example. You and I write email and write documents and whatever else. We have a spell checker continuously running behind the scenes, a grammar checker. And what that does is ensures that the finished product, the email message going out, looks supremely polished. Ron and Prasad look like star writers. We could be writing for the New Yorker or whatever else. And it is that combination of man and machine working together that elevates us to the point that we could potentially aspire to writing for New Yorker. I could aspire. It's the same thing, I think that we will see increasingly on the floor where the cognitive burden is going to change. The machine is going to gather data at scales that are humanly impossible. It's going to gather data of every single unit going down the line. What the machine is going to do is going to surface at scale abnormalities. This is almost like think of standardized work. You establish a baseline and now any deviation from the standardized work becomes a signal which the system will push to a human being. It'll push it with insights saying, hey, my hypothesis is this, but the ultimate determination for a while to come is going to be made by the human.

Being for a while to come. That gets my question. Will there come a point to where the machine has a hypothesis and the machine then tests the own hypotheses? Scientifically thinking we say, what's your next step? What do you think is going to happen? What actually happened? What did you learn? I guess my question is with a machine be able to form a question. Will the machine come up with their own prediction, run the experiment if you will and then see what they learn and then adjust accordingly.

Yeah, there's such a machine, it's called an airplane. Today when the pilot sets that autopilot on, the machine flies itself. It's got the benefit, it's up in the air. So there's a lot of latitude for error. Now self driving cars tend to replicate that on the road where you have much less latitude. Actually that day has already arrived. We don't recognize it because there's two human beings sitting in the seat called the pilot and the co pilot. But 85% of the time the plane flies itself. So I have no doubt that that's going to be the future. The question is in more the complexity of the scenario is what determines when you get to that point. Right. So the difference between the plane and the car is the environment in which the car lives. Much more complex random events, dog showing up the guy beside making a mistake. So if, if the roads today were only with automated cars, no doubt Google could have self driving cars or tunnels.

But Elon creates. Right. That are on.

Yeah, yeah, totally. You can put me in a tunnel, you put me the fast tunnel, the slow tunnel and you can take and I can move from one tunnel. All of that can be absolutely done today by machines. It's when you have the randomness of human beings arrive that you have the challenge. And that's the fundamental challenge that you see on the plant floor. If you were building that iPhone and you were given the same set of instructions, you will do it differently than I will. That's human nature. That's the beauty of a human being. You will do it differently at 8 o' clock in the morning than you do it at noon than at 4 o'. Clock. We're just built to be variational machines.

Yeah.

And that is where the challenge comes. That's why I think that's the part that the machine has a problem dealing with is this level of variation and gross variations. It can pick up the more subtler variations. I think you'll need human beings.

Yeah, yeah. So, so, so let's talk a little bit then about efficiency, you know, and improving efficiency. So I mean we've touched on it a little bit already. But the, I guess the, you know, the assumption in manufacturing today that they only. The only way to improve efficiency is to automate. So, first of all, do you agree with that statement? And if not, how would you tweak it?

Yeah, it's certainly a way to. It's not the only way. Okay. That's the position I take on that. And I think in the world of manufacturing, we made this assumption that we really tapped out on what we can do together as people and machines. And I think there's massive opportunity sitting in front of us. We've seen that with our own customers. When the machine can look at 100,000 cycles, 100,000 units, it can understand in a statistical sense what the variability is, where the outliers are, where people should focus their attention. Now, suddenly the machine is looking at this 100,000 and telling you, here's a 50. That matter. Yeah, right. Now, if you think about this whole notion of the turtle production system, of continuously running experiments, of scientific experiments, of then tweaking the experiments, as you learn more and more, you suddenly enable that at scale. You're not looking at a single station, you're looking at an entire line. You're looking at systems, just the way you can do all of that changes. The results are stunning. We find with our customers. In fact, our customers report to us that you can reduce defect rates by 50%, even at the single part per million defect rate level. Think about this. You've written this asymptote up. You're now at 9 parts per million defect rate, and you're trying to get to eight. And it gets asymptotically more expensive, both in effort and time, to get from 9 to 8. Now imagine if, like an NFL quarterback who threw an interception, you can actually look at the instant replay. You can now suddenly sort out and ask yourself, oh, why did. Why did that happen? And the human brain, being as capable as it is, can look at that and say, I get it. I tried to come in and turn the screw from this direction. If I just change the angle by 2 degrees, I'll be fine. That same feedback loop that you talk about in Toyota production system, that we're just digitizing and doing much faster and saying, hey, Mr. Line Associate, this defect crept in. Let's look at the video together and replay it and say, okay, I get what went wrong. Never happened. 50% reduction in rates. That's the kind of stuff you can look at half a million cycles and determine where your bottlenecks are on the line. And you can identify which particular step in your standardized work is actually driving most variability. Now you can focus your attention on just that one step and say, oh God, I can see I'm trying to put a label on this box that's going out, a computer that's going out. And I've told my people to get it absolutely parallel to the edge and I've given them no tooling to do it. So they're struggling over this now. Suddenly it becomes possible to fix that problem and you drive your variability down. That is the power, I think, where we can dramatically shift where people operate at on that performance curve. And it's really, you're not asking more of the person, you're just empowering them with better tools and better data.

