The last video, we're talking about firm practices such as organization structures and how skills should be distributed. In this video, I'm going to talk about other managerial practice and employee management practices associated with AI, okay. And this is really fundamentally motivated by the fact that growth and AI, related investment has been extremely rapid in the last couple years. And how does that change the employee compositions? Okay, specifically how should managers react to this growing investment in A I. And what kinds of firm practices are needed to accommodate the change not just for innovation, but generally this rapid automation on various type of work. And you probably heard a lot of these type of articles in popular press about how robots and AI are potentially causing a huge turbulence in both employment and managerial practices. You probably heard a lot of, red a lot of articles about jobs for robots. The fact that robotics company now than AI, robots will destroy all our jobs and we're still not ready for it. And this is published in foreign affairs, Harvard Business Reviews and even policymakers in European Union has proposed some kind of robot tax. In the 2020 presidential campaign, the Blasio famously proposed a robot tax to Buddhists campaign before he flamed out. And Bill Gates also proposed that maybe robots should be taxed as human. Okay, these are sounds interesting, but fundamentally we actually do not know to what extent is AI and robots really affect employment and to what extent do managerial practices need to accommodate these changes. Before we start with any kind of policy implications, we should actually have hard facts on how does AI and robots change nature of jobs? And we actually did a nation study an entire country of Canada. We actually captured all the robot adoption in specific firms in a very fine detail level, okay. And this again, this study is using tax taxation data and also census data. That means almost all firm needs to reply for compliance. So we have, we got like 89% all firms in Canada responding to the survey. I can see that robot adoption is rapidly growing in services compared manufacturing. So this is a little bit different than when we thought before, that robot is really affecting manufacturing. But here we actually see robots adoption speed growth is actually in services, especially in recent years. Of course, you see the one of the biggest adopter for robots are manufacturing, okay. And you see that in other industry machinery, heavy machinery industries, okay. And you also see the growth being quite rapid and healthcare in certain scientific research and all the other services. So again, this is not just a manufacturing story, but it's really a global economy story including both manufacturing and services. So we've seen that robot re-adoptions have increased dramatically over time, okay. And AI may mirror that adoption pattern and growth may even be faster as a new generation AI becoming more used to mainstream. And we in our study we actually seen that productivity associated with robot adoption is quite high is actually 10x more than factor share. So what does it mean? Is that if you adopt robots and a perfect equilibrium, you should get exactly the value of your robots quite similar to how much invest in it. But here I'm telling you is that the productivity associated with robot options 10x its value, 10x factor share. So that means there's a huge productivity improvement associated with robot, okay. And the dispersion of that is quite high. Some firm do really well, robots on firm don't. That's just complementary organization changes are happening associated with robot adoption. To give you and an idea what exactly is going on with robot unemployment. We break it down into both employment and also with managers. So contrary to the popular narrative that robots killing all jobs. You see the left hand side. This is the event study basically you can read the zero being the year your adopt robot and native number or the years before you adopt robot. So number two means two years before you adopt robot being a firm adopting robot and a positive value to means that two years after you adopt robot and the y axis is what is effective robot and a change in employment. You can see that before firm adopted robots. This virtually just hovering around zero. Robots shouldn't have an effect on unemployment when they don't investigating and on zero you see an increase in employment and gradually increase after two or three years. Even a plateau after 5, 6 years is still higher than initial level. So it's actually the opposite of popular narrative that robots actually increased employment overall, okay. But this effect is not uniform, okay. If you look at the right hand graph, you almost the exact opposite pattern. But this graph instead measuring total employment, you're measuring how many managers are there before and after you adopted robots. So before you got robots against hovering around zero, okay. And the year you adopted robots, managerial employment has gone down quite a bit. And plateau a little bit for a year two. And another big decrease after year three, four. It's different from what we think. So why would robots increased employment but decreased managers? It seems almost counterintuitive because managers by definition and manage other people. How could we replace them with robots? Are robots really are managers? What's going on here? And this is not a statistical fluke. If you look at managerial employees, okay. Between the two types of firms, the robot adopting firms, the blue lines, the solid blue lines and the non-robot adopting firm, the dash red lines. And we plot a share of manager employees over time you can see the robot adopting firm are gradually reducing the share of managers and employees. Where is the non-adopting firms for robots have pretty much constant share managers. This is really something about robot adopting firms that are changing, okay. But not the non-robot acting firms. So why managers? Okay, well, maybe because manager really expensive. So maybe if you want a cost cut you want to get rid of him first? Interesting, we can actually measure that. In part of our survey, we