The Manufacturing Executive
The Manufacturing Executive

Episode 103 · 5 months ago

Empowering Workers and Improving Workflow Through Video Data


People often worry that technology and automation will replace human workers.  

But what if the most powerful path forward is empowering workers with technology for better effectiveness and efficiency, instead of replacing them? 

In this episode, Dr. Prasad Akella, the Founder and CEO of Drishti, shares his view on the future of manufacturing. A future where video data not only improves the quality of a product, but also empowers workers on the plant floor and strengthens employee retention. 

Join us as we discuss:

  •  How technology empowers humans to be more effective
  •  What using data can do for your productivity
  •  Why using technology leads to better employee retention

I think they're sitting in the frontier of something new. I believe that in the NOCTO distant future, cameras will be an ubiquitous and every plant floor, everywhere, and the world will start seeing, and I play words and on seeing from all of this data and we will transform manufacturing in a very fundamental way. Welcome to the manufacturing executive podcast, where we explore the strategies and experiences that are driving midsize manufacturers forward. Here you'll discover new insights from passionate manufacturing leaders who have compelling stories to share about their successes and struggles, and you'll learn from B tob sales and marketing experts about how to apply actionable business development strategies inside your business. Let's get into the show. Welcome to another episode of the Manufacturing Executive podcast. I'm Joe Sullivan, your host and a CO founder of the Industrial Marketing Agency guerrilla seventy six. A few weeks ago, my seven year old daughter, grace, came home from school telling me about how her first grade teacher had brought in something called a typewriter that day. Grace went on to explain what it was and how it worked and how long long ago, people used to use these things instead of computers. While I entertained grace, I also thought to my thirty nine year old self, Huh, I'm kind of spoiled. I just missed that era and I can't really imagine typing without the ability to correct my mistakes. That happened just about every other word, and even those little red and green squiggly lines that point out my spelling mistakes and grammatical errors are pretty darn helpful at empowering me to produce a better end product. Well, my guest today is going to talk about this same idea, but in the setting of a manufacturing production environment.

In fact, the Microsoft words spell check. Analogy is one he'll share in this conversation to illustrate a key point, and that is technology is not there to replace people. It's there to empower them to be more effective and efficient and to be a contributor in producing a better final product. Let me introduce him. Dr Prasad, a Kella, founder and CEO of Drishti, is creating his third massive market category that uses technology to extend human capabilities. In the N S, Prasade led the General Motors team that built the world's first collaborative robots, or cobots, projected to be a twelve billion dollar market by two thousand and twenty five in the early two thousands, is cofounder of the social networking pioneer spoke. He envisioned and helped build the first massive social graph, a category now worth trillions today. At DRISHTI, he's working to combine find the cognition of AI with the flexibility of humans and factories in the form of aipowered the production Prasade, welcome to the show. Thank you, Joe, for having me. I'm looking forward to a fun conversation here. I am as well. I have. We had a had a great sort of precall a few weeks back and I'm really excited about what you guys are doing and and happy to get you on the show here to share your insights and even some of your product with our listeners here, because I think you're you're a disruptor here and doing some really interesting stuff. So so, Prasade, I can't overlook the fact that you were involved in the development of some of the first cobots with GM earlier in your career. I'd love for you to tell us more about your experience there. And how it plays into the conversation that we're about to have today. Yeah, you know, I think the study starts back as an Undergrad at the Indian Institute or technology and I decided to work into bodics. I like this interplay between hardware and software and changing behaviors, if you will, and that continued through my PhD at Stanford.