So I have college age kids and kids that are younger than that as well and haven't even gotten to college yet. Where do you see the future for our children as far as career decisions? And you know, because you know, there might be some jobs that don't exist 10 years from now that are currently there. Right. But some new jobs may appear that don't exist today. Like where do you see the future of education of our children and, and the future of just employment for our, you know, our, our young, our young, you know, teenagers today or our 20 year olds today? Like what, what are your thoughts there?

Yeah, yeah. So I have teenage, I have college kids at home as well and there's a conversation we have at the dinner table. I think the way I framed it up for them was data literacy is going to be increasingly the single most important thing skill that you will need in life. And if you, if you choose to be left behind on the data literacy curve, you're toast.

So define data literacy. What does that mean?

Yeah, so if I think about an industrial engineer even three years ago, what you needed to do. So think about a line balance problem. Just pick a simple problem that the audience relates to. You could go on the line, you could take 10 samples, you could eliminate the, the largest and the smallest, you get your set and you know how to compute an average on that. And you basically say, my average cycle time is this. Now suddenly when you have distribution functions, probably distribution functions, because you got a measurement on every single unit that went down that line, you start seeing long tails. There are cycles that take much longer than the tak time or the target time that you've got now the question is, what's going on in that long tail? Because there's information sitting there. It could be that the material flow, you have a material flow problem, you better diagnose that or you had an engineering change. Order go through, that's making it harder for the person to get the job done within the allocated time. Or you got a new person on the job. And now to be able to look at these distributions and actually extract from them the nugget, if you will, that lets you improve the process is going to become the critical skill. Right. And this is actually going to pervade society. Because if I'm a call center app taking a call on Zappos or something, maybe Zappos is a bad example because they can take as long as they want to answer the call of their focus on the customer. But any other call center, it's how long did it take? How long did it take to answer the question, what is your first. How quickly do you deal with the first call? You will have more reports coming your way that you got to understand and say, what's my behavior and how do I do what I need to do? In this world where data is being pushed at me, both is just at intuitive level, understanding the data and trying to figure out how do I make use of the data is what I think of as data literacy. You don't need to know how to run.

You don't have to be an industrial engineer, really.

No.

Right, yeah.

This is just, this is a very pervasive problem. And what's going to happen is the machines are going to do more of the grunt work that you and I do today. The kids will not have to do it going forward. What they will have to do is interpret the data.

Yeah.

And that's where most of the jobs are going to move, because just automation will keep moving us up and up the value chain and we've got to be at that point where we can tap into it and benefit from it.

Yeah.

And that's why I at least guide my kids to make sure that their math, basic math fundamentals are in place so they can interpret all this stuff.

Yeah.

You're a technician on the floor. This directly impacts you. Right. If robots going down, the lines going down and robots going down frequently. Now the question is why? And you need to sort it out.

So I am curious, you know, I know that you work with a lot of companies who are practicing lean, so I'm curious to get your perspective on what you're seeing out there with your customers who are practicing. Where do you see the lean journey right now? And, and what do you think? You know, what do you think the next, maybe the next, I don't know, 12 to 24 months look like?

Yeah. So I, I'll break the two problems apart too. What do I see? And what, what does the next 24 months look like? Okay, I see every company I've been to in China and Asia and in the us almost every company practices lean. And if they don't practice lean, at least they profess to practice lean. And almost every plant will have all the signage for lean up on the rafters. It's just amazing that the Toyota production system, lean, have made that kind of headway in the industry now. In direct contrast to that, true professionals of Lean are few and far between. The question then becomes, what's this massive gap that I'm seeing between the two? I think it comes down to a philosophy question, which is when you look at Toyota production system, it's all about people management. It's about empowering people. It's about understanding that the person on the front line understands the job and the challenges and such like, and therefore is best positioned to come up with a solution. So it's very much of a bottom up, empowering perspective. But I think what I see when people start implementing lean is they take the tools of lean, but they keep the mentality of a top down management system. Therefore, I, the process engineer, know more than you, the line associate, and I will tell you what to do. And I really don't give a damn about what you think or what you know and how you feel. Yeah. And I think that is the massive cultural gap that I see.

Yeah. In the end it's, it's, how do we treat one another, how do we respect one another, how do we honor one another? You know, if you will. Right. Maybe is what separates the, the pretenders from the, from the.

I'm not gonna, I'm not gonna go there, call them the pretenders, but I think they genuinely want to do what's right. But I don't think. Yeah, the full ethos or the philosophy has been absorbed.

Right. Yeah. Yeah.

