Podcast
Podcast
- 30 Mar 2022
- Managing the Future of Work
Working with software robots
Bill Kerr: Robotic process automation, sometimes called “software robotics,” automates repetitive and routine digital processes using more or less the same steps as a human user. This could be inventory management, resume matching, or even opening an email. While automation has been tied to increasing inequality, RPA isn’t intrinsically job killing. It can, in fact, position people to take on higher-value tasks. As the Covid-19 pandemic and tight labor market accelerate automation, how are companies using RPA? And what does its evolution mean for the future of work?
Welcome to the Managing the Future of Work podcast from Harvard Business School. I’m your host, Bill Kerr. Daniel Dines joins me today. He’s co-founder and CEO of RPA firm UiPath, which counts Google, NASA, Autodesk, and DHL among its customers. Daniel started the firm in his native Romania in 2005. We’ll talk about the workforce implications of RPA, how end users can program automation routines and sequences, and how UiPath maintains an extensive user and developer community. We’ll also discuss how Covid-19 has affected customer demand, how UiPath’s 3,000-strong global organization has handled the pandemic, and what’s next for RPA technology. Welcome to the podcast, Daniel.
Daniel Dines: Thank you so much for having me, Bill.
Kerr: Daniel, tell us a little bit about yourself, your background, and how you came to start UiPath.
Dines: I was born in Romania. My first passion was more around liberal arts. I fancied to become a writer. But I was raised in communist Romania, so that was not really a real profession. I went to the university, where I studied math and computer science. It was really a good choice for me. I got an offer from Microsoft pretty early in my career. I spent five years in Microsoft working for their database division. I learned a lot about building enterprise software. But then I decided to go back to Romania and start a company. It was very foolish, in retrospect—that was in 2005—not knowing anything about how to build a start-up. It was right at the beginning of the new venture capital industry reinvesting in start-ups. So it was very early. We made tons of mistakes, but somehow, after 10 years of hardship, we found good product market fits. We built a technology that we call today “Computer Vision.” This is the type of technology that powers our software. So, literally, we can look at the computer screen, and we can operate applications on the computer screen exactly like humans operate that.
Kerr: That’s great. So you may have said it’s foolish, but it was also highly successful, and congratulations on that. What’s your current geographic balance—Europe, U.S., and outside?
Dines: We were forced to think global to acquire a global base of customers. So we are one of the rare, and maybe the single, company that spreads equally on the three major continents. So even today, our revenue—it’s not equally, but it’s spread across the three major continents. And it helped us quite a bit to build a very diverse and inclusive culture. We got the best people we could have found across the globe. We have big teams in Japan, in China, in Singapore, in Germany, in U.K., in France, and of course, in Romania. And then in U.S., we have a few bases. We have our New York commercial headquarters. We have in Seattle our second-largest development office, where we build our cloud offering. So we have a presence in more than 20 countries. We sell into many, many other countries.
Kerr: Going back to Computer Vision, maybe take us on a slightly more panoramic perspective. On this podcast, we’ve talked about robotic process automation—or RPA—technologies, but can you paint for us a bit of the overall technology landscape and where in RPA you are most actively working.
Dines: When you integrate with another software system, you have different means at your disposal. You can use some API [application programming interface] that the system provides, which is the easiest. Second is to look at the user interface in a browser, for instance, which is the easiest. You can look at the code behind the user interface, and you can figure out what’s on the screen, so you can interact with it. And then there are systems that are virtualized. So what we see on the computer screen is just an image. So one of our initial niches was based on our really good Computer Vision approach of that day, where we could have automated a system that was virtualized. So you couldn’t integrate with it in any other way, but just reading the screen as an image. So that was very powerful. And we worked, especially initially, with BPO [business process outsourcing] companies that typically access their client system using this type of virtualization software, like Citrix, VMware, Remote Desktops, sort of this. But clearly, our niche started with Computer Vision. The secret sauce of our technology is the combination of this API automation, user interface automation, and Computer Vision that are packed in a low-code/no-code automated workflow designer. That makes it extremely easy for to build automation that scales, and very fast.
Kerr: BPO being business process outsourcing groups that you’re working with?
Dines: Yes.
Kerr: And tell us a little bit about that user interface and the low code/no-code type environment. I think we can all understand the APIs—application program interfaces—the ways that you’ve designed to connect. When you go into this more advanced space, how much does the end user of this need to be a programming expert or be able to work with those types of technologies directly?