Was Designing robotic hands. But what happened at GM was that my world turned completely upside down. Osha wanted us, wanted to find GM because people were messing up their backs on the production floor and GM hired me to come in and try and solve the problem. And it's one of those accidents in one's professional career that truly sets direction and I really appreciate that. And so the first thing I learned was management. I had a phenomenal management chain, Jim well, Steve Holland, Jimmy, sue, Gary Calgar, who ran all of GM manufacturing eventually, and they taught me what good management was. I had some phenomenal colleagues from northwestern and Berkeley that we work together on. But from a production perspective, from looking at what's happening today at the race da what I realized that there were thousands of people on the plant floor and robotics wasn't anywhere close to replacing them. And so all of us were coming at this mindset the robotics community, which was the ultimate goal is to chase productivity in quality in the only way. Maybe some thought the primary way was to automate everything and replace people, and what GM taught me was no, that's just not the only way. In fact, is a potentially more powerful way, which is you take these three hundred forty five million around the world who working on production floors and empower them. And I think this is probably the biggest takeaway for me, that you were getting a path to quality and productivity and flexibility, something that we're seeing here, because you put a robodic line down, if you take a six to twelve months to retool an entire line to build a new car. And when you look at things happening in China where you're making electric skillets and you making, you know, electric boilers, you switched, you tell the people on the line we're going to make one on the other and boom, the line is suddenly completely reprogrammed. So I think the message to me was human are the most awesome machines out there and...

...our job is to empower them and to it has proven this for a sixty, seventy years. They have something called autonomation, is what they call it. The idea is that you've machine in man work together to do stuff that together that is much harder for each one to do. So so I think that was probably the single biggest takeaway from me was everyone from in order Da Vinci has known that humans are capable. That just came into my head and that's been the story of my life ever since. So when we started with cobarts, we really started empowering them in the physical way, just making physical jobs easier to do. And then I took a bit of a detour. I went off to build a software company looking at business social networks. In essentially, how do you take data, put in a graph and make it searchable so I can discover a path to Joe Sullivan? I want to be on Joe's talk show, and who can take me to Joe that Joe will take the meeting with me? And today Linkedin is the prime example of that. But really at spoke we created that Reed Hoffman, so that six months after us and beats at our own game. But you know, well, that's a sad story. Transformed that in the white color world. So when it came to Drishti, the question that I was asking myself at one level was how do I take these concepts in the white color world? How do I take them to the into the plant floor? And that was really what, for me, General Modis did. It changed my worldview. Yeah, I love that and I think it's such a smart way of looking at things that it, you know, doesn't have to be so dramatic. It doesn't have to be humans or robots, right, or no technology or all the way in with some technology. You made a great analogy actually, I think, in a in our previous conversation, when you reference the way that you know, a spell check function in Microsoft. Word empowers the human to be more effective, right, like how's that apply in this manufacturing environment? Yeah,...

I mean the let me just continue that example. So you got a spell checker, Grammar Checker, right. You and I write a lot and the beauty of the grammar spell checker is it helps people like you and me produce artifacts that look almost perfect, but nobody in the outside knows that you know that. You know my spelling is terrible or grammar is in the best. The output is what everybody looks at and that's fantastic for us. Word Color Times. But really for me it was the question at risk to you was can I do something similar for the line associate, for the surgeon in the operating theater and for the pharmacist dispensing medications? Right. So the real if I would have frame it up slightly more abstract as I was looking for the gentle purpose spell checker or Edor checker for a range of tasks where people are at work. That was my central motivation to start with, and people think I those manufacturing because of my background. Actually, I chose it because it was the place where I could find the most repeatable work. Right, we're building a completely new technology and if the data was also moving it would be very hard to build a working system. Right. I mean you think about is an operating theater. It's not always the same that the different doctors with different styles and the same doctor will perform different surgeries. Just too much variation in the operating theater, but on the plant floor it's the same product being built and yes, people are moving, but basic fundamental geometry and process is the same. So that's really why we went with this. And but one day I agreement the situation where we can answer out though. The one day is request of all operating theaters follow standard work, because your follow standard work, the outcome in the miracle operating theaters much higher. If we have fewer debts, we have fewer errors. Concept is still the same. So I think the spell checker is a fantastic analogy for what we do here.

