 Now I'm excited about this next one. Please help me welcome Professor Panish Puranam from INSEAD. In a distributed virtual or asynchronous environment, learning takes on new importance and new challenges. Any parent, student, or instructor who was involved with the sudden shift to remote learning in 2020 certainly saw firsthand what that really means. But what about organizational learning? What lessons can we take from the past year and how can we use this experience to create better learning environments? Panish is a leading expert in organizational science with new research to share. This talk will explore how to excel at distributed learning by not just addressing its challenges, but also seizing its opportunities. Hi, my name is Panish Puranam, and I'm here to talk to you about organizational learning in the age of distributed work. I'm a professor at INSEAD and I hold the Roland Burger Chair in Strategy and Organization Design at INSEAD. I've been working there for about a decade now before which I was also at London Business School, also for about a decade after getting my PhD at the Wharton School. I've been actively involved in doctoral program supervision both at London Business School and at INSEAD. My broad areas of interest are around organizational design, and as you might imagine, the great organizational design experiment that all of us have been through since the beginning of the pandemic last year has been an opportunity both to cope with personal disruption and difficulty, but also as a source of interest to learn what we can about this extraordinary experiment and remote working that has been affecting the organization designs of companies all over the world. And that's what my collaborators and I have been working on for the last 18 months or so. The kind of work I do is not possible without collaborators. And I think it's a good point to stop and acknowledge some of them, all of them in fact, who are associated with the topic of distributed work. And you'll notice two things from this panel of collaborators that I'm showing you here. They're themselves highly distributed. So a few of them are here on the island of Singapore with me, but many of them are not. Some are in the US and many are in Europe. So we kind of live our own work in the sense we're all working on distributed work of various kinds, but we're doing it in a distributed way. And this team has been working on software development using distributed models, much like our friends at GitLab do, open source communities as well, and innovation contests. So these are all different forms of collaboration where people are not co-located. And that's been the theme of the lot of the work that we have been doing as a team for several years now. In fact, our engagement with GitLab precedes the COVID shock. So some of us on this page were actually working on developing a case to understand how an all remote organization like GitLab might work. And we started work on that a good six months before the pandemic hit. So in some sense we were missing it. Now, when people talk about remote collaboration, and very often I think these are the two extremes that people might have in mind. So on one end is the case of all remote where everybody is off working in their own little isolated unit. And the contrast is that of everybody being co-located and in one space. And maybe the picture on the left is a bit of an exaggeration because even before the pandemic of course what was much more common was not everybody in one location but clusters of people in different locations. So that for instance would have been the typical organizational design if you like of a multinational company or even a national company which was regionally distributed in different parts of a country. What I'm convinced about having observed everything I have over the last year and a half is that these extreme cases where everybody's all remote or everybody's in one location, they are not going to be the model that most of us will employ. So most of us are going to live somewhere in the middle and that could be something like the picture here where we have clusters of people or it could be something a bit more like teleworking or it could even be something that looks like the picture on the right here but is changing over time. By which I mean different people come into work on different days or we even change the sets of people who see each other face to face by rotating through who comes into work face to face and who does not. So there's an entire space of possibilities between these two extremes of all located in one place or all remote. And I think for most of us the action is going to be about finding and tuning the distributed work model that is just right for you. That's really the core point that I want to make in this talk that we want to think about the organizational design challenge going ahead not as how do we transition everybody to this model or how do we go back as quickly as possible to that one but how do we find the right spot for you in the spectrum of distributed working? So it's really about understanding the distribution in statistical terms of distributed working. So that's why I call this the distribution of distribution and finding out where in that distribution you want to rely. So locating that point and tuning yourself to be on that point using data is really going to be the core point that I want to make. Now some of my arguments are going to draw on the research I've done but some will also come from the research others have done as well as my observations as an educator. So INSEAD in that sense is a very unusual organization because we were very distributed much before the pandemic ever hit. So as you might know it's one of the few global business schools in the world which really has global footprint. So we have offices in Singapore, in France in Abu Dhabi and in San Francisco now. So we've been working remotely for a long time and when I was running the PhD program people would ask me from our peer schools how can you run a doctoral program in a remote collaboration model? And the answer is well we figured out how to do it because our doctoral students were absolutely world class and were being placed in the top universities around the world as professors. So we know a thing or two about working in a distributed fashion even at the institutional level but the big push going forward as I said is to help develop tools and ideas that can help any organization figure out exactly where in this distribution of