Dines: This is where we have reduced a lot of the technical skills required to build automation. Using the user interface, in order to automate other applications, is not something necessarily very new. It’s a technique that was used widely in regression testing. And I’m sure you heard about screen scraping of the older days. But what we did was to abstract away the complexity of dealing with the user interface. So in our technology, literally, you have a selection tool. You go on the screen, you see the tool that shows how it sees all the elements that compose an application. You just select an element, and behind the scenes, we capture a lot of information about that element. And when you play the script, we find this element with a high degree of accuracy. We look at the screen. So we capture as much information. We look at the surrounding elements. It’s mostly like a human user would do. If I ask you to get the invoice number from a screen or a document, you’ll find first the label, the invoice number, and then you’ll get the number. You don’t have to remember the fixed position where the number is. And we do the same, because if we remember the fixed position, maybe the next invoice, it’s a bit skewed. So we always find invoice number, and then we search next to it, and get the right number. It was an evolution in the way these tools work. Packing it in a low-code/no-code [tool] also reduces the technical skills to create some business logic within the automation itself. Take input, data input, do some data transformation and enter in the data into other systems.
Kerr: Yeah. So give us some tangible examples of customers and how they’re using this technology.
Dines: We have started mostly in the financial industry. Big banks, big insurance companies, were our initial customers. And we have some of the largest banks in the world, from Japan to Europe to United States, Latin America. What they do is, they work from everything—invoice processing to core banking systems to credit card reconciliations, all sorts of manual tasks. A lot of compliance tasks they are doing. But also we have modern companies—Pandora or Spotify, Uber, CrowdStrike—that are using our technology in their finance, for sales automation, for marketing automation, HR automation, legal automation. Plus, we are very proud of what we’ve done during Covid to help the healthcare systems and the local governments. To give you a few examples, the Cleveland Clinic is one of our major customers, and we work with them really early on for cutting waiting times at drive-thru Covid-19 testing facilities. So, while the process takes a human two, three minutes to execute, with our robots, we execute it in 14 seconds. And then we work with Ecolab that uses our robots to handle the 10X spike in hand sanitizer orders. You remember how difficult it was to get the sanitizers early on.
Kerr: About two years ago, from when we’re recording this podcast.
Dines: Exactly.
Kerr: Hand sanitizers were a very hot commodity.
Dines: And one of the most striking examples comes from a hospital called the Mater Hospital. So we used our technology to help with the administrative burden regarding the Covid tests. So we were capable of saving three hours a day per each nurse. I think this is huge. And they, in turn, were [more] capable of helping their patients. And also from government, like New York State was really in a messy situation early on with all these unemployment claims. So over a single weekend, we were capable of bringing on 200 robots to help getting the relief into the hands of their constituents.
Kerr: Amazing examples. And maybe let me frame it in a slightly different way. When you go and start working with a company for the first time, clearly there’s part of the interface that you’re going to do that will always be customized or bespoke. But broadly speaking, how common are the use cases that one customer has with the next customer with the next customer, and it’s a matter of just fitting into the computer systems, versus each new customer wants to do something a little bit different with the RPA?
Dines: Well, our software is more like integrated development environment. It’s a tool that generates software, which is the automation. When we go to a customer, first of all, we have all the blueprints of how to scale, where to find use cases, how to put them in production, how to maintain them, how to work with partners. And we have a lot of accelerators that can help you get started with typical use cases—like order to cash or procure to pay and many others in healthcare, in banking. But it’s not a ready-made system, for sure. So our customers typically build an internal center of excellence that is in charge of working with business lines to source opportunities and then to implement those opportunities and later on to maintain [them]. Usually this center of excellence is in charge of the internal developers, but also they work sometimes with one or many partners that help them with the entire process.
Kerr: Understood. Let’s go back to some of your examples and talk a little bit about job implications. You gave some that were happening during the pandemic, when the healthcare system was stretched beyond belief, and saving those three hours was very important. But in other settings, maybe you’re needing to reduce the overall size of the financial services company that you’re working with. How do you think about automation and its impact for jobs?
Dines: I think this question is relevant, not only for RPA, but for any modern technology. RPA raises more questions, because it’s a technology that emulates human users very directly. So the impact can be measured. But we are in one of the best labor markets for employees ever. We are facing the phenomenon of aging of the population. Plus the new generation is less prone to work on these low-level mundane tasks that our robots automate. Automation—and to a certain extent, RPA—is one of the only ways to cope with the pressure on the labor market.