It's this notion of assisting people with a running the back, people in the front and together the output you can tell. I think you've summarized the concept and the idea really well here. I don't often have my guests go a whole bunch in into their product or service on the shows. I don't wanted to be a product pitch show or nobody be listening right, but I'm going to give you the green light here. Did you exactly that, because I want you to be able to apply what you've talked about here so far in a tangible setting and tell us how drest these technology kind of brings this whole concept to white. Absolutely absolutely be glad to do that. So okay. So the first thing is we can compress space in time. What you're looking at here is a customer Barus running live in Mexico, but you're sitting in Centatos, I'm sitting here in Paul. All turned we together able to look at a line somewhere else in the world. There's almost like zoom for manufacturing and and if you think about cover times of covid people can travel. Yet you want the you want that data flowing, you want that insight flowing to your people, and so this sort of is a very basic capability that video gives us, because videos very, very powerful. In fact, I make the case that you know, there's the adage that says a pictures worth a thousand words. I say videos worth a million because the information content sitting in video. The second reason is independent of what your background is, you and I can visually tell what's going on. No paper written, no pictures, no interpretation required, no language translations, because videos and universal language. So this is bringing the world to you. Now a second case that we often run into is you have a customer calling in and saying I received a unit and it's defective. And we had the example of Samson blowing up billions of dollars because they couldn't figure out why phones were blowing up. So let's just assume you to your phone with the serial number that you know you're reporting in. What we can do in our system is essentially youtube like bring back the..., but certificate of your phone being built on the production line. So you can see all the stations here in Prasade. Let me, let me stop you for a second. I'm gonna just I want to explain to those who are only listening right now sort of what prasade has up on screen here. We're essentially looking at. You just put in a serial number for, say, a specific phone, right and now what we have here is, if you can picture, you know, what it would look like if you were on Youtube and you've got a big video up in the middle and you've got some small videos on the side. We are looking at an actual overhead camera looking down on a production line with somebody assembling at that station. Correct, if you can elaborate on that if you'd like, Prasade, but we're looking at actual footage back from what? What's the date here? August eighteen, two thousand and twenty. Okay, great, so the entire so everything when this production lines running is is essentially captured on video time stamped. You can go back, even with a serial number, and see what's going on. So feel free to elaborate if you'd like, but I want to make sure that anybody who's listening only understand sort of what we're looking at. Yeah, thank you for that translation. I'll do that as well. So the whole story here is that you can retrieve information of past events to do root cause analyzes, to figure out what your warranty might look like, warranty exposure might look like, to understand I made a design change, to that design change work or did not work, and to be able to pass this information around the entire company so teams can collaborate on visual information. So the whole idea is you, we are streaming this video from every station on the floor where analyzing it using our own proprietary and next generational computer vision date models. We're creating video analytics that are searchable, that are creating a new data set that you can use to design, to tune and to optimize your production system...

...on the floor. That's funny, mate, what we're doing. So if I take you as an example to I mean there's another company here where you can actually look into their lines and you can see this live. Again, you're seeing them produce whatever they're producing on the plant floor, but all of this you're watching as it is running live and you can actually see see people moving around. So you can see multiple stations. So you really have even this almost it's like you're setting the rafters looking down watching this whole process flow. What that gives us also is the ability to think about balancing lines, so you can look across multiple stations and ask yourself, how well is this line balanced? Where are my issues? How do I rebalance these lines so that I'm optimizing through both through my lines? So essentially you're looking at what's called a line balance chart, identifying the bottle next stations on this chart and you now know where to take your industrial engineers and focus their attention, which is essentially means that they're getting the data that they would normally run around the floor trying to get. had half the time to be spending spend trying to get data and now hundred percent of the time is being spent fixing problems because the data is being delivered to them and the last thing I want to show you here is training. It turns out that companies have a huge problem with training. You know, you're getting people coming across from a Walmart is an example, and I want to know what the difference is. How well am I doing against next pert? So what you're seeing here is video ab examples of the expert on the left and me on the right, and I can visually start saying, oh, this guy seems to have a knack of doing this. Well, what's he doing? I can zoom in, I can study the expert doing it. I can learn from that myself and get much better. So essentially this raw videos being trained into finding examples of videos for me to look at, for me to analyze and for me to play with. Right. And then I'll close off with one little example here, which is the spell checker analogy that I gave you earlier. So if I look...