distribution is the right spot for them. When the pandemic hit last year one of the first things that we decided to do in our research team was to dig a bit deeper into exactly how the first responses were shaping up around the world to this pandemic. So in a survey that I conducted along with my collaborator Marco Minermini who's here at INSEAD as a postdoc we ended up surveying close to a thousand people from around the world. Mostly INSEAD alumni and people who look like INSEAD alumni by which I mean people working in consulting or in banking or in senior managerial positions mostly knowledge work but from all around the world. And we discovered a few very interesting things and this was very early in the pandemic somewhere around March or April last year and we learned that here's a pretty systematic bunch of correlates of self-reported productivity. Okay, perhaps the high level message to emphasize here is that the hierarchical level of our respondent and the previous experience of working remotely were much more important in predicting whether they would be productive or not in distributed working models. Then the industry they were in the company size that they were working in or even the nature of work. And that was quite surprising for many people because most people thought there must be huge industry differences. So some industries are gonna get hit very hard and some not so much. The reality actually was that once you get past a couple of layers in the managerial hierarchy, right? It really doesn't matter whether you're in healthcare transportation or consulting or research. Once you are a few levels above the direct contact point with either the customer or with a supply chain or with some other external entity. Then we're talking about managers. Management is knowledge work and knowledge work looks roughly the same across different industries. So we found for instance that being higher in the hierarchy was correlated with more self-reported productivity. Having previous work experience in working remotely was correlated with higher productivity. Better work-life balance which was accomplished because of working remotely was correlated with higher self-reported productivity. And of course the big one avoiding commuting that was correlated with higher productivity as well. So what acted as a drag on productivity in this distributed world that was forced on us after the pandemic? First missing social interaction. So this was extremely important and very repeatedly emphasized to us in our survey that one of the big downsides of this new way of working was the missing out on the social interaction that people took for granted. And the other which actually is a very unequal affliction in the sense it doesn't affect everybody the same is the extent to which you are beset by distractions at home. And here there's a worrying factor to keep in mind that distributed working might actually accentuate some of the pre-existing inequality. So people who don't have a large home or who don't have enough space or who don't have the capacity to hire help to take care of their children for instance have to take primary responsibility in these roles. And in many cases this falls unfairly heavily on women rather than on men. Then distributed working can actually be a force that accentuates inequality rather than democratizes the workplace. And as I already mentioned, the surprising thing is neither the industry nor the company size or the nature of work were able to predict variations in self-reported productivity. So just to sum up, where we are today is the following. We now understand I think with a fair degree of confidence what the trade-offs are to distributed working. The big positive of course that we all learned about and have found quite heady and addictive almost is flexibility. We want to discover that we can be just as productive as before or sometimes even more but with significantly more flexibility in how we allocate our time when we do what we are supposed to do and exactly how we balance our work and our life and that's been great for most people. The idea that by being distributed one can expand enormously the talent market that one can access is of course not new. This was well known even in the world of offshore software development two decades ago or in offshoring and outsourcing 15 years ago but the distributed working of the last year and a half has really made that very salient for everybody. Even in my world I'm getting internship requests from students all around the world every year and I used to get them before but I never took them very seriously because I was implicitly worried about how would I supervise them, would they do a good job, would I be able to monitor them but now I can't do any of those things even for interns who are in Singapore. So once that constraint was relaxed of course I'm thinking why can't I use the smartest kids from anywhere in the world who want to spend a month working for me on a research project so that's what I'm doing. The unstated advantage which I think requires thinking a little bit about is sustainability and perhaps you've made the connection in your mind perhaps you've not but there's a TEDx talk that I gave relatively recently that you might want to look up where I make the argument, a very simple argument that if you just look at the impact on the environment that working from home has had it's going to be very difficult to go back after the pandemic to a world where we assume the default is lots of commuting and lots of business traffic that just will not cut it. Our impact on the environment in terms of carbon emissions through commuting and business travel is not trivial just to give you a rough back of the envelope type calculation. If you travel 10 miles each way, five days a week. You're adding about four tons per annum of carbon to the environment and just to put that in perspective that's roughly between a third to a half of total carbon emissions for a person living in Western Europe or the United States. So that's just the commute. Now through in a long distance business journey something like between our campuses in Singapore and France which I do routinely or used to do routinely about three to four times a year and you could end up with the same magnitude of carbon footprint as that commute that you were doing throughout the year. So this is significant. So even if we had no other reasons such as cutting down on real estate costs or improving flexibility for employees or talent market access just improving