Kerr: We talked about the pandemic and its impact on certain sector spaces. There’s a lot of conversations in the U.S. and elsewhere about what’s been called the Great Resignation and this scarcity of labor. And some people have this belief, it’s a transient phenomenon, and we’re going to go back. Others of us think it’s more of part of a long-term trend. Do you see more employers coming to you with, “The reason why I’m doing this is because I can’t find the talent that I needed to do it through humans on the job, and so technology is supplementing it?”
Dines: Just a couple of months ago, I had a discussion with one of our clients, a pretty large investment bank. And they told me that in the past 12 months, they had 40 percent attrition. This is kind of unheard of, really—and it’s a good investment bank. It has really good, well-paid jobs. So they were concerned they cannot retain and cannot attract talent. So we are talking, besides the obvious benefits of automation—because it reduces your dependency of human labor. So, even in the face of a massive migration, your business still can operate at some basic level. But we were talking about offering automation as a perk—offering robots as a perk—to their employees because we believe automation literacy will be for the current workforce as important as I think the office tools literacy was 30 years back. So automation is a way to retain and attract talent, that it’s amplified by this Great Resignation phenomenon.
Kerr: Okay. So just as we think about Word and Excel and all the other tools that came, this is going to be a tool you envision on many, many desks going forward.
Dines: Yes, certainly. And our technology addresses different personas. We have tools for more of professional developers, but also we have tools for Citizen Developers. And we’ve built our approach of building Citizen Developer tools was to kind of emulate the Microsoft Excel as in terms of simplicity. And actually, we kind of put Microsoft Excel at the center of automation. So how our Citizen Developer tool works is, it allows a business users to kind of automate simple use cases—like take data from one document or one system, copy it into an Excel spreadsheet, where they can do some formulas using a tool they know, and then take the same spreadsheet and put it into a destination application.
Kerr: Daniel, tell me a little bit also about the broader ecosystem that surrounds UiPath. As you think about contingent workforces, contract workers, gig workers, you have a very porous organization. And so how do you keep that ecosystem aligned? How do you incentivize it, encourage it?
Dines: Well, I think building an ecosystem, a large ecosystem, was a business goal since the beginning of our company. We offered our tools for free to the community. Even small businesses that have revenue up to $5 million can use our technology for free. So we helped a lot of individuals, practitioners, hobbyists to get to understand how RPA works. It helps a lot of people within the service companies to get started by themselves, even without having a formal program with their employer. So we amassed a community in excess of 1.5 million people that are familiar with our technology. That helped quite a bit our customers to source talent and to work with partners that, in turn, source talent from our community. I can say that in RPA, one of the biggest bottlenecks is having people to implement the processes. One bottleneck is finding all the use cases, because sometimes people don’t even realize how much can be automated. And second is finding the right people that can deliver to automations.
Kerr: And going back to your global workforce—and you were describing the inclusive nature that you’ve sought to build in a culture—how do you manage that ecosystem on such a broad and globally diverse scale?
Dines: Well, we have diverse programs to help our people. We have an MVP program for our ecosystem. We are doing various hackathons. We have a huge Academic Alliance program. We have more than 1,700 colleges and universities that are using our program. I truly believe that this is part of the literacy for new employees. Automation is a must-have; it’s going to enhance automatically your job prospects. It’s kind of our passion, the community that we have. It’s a big differentiator. It’s a big competitive mode. So we continue to invest in it.
Kerr: Daniel, think with me a little bit ahead, maybe five years ahead, 10 years ahead. I’d love to know what you’re thinking about and what UiPath is kind of preparing, not just as the next version that’ll come out in 18 months, but in 10 years’ time, where will RPA be? And what will Computer Vision be doing? And how do you see it sort of transforming our workplace further?