...across this customer, it turns out they made fifty five errors over sixteenzero cycles, which is incredibly good. And if I look at what these guys are doing, they've got these two errors they made on me the twelve, and I can actually look at this and say what went wrong so you can actually see our a believe, to deconstruct the visual image on the right hand side, the steps that the operators going through. Right. This is a core of our technology. We call this action recognition. And down below here you can see why we threw a flag. It turns out this person was supposed to only tighten the fuel by bottom and the top, one s each in that sequence. But you can see our system identified it that bottom, top, top, bottom. He's not following standardized procedures. And you don't, you can introduce errors, you could over talk that particular pipe and who knows what happens in the field. And there's absolutely no human being in all of this. This is automatically detected by the eye system for the operator, for the operator, for the company, everybody to work off of. Right. So this is literally there's a little ipad hanging there that tells them, Hey, you went through the sequence strong. It sends a trigger downstream to the QUA team, so you can do a double check on this this this engine before it leaves to the customer. That's really amazing. I imagine it becomes really powerful when you start to see patterns in the errors to and you can say wow. By correcting this one common error, we can scale our our quality control to write. Oh, absolutely, and I think that is that is the power of what we've built because essentially, by looking across an entire system, we're solving the problem at a system level. Most every other computer vision system out there in the world looks at single stations and we all know in manufacting that the symptom may show up in one place but the problem really is elsewhere. So by thinking of videos, video cameras is core infrastructure on the plan and thinking of video data as the next generation. You know,...

Google figured out what to do with text, apple and city figured out how to mind audio. I think what brish Ti has done is figured out how to mind video and that is the core, central story of the Brish Tis is taking us to the next level of sensing, of figuring out what's scenes and going from there. Yeah, I love that. You know, you told me a quite a few stories when we first talked about this. One that stood out to me was you talked about a production line worker who was fitting hoses to nozzles at a way faster rate than anybody else and I think the company looked at and settle how is this particular guy so darned productive? And while they pull up a video and and the sort, the answer was pretty clear, right. You want to tell that story? Yeah, I do, I do. The story is fairly simple. It is the story of data pointing out a particular rotation, a set of rotations, where that station was meeting cycle time, in fact beating cycle time quite dramatically, and every other rotation was having a trouble trying to make cycle time. And so an investigation discovered that the same a particular operate on the floor was on those rotations that always made cycle time. And so, digging in further into the video, they realize that this this gentleman would show up with a cup of water, and turns out to be soap water, and so what he was doing was he's dipping the hose in the soap water and then, you know, getting it over the nozzle. And it was a breeze for this guy and there was excitement because now you suddenly realize there's creativity on the plant floor. Right. In fact, we call him the brilliant outliner and most most of us think of people on the floor is fundamentally interchangeable. That's the thing that Henry Ford tried to drill into our heads by making the parts completely replaceable, interchangeable, then people could also become interchangeable. It turns are humans are very different and we're not quite interchangeable. And you got these people who are...

...very bright sort of think differently, and so what happened was they realize here's the right way to do it. They made that a standard process across the line and across all of the company and it actually got the HR team thinking differently. It said, how do we differentiate these kinds of people on the floor? How do we give them promotion paths? How do we put them on the hardest problems that we run into, because this creative character was all your problems for you. So you essentially started getting a rethink of people on the floor from multiple perspectives, and that was the exciting part, that what appears to be oversight of the human being is actually a nick is an opportunity to highlight capabilities, to grow people and to change how we build stuff on the floor. Yeah, I love that. Let's stay on this people topic for a moment here. You know, we are living in a time where employee turnover is at an all time high. Finding frontline and Labor has become pretty much the number one challenge for almost every manufacturer I talked to. Can you talk about how technology like drift sees can help manufacturing organizations? You got into this a little bit with the example of the track of training, right, but talk about how this can empower manufacturing organizations to help their people be as successful as possible and to, you know, help them sort of emulate the true pros who are maybe are working on those same lines. Yeah, it's funny. There isn't a single phone call with the prospect or customer where they tell me, you know, that annual labored to turn over as between ten and thirty percent. And I remember talking to a few Chinese prospects and they're telling me fifty percent one week in golden week, weekend, have the company does not show up on the following week, Monday, and then they also tell me there's almost always a ten percent absenteas on a daily basis.