our footprint on the environment and pushing the sustainability agenda suggests that distributed working is here to stay. There's no real way to get away from that. Now there are downsides and I think we've all begun thinking harder about them in the last few months I should say. Firstly, what's distributed working going to do to learning? Can people learn as well in distributed environments? And this is both learning in classrooms like we do in NCI but also within organizations and at the hybrid junction of the two when we have execs come to our classrooms for instance in our exec at program settings here. What does distributed working do for creativity? It's one thing to execute on fairly standardized organized work in a completely distributed environment but what about when groups have to come up with new solutions and new creative ways of solving a problem? Is that still possible in a distributed work environment? What's going to happen to our organizational culture and our networks? And this is really a very important point to underline which is in some sense the last year and a half we've all been drawing down on the stock of our culture and networks that we built up when we used to work face to face. Think of it like money in the bank. So we've been drawing it down over the last year and a half with very few opportunities for putting anything back in. And that means at some point we might run out. So this is really a problem going forward if you're going to bring people on board who've never met each other and who will continue to work with each other without necessarily meeting each other face to face for long periods of time. So how do we manage that? And there have been great strides made on this by different organizations over the last couple of years but GitLab has been doing this for even longer which is to think about onboarding people and building a culture on a completely online context. And I'll talk a little bit about that as well. There's also some work I want to shout out here which is in management science which appeared a couple of years ago where a collaborator Julian Clemore and I worked on this idea of the formal structure in most organizations really playing a very important role in being a random encounter generator. What I mean by that is that a big role of the organization design is not just determining who sits next to whom and who reports with whom or who reports to whom but also about creating opportunities for spontaneous unanticipated interactions between people in different parts of the organization. So think about meetings, think about the water cooler, think about the lift lobby and of course think about the annual picnic. So many of these events are really designed to create random encounters between people and that is so crucial because very often innovation comes when people who were not meant to be talking to each other end up talking to each other. So let me talk a little bit about what we know about the impact of distributed working in particular on learning and creativity. This has been a very important question for us not just because we know it matters for organizations but it's very important even in the work we do. As academics we are trying to work as a team to learn something about the world and be creative as we're doing it and we've had to rely very extensively on distributed working because we had no choice and so it became partly a self-interest driven process of trying to answer questions about learning and creativity but here's what we've been spending our time on over the last year. We're trying to answer the question does restricting social interactions which is obviously one of the negative consequences of being distributed. Does that suppress creativity and learning and by how much and why? And the intuition is pretty clear. One assumes that being able to interact rapidly in real time and with sufficient bandwidth in terms of getting all the signals not just of the words but also the body language and the facial expression. All of that goes into the process of coming up with creative new ideas and learning from experience that groups routinely engage in whether we are talking about contexts which are within a company or within a classroom. So the question is if you were to impose restrictions on social interaction, the simplest form of restriction just might be that you can only work through Zoom as we are doing now or some format like that. A more complex form of restriction might be that you don't even use cameras. You work mostly asynchronously. So people contribute to a knowledge product at different times from different places in the world. So there's very little direct real time interaction. That's for instance very common in open source software development. That's also a big part of how our friends at GitLab work. So that could be a different way of thinking about restricting social interaction. So these are different ways in which social interaction gets curtailed once we have to work in a distributed fashion. And the naive intuition might be that we try to minimize that restriction as much as possible. In other words, if we could, we'd all use virtual reality and holograms so that we could completely simulate being in each other's physical presence. That's my intuition. And what we wanted to do with this research was to push that and ask, is that really true? Does restriction necessarily suppress creativity and work? So here's what we did. We looked at a contest that you might have heard about called Kaggle. So Kaggle is a data scientist's dream in the sense that what we have in this is a contest platform where companies and individuals can post their data and teams of individuals from all around the world will compete to try and find the patterns in the data using machine learning algorithms. So what we did was we essentially got hold of the data on literally tens of thousands of teams and individuals who participate in these contests and did one simple statistical trick. We compared real teams in Kaggle, which are groups of people who work together as a team to solve a problem with what we call synthetic teams. So these synthetic teams are composed of individuals, each of whom looks like a statistical twin of an individual in the real team. So that's the advantage of Kaggle. There are people contesting in these competitions as teams, but also as individuals. So for every real team, we could find a statistical twin, in other words, somebody who looks statistically like this individual and created a