actually ask this question which of the following factors with respect to the relative important to your workplace, general business strategies. And one of the questions one asks, is it reducing labor costs a really important strategic priorities or is improving products, services and a few others too. A notice when we linked reducing labor costs as a strategic priority to robot production, the fact size zero and almost no relationship between the two. Okay, but if your firm strategic priorities was to improve product services that's highly correlated with the propensity to adopt robot. Okay, so it's really not about cost cutting here, but it's really about how do firms improve product services that ultimately drives the firms adopt robot. So it's not about managers being really expensive. So why does AI and robots have profound effect of managerial employees and also manage your practices? Their true reason for that. Okay, number one, remember the graph I showed you earlier where decrease dramatically first and then a plateau and decreased again two years later. If you think about automation technology has enabled by robots and ai a lot of reporting tools like who showed up on time, whether they done the thing they should be doing clock in clock out. And this type of work are can be relatively managed by technologies already. So it kind of leaves a lot of managers with other type of work because the supervising role has decreased quite a bit. So there's a probably immediate effect on technology effect basically, technology replace some of managerial functions. Okay, but still they're managing people. Technology can capture some of the monitoring, technologies can capture some of the aspect of them is supervising roles but it cannot be capturing already if not most of it. Okay, remember two years later there's another dip. Okay, and that is likely to come from the fact that employment composition has fundamentally changed after a robot that's been adopted. So what do I mean by that? So I actually disaggregated in total employment by the skill type they have. Are they highly educated? Are the middle skilled or are they low skilled? High skill being you have university degree for your graduate degree and middle skill being you have some- you graduate high school and you have some kind of accreditation. Okay, two years social degree. Some kind of trade skill accreditation. A low skill being people who have just high school degree or less. Okay, so if you look at the robots effect on the high middle low skill workers, look at how different they look. Okay, the first bar, just total employment, I showed you that already. So each bar here is we're measuring the effect of robots on the total employment on high school employees and middle school employees and low school employees. So each bar is effect. Okay, so notice that high school has gone up, the demand for high skilled workers are has gone up quite a bit after a robot compared to low skill workers is much smaller. Low skill has increased quite a bit too. So both high school increase was low being much more than high skills. But notice the middle skill workers on that bar, it's a big negative number. So although robot has increased unemployment for all workers, but the fact is varying his heterogeneous. You see a huge effect, positive effect on low skilled workers, some effects positive effect on high skilled workers, but very negative effect on the middle skilled workers. Okay, that's problematic because that changes the type of people that managers are needing to manage now, do you supervise now. When you have such a large change employment composition, okay? The type of managers you need will also need to evolve. So what does it mean to manage a very high and very low skill workers are going to be very different from managing middle skilled workers. Okay, low skilled workers are working very much standardized. You can like think the pick/pack at an Amazon warehouse, you have camera looking at, they're showing up on time or not. You can have each scan code on the products. You know exactly how many boxes were products of process. So all of that can be traced and monitored. Okay so you probably don't need many supervisors to teach low skill worker how to do their jobs, little training needed and much of its standardized and can be easily monitored using technologies. Okay, so compare middle skill work. Okay, you probably need a lot less supervisors. Okay, but managing high school work is also very different because these people are probably expert themselves. They probably know how to do their jobs better than their supervisors and managers. So their managers are probably more acting like a advisor, a coach and helping them to do their job, supposed to telling them exactly how to do it. So again, the type of manager managerial skills needed to manage high skill workers can be quite different from middle skilled, low skill workers. So again, because of this employment composition change, you see a lot higher low skill workers, a lot lower middle skill work and moderately high high skill workers. You will need a different type of management. So as a result, you see that manager on average has gone down and that is because most of the employment increases coming from low skill workers. If most employment has gone to high school worker, you may have a different result. You may not have seen a drastic decrease in manager that's before. But either way the type of people we manage can be different, therefore type of managers you need It are different number is going down. But more importantly, firm practices are changing to accommodate or to co-evolve with robot adoption. So specifically, we looked at three different firm practices. First look at span of control, how many people are being managed by a single manager on average? So if our results are correct in the sense that there are many more employees that a lot of fewer managers by average span of control should also be going up right, a lot more people under single manager. We independently measure that. We actually asked manager how people are managing? Okay, and we have independently confirmed that span of control has indeed have gone up. So managers have to manage