So that's a pretty tough, you know, variable variation to deal with, because you want to build product that is absolutely invariable, right. So that's the heart problem that they're all dealing with. So it turns out training is the most fundamental element that helps them deliver. And like I was giving you the example of golf, I meant Tiger Woods, pre sade, no, no comparison. Supposed to look down to hit the ball, but of course I'm drifting off and looking into the distant Mons and a video recording highlights what I'm doing wrong. Even though the instructors told me look down, I'm instinctively looking forward and when I see myself doing that two or three times, I can and I realize, okay, this is what I really do. My ability to hit the ball in the right direction, the right range fundamentally changes. I've tried this out myself and I can vouch for that right. So I think it is this core idea of taking somebody who walk across the street from Walmart because he's being paid one dollar more for our not quite the most dexterous person, not a natural mechanic, and having to first start by teaching them the fundamentals of manufacturing, the Fundamentals of assembly, getting them to understand the importance of following sort of the right sequence and getting quality right, and then putting them on the floor. So this two parts of the training problem. One is before you get on the floor and the second is on the job feedback. Right there's like on the job as I'm typing. Is One thing to learn how to write in school, the grammar, and other than nothing to write an article and have a spell checker running behind you. And I think that is really the huge opportunities sitting in front of us, which is a priori training and on the job training and the use of video and ai to enable that. And and the results are very stunning. Manufacturers typically take six weeks or so to train and new employee. So look at the cost of that. You're losing a third of your workforce, hitting six weeks...

...out of fifty four weeks, ten percent of roughly the time to train them, and then you're going to lose them again. So what our customers have found is they can cut that six week in half, so in three weeks you can have somebody up and running because the video is a faster feedback loop. But what's more interesting is even as you cut the time in half, you can reduce the defect rate in half. So you know, trading off the two. Oftentimes, you know, if you shrink the training side time you the defects on the floor are higher. Here you're getting shorter training time and reduce defect rates, which is, I think, really fascinating from a business perspective and and I think that is why I get so excited about all of this. You can just start doing the math and what I just describe for you. That's huge payback. And I imagine there are things that happen downstream to on the on the employees side or the workers side, where you walk into a job you understand what you're supposed to do, you can learn faster, you gain confidence, you get good feedback, you feel more secure in your job, you feel better about coming to work and all these things are huge. And you know when when you look at the challenges companies are having with retention and getting people to show up. So some of that stuff may not show up in a spreadsheet, at least immediately, but probably have a really long term, positive long term effect on employee wellness, happiness retention. Right, yeah, totally. I mean if you take this example, I gave you the hose in the and the soap water. HMM. Now, escentially, what's happening is that you're getting the front line empowered to actually Redo what the process is. It's not an industrial engineer sitting somewhere in headquarters of telling you what to do. Is You educating them and saying this is the better way to do it. And you know, one of the things nice things about this data is that you can actually conduct experiments. So go to the core concept of of continuous improvement. In lean typically run a cousin event once once a year, maybe once a quarter. But what we really...