synthetic team. So the synthetic team is actually just a bunch of individuals competing in these contests who have no social interaction with each other and they don't even know of the existence of the other, which is obvious because they didn't compete as a group. They were competing as individuals, but we could run this experiment essentially of comparing these real teams to these synthetic teams. And because Kaggle is such a well-structured problem domain where creativity and learning are fully measurable because you're literally looking at groups of people trying to solve a hard technical problem so we can measure how creative and how successful they've been. We can measure how much they learned. We can measure how effective their learning was. Everything is available data-wise and at enormous scale. So we're talking about really tens of thousands of teams. So this is big data brought to bear to organization design, a theme that I'll come back to. And what we found here in this work with Marco Minervini and Tia Nuhay at NCL, is the following. When you move from a real team to a synthetic team, you get less effort, which means social interaction is probably a big driver of motivation. So people do get motivated through the social interactions they experience with each other, which is hard to replicate in distributed work, not impossible, but harder. But on the other hand, they actually get more variety in their solutions. So that's really surprising, right? Because what it says is restricting social interaction might actually make groups more creative, not less. Even though the resulting groups where you restrict social interaction might not work as hard because they feel less motivated. So that's an interesting trade-off. And of course, the advantage of something like Kaggle is we have so much data. It's very natural in a context where people are competing for high stakes. So it is about the real world, but we don't have control. By which I mean, we cannot really account for the possibility that the people who compete as real teams are just different from the people who compete as individuals. Now we try to make them as similar as possible using the statistical matching methodology, but we can only match on the things we observe. We can't match on the things we can't observe. So there is a limit to how much we can make these two things comparable. And one way to get around that is to instead use a technique which probably many of you are familiar with as A-B testing. But in the world of academia, we call these randomized control trials. So what we did was essentially create an A-B test in which we're able to contrast teams of people who are put into these two different structures and they're engaged in a creative problem-solving task. So this is research with the same team, Tia Nuhay, Marco Minervini, and also ads Professor Jay Narayanan from NUS here in Singapore. And we ran this with both undergrad students as well as senior managers on training programs at NCI. So it's really spanning the range of relatively inexperienced entrance to the labor market to fairly senior, sophisticated people who have been working both remotely and face-to-face in senior roles for a long time. And here the experiment was to compare how groups which were randomly put into these two structures did in terms of creativity and in terms of the problem-solving they engage in. So on the left, we have a structure which one might think of as a Slack-based team. So I'm putting Slack in quotes because it doesn't have to be Slack. It could be Microsoft Teams or it could be GitHub or even any other product like that where essentially everybody can communicate with everybody else and leaves a complete written record of the conversation. So these are essentially like Slack channels. That's why we call these Slack-based teams. And on the right, we have teams which have the same authority structure as the one on the left. There's still a bunch of four people and there's a leader. But the difference is that here the subordinates cannot directly communicate with each other. Instead, there's a series of dyadic exchanges more like an email-based team where the leader is communicating dyadically with every team member. So in some sense, pre-pandemic, most organizations which had not yet adopted techniques like Slack or Microsoft Teams looked like this. And post-pandemic, because of the opportunities created by the adoption of technologies like Slack or like Microsoft Teams, many more teams now look like this. So again, the question is, is there a difference between the two? Because the one on the left has unrestricted social interaction. The one on the right restricts it. It's kind of a Harvard-spoke system where everything has to go through the leader. And that can create silos and bottlenecks as we understand them. But the question really was, would it have an effect on creativity? And we found quite remarkably the same pattern that we had before. So as you move to these email teams or if you compare these Slack teams to the email teams, what you get is less integration of ideas. So the ideas tend to be a lot of unique kind of bundles with very little coherence across them. This is an idea formation task we put them in. But you get a lot more variety. Again, the same surprising result, which is when you put restrictions on social interaction, far from dampening creativity, it actually seems to increase the variation and the variety of ideas that people come up with. Even though of course this is not entirely free of cost. When we looked at Kaggle, what we found is that teams without social interaction worked a little less hard. So they just put in less effort because they were less motivated. And in this context we found that the email teams, which were essentially working in a hub and spoke system without complete transparent communication among themselves, were producing ideas that on aggregate were less integrated. Because in some sense, the entire burden of integration fell on the leader, whereas everybody was doing it on the structures in the lab. But they were producing more varied and diverse ideas. In other words, they were being in some sense more creative. So what's going on? Why do we get this surprising result? The answer lies in a very old idea in social psychology, which needs to be very widely known and appreciated, in my view, given what we can do today in terms of tuning our organization designs to work remotely. Here's the point. The challenge is something called the Curse of Conformity. And