a lot more people than they had before. Okay, we also look at work predictability. To what extent can you actually predict your hours, what you need to do? And if you think about that robots are doing all the predictable part because you remember AI machine learning a great prediction machine mining patterns that we haven't seen before and by using that pattern to predict what's going to happen next. If all the predictable elements are going to be done by robots may be leaving workers with all the unpredictable parts and you have indeed seen that work. And predictability has gone up quite a bit, have a robot. So this also just the nature work is changing. And lastly, if you think about robots can decrease the various production, remember robots using algorithm can do the same thing over and over again without being tired, without taking a bathroom break, without taking a lunch break, they can do it 24/7 a dark room without seeing you. Okay, and that reduction variance could also mean that The fact that what you produced can be more attributed it to your effort, your abilities as opposed to some random noise, right? So now the firm may be able to more accurately capture what exactly is your contribution of production process, okay? By more easily attributing to your performance to the actual effort what you actually did. Individual pay performance could also go up and we have in deed shown that performance pay have gone up after robot has been adopted. But again, these are evolving we're just scratching the surface here. But we overall we do see that robot is dramatically changing from practices, especially in how people are being managed. And lastly, I want you to show that this dramatic change is not just about people. But also decision rights allocations, who are making decisions for these tasks in the firm's? You see that this is robot adoption is also changing who deciding what? So in the first one we asked who deciding who gets trained for what? Before and I've heard robot adoption, we this is what we see. We see that middle bars management, managers are making a lot less decisions on what you should be trained for. Employees should be trained for where employees are doing most of the decision for themselves. He's in the left most bar nine managers. So decision rights are shifting from managers to non managers on the decision on what type of thing they need to learn, what kind of thing they need to train for, okay? For CEO upper level management focus pretty much zero. So it's really about decentralizing that decision rights for managers to non managers, okay? And we also see decision rights for the choice of production technology, who is making the decision on what production technology to use. Here again, we see manager losing that decision rights, okay? We see that manager no longer making the decisions as much as that decision is moving towards the CEO or the C suite and the C suite managers. We see that from nine managers there's really no effect there for managers. Middle middle level managers a big decrease for owners CEO level that decision rights has gone up. You can see not only the managers are decreasing in their pure number of head counts. But what they do have a fundamentally changed and the decision rights are also being centralized, decentralized away, okay? This is a very subtle story than before because when we think about it robot and AI are not supposed to changed managers. Because they manage other people, people manage other people, right? We haven't the fact that robots can performing managerial task is something kind of far fetched. But here when I'm showing is not necessarily because robot can do so much what managers can do. But fundamentally because the type of management pack of employees have changed that ultimately meaning that the type of manager need are different. So when your adopt robots and AI is not just about plugging existing, take it take out of this system plug AI in. It's not just that you actually have to examine the entire process because it would be ripple effects. Everything you do on decision rights, employment hiring, turnover type of managers you need, various from practice and compensation etc, okay? So say important and profound decision and you do that you need to figure out what impact are these technologies are really changing. And how are you supposed to accommodate that change? So I'm giving you some idea where it could be, okay? But it's really important for managers themselves examine specifically in their context, how these processes are changing. And you may see some resistant in managers if they believe that these kind of technology could replace managers, okay? And that could create situation where firms should adopt robots and don't, okay? And that's actually precisely what had sort of happened in some two exams in our data. We actually see that the robot adopting firms are far more productive hire, a lot more people than non robot adopting firms. And whatever loss you have seen in the popular press or an earlier study on country industry level studies that show a employment loss. It actually comes from the non robot adopting firms losing market shares, losing competitiveness because they didn't adopt robot for whatever reasons, right? Could be resistance from workers or managers could be many other reasons. But again robot adoption AI is not a small decision, it has ripple effects to an organization. So it's conclusion that nature of manager work is changing and practice are evolving. So we have shown that robot adopting firms increased productively and hired more people. But only for the firm adopted, the firm that didn't adopt in the same industry actually experience a greater employment loss, okay? And we see that this is really shows that managers in the sense that robots associated with less managers hiring greater turnovers. And it's not cost driven, its quality driven, okay? As a result, the type of people after robot adoption are very different from the type of people you manage beforehand. That means you're going to see a lot of organization changes associated with management. And it's important to continue and monitor how there's change involved and customize it for your particular firms.