...saying with this digital measurement technique is it you can run it every shift right so you're compressing time again, which means you can an exponential benefit from that any change you make. You can test it, you can validate it and you can either reject it or accept it and start seeing the benefits right away, which is completely different than trying to automate putting a robot on. It takes you multiple months to get the robot take multiple weeks to get it up and going. Here you can conduct experiments all the time because of human ingenuity and and are ably to measure, which I think just returns, you know, gives the company returns almost right away. So we're we're shifting the whole thing that's why I when I started, I said the light bulb went off me at GM that if I can get the three hundred forty five million people and power to do stuff, we will see very different productivity levels. In fact, it makes it possible for America, as an example, to start sort of reshoring work. You take the skills of the people that could believe to cammunicate, the will to interpret data, and now you suddenly start off setting, you know, the behavior. So in fact, one of our customers are telling as a three percent improvement in quality more than pays off the car Labo coast of friendship between the US and anyone else in the world. So if you can give if I can give them, a three percent left in quality, and I just totally would give as much as fifty percent. It's a whole different ball game from supply chain network design. You know, do you do? You want to be in China it can you do in America? Can you all of that equation that everybody's looking at in the face of covid suddenly changes. Yeah, that's really powerful. Prasada, I heard you reference in our last conversation the money ball story that many of us know about how the two thousand and two Oakland Age team, I believe it was, was constructed. How's that apply here? Yeah, I mean if I run with that example that you gave, they had the a young Harvard NBA show up and the a's were up broke. They wanted to build a good... and this guy went around the different scouting events and he realized that the scouts were measuring people with the wrong yard stick. So, as an example, there's pen measurements. Every player had to go through at how fast you can run, how fast you can hit, how fast you can throw a ball. But how fast you can throw a ball doesn't matter if you're if you're eiven to be a battery trying to hit the ball or of the ball park, how fast you can throw the balls. That element. What matters is the picture knows. Should know how to fast the ball is and how much to move the ball bad it doesn't need to know that. So the revelation he had was if I can take specialists at a lower price, in his case, and constructor team, I can get a fantastic output. So that's the concept that we are taken back into. Brish tea is saying, if I know that Joe Specializes. Naturally is inclined to the task and stations one hundred and five and eight, and Prasade is good at two, four and seven and others. Brandon is good at whatever else. Now the system can automatically allocate the right people to the right task. So you're the batter, I'm the I'm the Pitcher, and he's the he's the casher. Right, and so suddenly you're constructing a team based on what's available to you and you're not asking any member of the team to do something that you know it's hard for them to do. In fact, you're doing the opposity making giving them as assignments that are easy for them, that they're good at, and the natural output of this is higher productivity, higher quality, just like the a's would be able to do very well. And so it is this concept, this is what's truly exciting about what we're doing here, is that we are essentially using math, we using data to have preconstruct teams in the face of absoluteism, in the face of Labor Churn, so that the end product comes out absolutely perfect. There's a lot of math. I mean, if you think about logistics, where you apply a lot of LINEAR ALGEBRA and and...

...constrained organization techniques and such like. All of that math is waiting to be applied here on this new data set that rift is created for sad this is really been great. Is there anything you'd like to add to this conversation that I did not ask you about? I think we're sitting on the frontier of something new. I believe that in the not too distant future cameras will be a ubiquitous and every plant floor everywhere and the world will start seeing, and I play a play of words and on seeing from all of this data and we will transform manufacturing in a very fundamental way. You Know Henry Ford and and Credit Taylor. God is going one way. I think this is the next bit leak and I can't wait to see all of that. Transmission happened the floor and I invite your listeners who curious want to talk to us to call us up. It's Rish tecom Drishticom, or drop me mail and love to chat with the well beautiful that you kind of got to my name. My last question pre sade, which is what's the best way to get in touch with you? And you just mentioned drish tea and make. You can maybe requested demo. There is a correct for anybody entrusted. Absolutely, yeah, you can request a demo, you can write, put a send an email to info at rishtecom and somebody'll get back to you and and we'd love to engage beautiful well, thanks for doing this is super interesting. I love what you guys are doing. I think it's you. I completely agree that you are at the front of something huge that is probably going to be common practice at some point, and so thanks for what you're doing and in the manufacturing sector, my pleasure. I said, I just happened to follow in the footsteps of something interesting here and and we's kind of great to see what the future holds. Well, thanks again, prasade. Thank you doing for... me this opportunity and as for the rest of you, I hope to catch you on the next episode of the Manufacturing Executive. You've been listening to the manufacturing executive podcast. To ensure that you never miss an episode, subscribe to the show in your favorite podcast player. If you'd like to learn more about industrial marketing and sales strategy, you'll find an ever expanding collection of articles, videos, guides and tools specifically for bedb manufacturers at Gorilla, seventy sixcom learn thank you so much for listening. UNTIL NEXT TIME.

In-Stream Audio Search


Search across all episodes within this podcast

Episodes (129)