So what, so what does this mean for my customers or for companies that talk to us? I think there's a genuine desire to be Toyota like and therefore there's a desire to jump on what I call the lean maturity curve. Literally, instead of crawling and walking and then running, the desire is to go run at the get go. What I realized is there isn't an easy path. You've got to pay your dues. So if I take something as very fundamental as standard work, it's quite surprising how often if you ask them to give you a definition of standard work, most of them cannot give you a definition of standard work, right? It's just as fundamental as that. Because if you don't have standard work, then when you're trying to do your training, what are you training your people to? When you're trying to do root cause analysis, how do you find deviations from that? Because if everybody's doing it every which way, there is no standard path to find deviations from, right? So it just starts as. Just as fundamental away as that. And then when you basically get to the next level saying, okay, I'm trying to measure cycle time, and I'm thinking of cycle time as the first predictor of lack of control on the process, right? I mean, if you have a lot of variation in cycle time, then you know you're not controlling your process. It turns out that the absence of standard work hurts you there. And now the question is, okay, how do I fix this problem? Yeah, I want to run a Kaizen event, but what's my Kaizen event going to do for me? Right. I don't know what baseline to work against. And it gets even more interesting because when you look within the cycle and you look at the steps, if you really don't know what you're trying to do, you can't manage quality. If I'm going sequence ABC and you're going bca, who knows? I might be introducing issues and you may be doing the right way and I'm doing the wrong way. Just don't know. And off we go to the races. This is the kind of thing I see over the next 24 months, even though we're a small company, recognizing that we actually have a fundamental understanding of the issue, have tools to measure and guide our customers, to increasingly take more customers down this path, saying, let's get the fundamentals in place. Let's figure out how you empower your workers and how you start thinking about workflow value streams and then how do you actually, from that base build up so you actually get to Pokeyoke, where you can drive the quality aspect of what you've got. It's a journey that I think of ourselves taking, working with our customers to go down, which is a tough one, but that's why we're here.

Nice.

Nice.

Well, good. Well, this has been fun. You know, it's always interesting kind of speculating on the future and also, you know, reflecting on where we've been. And, you know, that's why I like sometimes when I like to what's the future? I like to, I don't want to say five years, that's too far. You know, maybe 12 months, 24 months, you know, maybe this is more, more realistic, right. To make some predictions on. So. But I really enjoyed it. Thank you for, for coming on. Why don't we wrap things up Prasad with you sharing maybe just some final words of wisdom and then defin definitely tell folks how they can connect with you.

Got it, got it. I want to end actually with a little thought provoking idea for the audience. McKinsey speaks about what they call pilot purgatory in the industry 4.0 world. The idea basically being that all of the companies are curious, they want to test the waters a little bit, drop their toes in there, but they never really get to broad deployments. That's what McKinsey calls pilot purgatory because you're stuck in pilots. As I've looked at this whole problem more the last few years, what's become very clear to me is that companies I think need to be more disciplined about their processing. Basically say if I'm investing I the company, I'm investing in this effort. With a company like Drishti, I better have a long term view of what success will look like. So let's assume Drishti as the example, but any company that they're working with in this area, let's assume success and behind success, let's plan the broader deployment because at the end of the day it's a financial thing. The success has to translate into actual financial benefits to the customer, to the company. And so it starts with even empowering that team. You have lots of industry photo teams, you have factory of the future teams that each one has a different name process R and D. Now that team needs to think and be empowered to taking any new technology that they bring in, once proven both technology and economically down across their 230 plants around the world. That's the seed I want to plant in people's minds that true success comes from thinking of scale. It doesn't come from just simply testing the waters with 10 different pilots running in your labs. Doesn't help anybody. It doesn't help me because I've got investors asking, okay, where's, where's progress here? Right? And so I would love to challenge your audience to think about this, to plan for success and for scale and actually narrow down what they work on so they can actually support this kind of scale coming down the pike. I think that's good for the industry, it'll be good for them, it'll be good for companies like myself. So that's the thought provoking comment I want to lay out there in terms of reaching us. We are at Drishti.com,-R-I S H-T-I.com we're on LinkedIn, we're, you know, on Twitter, all of the usual social handles. You can find us there. Okay. And if you want to drop us mail, it's infodrishti.com.

Okay. And are you on LinkedIn yourself personally as well?

Absolutely. P. Akela, you should be able to find me there. I'm happy to engage. Just reach out to me anytime. And my email is P. Akela rishti dot com.

Got it? Yeah, we'll link to it in the show notes. So let me see, this is going to be going to be episode 355. So GembaPodcast.com go to episode 355. We'll have you send us all those links, Prasad, and we'll put them into the in the show notes and folks can go there to get all the good stuff. But, but yeah, thanks again for coming on. It was fun. I enjoyed it. Keep up the great work and maybe we'll connect again down the road sometime.

Absolutely. This time I'll come on with one of my customers.

Okay.

And that'll be a fun conversation to have as well. So Ron, thank you for hosting me today. I've enjoyed it totally and look forward to the next one.

Fantastic. All right, thank you.

Take care. Bye bye.

Thanks for listening. Whether you've been on the continuous improvement journey for many years or perhaps you're just getting started, Gemba Academy is here to support you. And while we're best known for our more than 1500 Lean and Six Sigma teaching and virtual tour videos, we also have a team of experienced Lean and Six Sigma practitioners available for one on one coaching, as well as a variety of Lean and Six Sigma certification options. To learn more and to schedule a demo, head on over to gembaacademy.com.