you might be familiar with this in the guise of the line experiment or the ash experiment, but it dates back to the 1950s. And what you see here is an experiment where a group of people are looking at a screen and they're actually looking at the screen that we see on the right. And the question being asked is, which line is the same length as the line on the left? Now, looking at it, it's pretty obvious that it's line A, right? But people would systematically end up saying it's B. So how could you get this insane result where people would say, in fact, B is like the line on the left when it's so obvious to us looking at this that it's actually A? There's no visual illusion here, right? So we're not trying to play with your visual system or your optics in any way. What's going on here is a purely social phenomenon. There was one person in this group that you look at here who's the real subject of this study. The rest are confederates of the experimenter. And what these confederates are saying is telling this one real individual who's actually the subject that we think the answer is B. So for that one individual to defy the crowd and go with their own private opinion that in fact it's A is quite hard. So effectively, what the two experiments that I just discussed and described to you are telling us is that in the world of distributed work, we might lose out on social interaction and that might come with some real costs, no doubt. But it might carry the silver lining of releasing some of the pressure of conformity, which might allow our groups to become in some sense more creative because of the social restrictions that we are placing on interaction between them. So does restricting social interaction suppress creativity and learning? Our answer at this point is not always. We can use the restrictions posed by distributed work to tune up the variation in ideas that we would normally get just by having people all located at one setting. And the point about tuning I think is really important because when we work in a distributed way because we're all working electronically, the tuning of the social system is actually a lot more feasible than one could do if everybody is face to face. Imagine how weird it would be to tell people in the same office to not meet each other when they are in the ideation phase, right? This is not news. People knew about this from the literature on brainstorming a long time ago, that in the early ideation phase, it's better to work quietly and silently. But that comes hard for us. We are social creatures. If you put us in a room, the temptation is to talk, the temptation is to chat and that's what makes us human. But it also acts as a source of potential conformity that might restrict the variation in the ideas we come up with when we are in that social group. There is a point in time that we want to switch to that social group which is around execution, which is around converging on one among several creative ideas. But at the ideation stage, actually we can use the restrictions that distributed work offers us to tune up the variation in ideas. That's really the core and I think somewhat surprising result of what we'll be doing. So to wrap it up, if you think back to this point I made about the distribution of distribution that the world is not just going to end up after the pandemic being either here or here, but it's really about finding the point in the middle which is exactly right for us. Then it's finding and tuning the distributed work model that's right for you. I think that's going to be where the action for the future of work and organization design, that's where it's going to lie in the next coming few years and months. To help with that process, I also want to share with you a resource which is a book on organizational analytics. This is a free book which is available for free download from the link I show you here, written by Julian Clemore at Stanford University. And what we do is we try to give a guide to data-driven organization design which allows anybody, even with somebody with zero background in data science to understand what the opportunities are to use data, for instance, from networks, for from surveys, from sources like Glassdoor, which are fantastic sources of data about the culture of an organization because employees like to use from running AB tests. So this is really where the frontiers of data-driven organization design are. And our goal is to bring any reader, regardless of prior background knowledge of statistics of these techniques to the forefront of thinking about how to use these different techniques, particularly to tune their own organizations and their organization designs. Let me conclude by pointing out that we are sitting at a particularly unique point in the history of thinking about organizations. And that unique point comes in some measure because of the pandemic and the force move to distribution. Why do I say that? Historically, the kind of data we had to rely on for organizational design and development was restricted to non-scalable data. By which I mean, you could run some surveys, but it's very hard to get high response rates. You could interview people, but that's even harder to scale. But now, because over the last year, we've all been working remotely in various degrees, it's meant we've been working electronically. And that's left an enormous digital exhaust, if you like, of data on networks, of text, of interaction, of video, which using the right guidelines, both for data privacy and protection, with the right kind of filters, the right kind of aggregation, the right kind of anonymization, is really opening up a goldmine of opportunities for how to do data-driven organization design. That's where I think the future of work is headed, and that's where I think working remotely is nudging us towards moving going forward. Thanks very much. I hope you enjoyed listening. There are a few other resources which are available through our slides here, which I welcome you to take a look at. In particular, at NCAD, I want to point out the work of my colleague, Professor Ithaista, who's developed a fantastic teaching experience based on virtual reality. Again, this was before the pandemic hit us, but as you can imagine, this has become of great interest now, where he's developed entire courses, entire cases that people can take in virtual settings from anywhere in the world. And here are a few other resources that you might find useful that you can click through. And I hope you enjoy the rest of the conference.