Dines: Well, I think if the business world would be static, at some point, we will reach a state where everything can be fully automated by using machine-to-machine API-type of automation. But this isn’t a static world. In a dynamic world, where people imagine new businesses, you do M&A, it’s a lot of movement. You simply cannot stay on the top of it unless you have a technology similar to the human mind. Who is going to create a new business process? It’s businesspeople. How are they going to create a new process? They are going to use an interface that is familiar to them. It’s the user interfaces—natural languages, voice, whatever, but it’s a human interface. It’s the fastest way to cope with the complexity and with the speeds required. So if this is true, you will always need a tool for emulating people. But also, I don’t see our world being in a position where we will be tasked with the same level of mundane tasks that we have today. I don’t imagine a world 10 years from now where a guy freshly from college will just go to websites, find some tables with information, and copy-paste in a PowerPoint. This is not really a job for humans.
Kerr: Yeah, but stay with me there. What are the next few tasks or things that currently humans are doing it, but you look at it and say we could do better? Or we should have technologies to be able to do that part of it.
Dines: Well, we can do better in terms of less errors, for sure. In terms of the speed of execution. It’s clearly technology at a low-level tasks, technology beats humans. Where we cannot do better is when creating new processes and thinking strategically. And this is nowhere. I don’t think technology is near to cope with these type of tasks.
Kerr: So, Daniel, even at the risk of being a little bit too technical, what are a couple of things that are on the more-specific roadmap you’re seeing?
Dines: One of the major initiatives that we have here at UiPath is called “semantic automation.” Our robots today are kind of dumb in executing a process. They don’t know what they do. They blindly follow the steps that were described to them. As an example, for a human user, for a person that knows English and is literate in this world and understands what an invoice is, to describe a task to copy an invoice into an ERP system is just as simple as saying, “The invoices are in this folder, take it one by one, open the ERP, go to this form, and copy the information.” That’s all you have to say. Now imagine, take the same person and do this in Japanese. They will be lost. You will have to tell them every single step: “Find this icon, find this ideogram, and then get to the number next to it, go to this application, same, find, paste it there.” Hundreds of statements. If one of these little steps is not there, the person will fail who doesn’t know how to handle any exception. That’s true for humans. But our robots are more like English people dealing with Japanese invoices. We want to bring them the notion of the language, of the semantics, of the context, because that will make describing automations much easier—like you tell to a person in English—and that will make the automation much more reliable, because you are understanding the context. You can deal with small changes that happens on the user interface. So in this regard, one of the first big features that we are about to launch is what we call the—it’s a semantic copy-pasting feature. So that means any business user can go to any document that you have on your computer or any form, any PDF. It doesn’t matter. Any source of information. You just go to the source and you say, “Copy this information for me.” You don’t have to explain, “Where are the labels? What kind of information?” You just say, “Copy.” Imagine, go to a webpage. “Copy.” And then you go to a destination—for a PDF, whatever—and you say, “Paste the information.” And now we figure out all the mapping between source and destination, all the fields, which field corresponds to which information, from target to destination.
Kerr: I look forward to when that product is available. Daniel, maybe I can close or ask you to ponder one last question. We have a lot of students at Harvard Business School and a lot of entrepreneurs around us that are trying to build companies for the future of work, start something new. And so, if you reflect back on your journey from 2005 until now with UiPath, is there advice that you would leave them with as somebody that has thought deeply about how technology enters the workplace and changes the workplace and so forth?
Dines: First of all, one thing that was very particular to me, leaving a good job at Microsoft and spending 10 years of hardship until we saw some stage of success. To me, I was very afraid of failure and basically becoming unemployable. First of all, you—especially as the CEO of the business—you’ll have to get to know a lot of things from many departments. You are not going to become a specialist into one domain, but you are becoming a generalist that understands many aspects of the business. I was afraid that this is not the skill that people will actually want. I want to alleviate this type of fear that they might feel during building a company. This is one of the most valuable skills that you can build in life. And great companies are going to fight for you even if you fail in your journey. Now in terms of the future of work, I think this is the best opportunity that exists maybe in, I don’t know, since the invention of computers right now. People have not taken for granted the way their parents and grandfather worked. People are willing to leave their jobs if it’s not in according to their principles and their way of conducting life. So it’s a tremendous opportunity right now in this space reimagining work for enterprises.
Kerr: Daniel Dines is the co-founder and CEO of UiPath. Daniel, thanks so much for joining us today.
Dines: Thank you so much, Bill. It was a great conversation. Thank you.
Kerr: We hope you enjoy the Managing the Future of Work podcast. If you haven’t already, please subscribe and rate the show wherever you get your podcasts. You can find out more about the Managing the Future of Work Project at our website hbs.edu/managingthefutureofwork. While you’re there, sign up for our newsletter.