Daniel Dines is the founder and CEO of UiPath, one of the biggest automation companies around. UiPath sells software automation or what consultants call robotic process automation, or RPA, so they can sound fancy and charge higher fees.
RPA is actually very simple to understand. Lets say you have something in your business that relies on older software to do some repetitive task like entering billing information or moving data from one system to another. Now, the intuitive way most of us would think about making all that more efficient would be to upgrade or replace that old software with something with more capabilities. But, as weve all learned by now, new software often causes more problems than it solves; there are compatibility issues, stability issues, and the general chaos of rolling it out and making sure it all works.
UiPath and other software automation companies have a different approach: just hire another computer to use software for you. Seriously. UiPath uses computer vision to look at whats on a screen, and then it uses a virtual mouse and keyboard to click around and do things in apps like Excel and Salesforce. The automations can be mundane, like generating lists of people to contact from public records, or intensely complicated: UiPath can actually monitor how different software is used throughout a company and suggest automations. Huge companies like Uber, Facebook, Spotify, and Google all use UiPath.
Last year, I talked a lot about the social consequences of automation with New York Times reporter Kevin Roose hed just published a book on automation and the workforce. But I was really excited to talk to Daniel about it directly and hear his perspective on competition in particular, especially because UiPath has had a pretty up and down year.
The company went public almost exactly a year ago in one of the biggest software IPOs ever. Since then, the stock has taken a nosedive. (It IPOd at $74.84 a share, but at the time Im recording this, its just $16.34.) I wanted to ask Daniel how UiPath was built for this moment, how mainstream he thinks automation like this can be, and how hes thinking about big competitors like Microsoft and Salesforce.
Of course, we also talked about those social impacts. Daniel had some pretty interesting responses to those questions. He thinks giving boring work to robots instead of people makes the people much happier and might keep them from looking for new work. Daniel tells the story of a company seeing upwards of 40 percent attrition. They eventually turned to UiPath to give their people a lighter workload to hopefully make them like their existing job more and keep them from quitting. We often worry about automation taking jobs away, but its interesting to hear Daniel talk about how automation might help companies retain their employees.
Daniel Dines, founder and CEO of UiPath. Here we go.
This transcript has been lightly edited for clarity.
Daniel Dines is the founder and CEO of UiPath. Welcome to Decoder.
Thank you so much for having me, Nilay.
I have been excited to talk to you for a long time. Robotic process automation is one of my favorite subjects, and you are the guy to talk about it with. It is almost exactly the one-year anniversary of UiPath going public as one of the biggest US software IPOs in history, at $1.3 billion last April. I think its important to dig into software automation, which has the potential to reshape how everyone works. I want to talk to you about what the last year has been like, and where UiPath as a company goes from here as the market reacts to automation existing. I think one of the big challenges for UiPath and for everyone else is that robots are going to become a class of users right alongside people. Lets start at the beginning. What is UiPath, and how did you end up starting it?
Well, we are an enterprise software company that specializes in enterprise automation, using a very different flavor to automating business processes compared to any other technologies. We simply emulate human users. Its more like the self-driving car version of the automation, and that has the tremendous advantage that it reuses the existing workflow operations that are proven in enterprise. We just code them into a software workflow, so it considerably speeds up the process. Its more cost-effective than any other type of automation, and it proves that its the only technology that can scale for the long tail of manual processes.
I just want to unpack what it means to emulate a human user. Correct me if Im wrong, but in the world of enterprise software, there is a tendency to dress up simple ideas into language for CIOs and finance people. When you mean emulate human users, you mean your software uses a mouse and keyboard to use other peoples software.
Yes, thats very well said.
Okay. I just want to be clear. You have a robot that moves a mouse around a screen and clicks on things in Microsoft Word or wherever.
Yes, but its not a physical robot. We do not use a physical mouse and a physical keyboard; we use an emulation of mouse and keyboard on a computer. Its virtual. You can move the mouse on the screen without physically moving the mouse. This is technology that any computer provides. We are within the computer, but we are seeing the screen and we are operating like a human user.
So again, just to make this as dumb as possible and as accessible to many people: You have an application that emulates a mouse, a keyboard, and a display, that can see whatever application is on a screen whether its Salesforce or Microsoft Word or Adobe Photoshop and you write automations there to operate that software.
Yes. If you want to send an email to someone, you would open your email client, start a new email, enter the title, subject, to whom, and then type what you want to say and press send. We can code this workflow in our software, and you can literally see on the screen how the email client opens and these actions are replicated to send the mail.
There are enormous implications for how all this works, but this is actually a pretty simple, easy-to-understand idea for how you might automate software. How did you come up with it? What was the genesis of this idea? How did you end up starting this huge company?
The idea is old, but it has not been used to automate business processes; it comes more from the world of testing. As a software engineer, you have to test your applications, then many people use this technique which is called regression testing to simply emulate a user doing this. That is where the idea came from.
Initially, I really didnt understand how big the market would be because I was naive. I was thinking that all these manual processes did not exist in big enterprises, and that they had already been automated by various other technologies. We thought this was a small use case more in IT automation, ITSM [information technology service management] type of use cases initially.
In order to unlock this idea of emulating people, we built a low-code/no-code environment because this was key to what we were doing. We reduced the technical skills required to build end-to-end process automation for people that have some kind of knowledge about programming but are not experts. Paired with this approach, we saw our first big usage in the BPO industry, business process outsourcing. When we saw this market for the first time in 2014 or 2015, they were under big pressure to continuously deliver more benefits to their clients year over year. They squeezed everything they could from the process optimization, using techniques like Lean, Six Sigma, or similar. Back then, automation was the only way to get more and to reduce the cost of their offering.
They started to use this type of technology, and very soon, their clients got wind of it. Oh, this is something very interesting that can give us back the leverage in the relationship with outsourcers. People realized what a huge return on investment it generates. And by our knowledge, this is the highest return on investment technology that exists. We are currently seeing people that invest $200,000 and generate $5 million in return; we have companies that invest $10 million and generate to the tune of hundreds of millions of dollars in return on investment.
Let me ask you about that real quick, because I have always wanted to dive into that specific return on investment. You are saying that a business process outsourcer, a BPO, is your external accounting firm or your tax provider; they are someone on the outside who is running some back office function such as accounting, billing, or invoicing that needs to be done to run a business. You sell to them because automating some of that work is the easiest way for them to increase their margins, lower their costs, provide more services, and get rid of their warehouses full of accountants. This work comes into a company because they see it happening and that they can generate a $500,000 investment for millions of dollars in return. Are they generating more revenue, or are they cutting costs to generate that return?
Well, I think its both. Let me explain the journey in an enterprise, because that will make you realize how the flywheel of automation works. We land into a department like finance, HR, or procurement and lets say we help them build a center of excellence. They have a certain number of KPIs about their automation program they use to measure the results. The most important one for us is the number of manual hours that we return to the business. We need to have someone like a controller or CFO in that suite that signs for these KPIs, which are strictly measured. Once the program is put into place and this unit returns the manual hours, its very usual that the controller will go back to them and say, I want to invest more in this. This is the technology that generates so much for me.
The cost is reduced because you can deploy people from low-level tasks to higher tasks; they can produce more and actually increase your revenue in the company. For example, if we talk to one of our big telco customers that has a negative NPS [net promoter score], for instance, we help in their contact center.
They all do. Every telco customer has a negative net promoter score its how much people talk about how much they like your company.
If we help their agents to engage the customer more and help their NPS grow, its actually a good way to increase the revenue for providing better services. So its both.
I want to come back to the very simple idea of what the core product does to help these service reps. If you are a telco customer service rep, you use some piece of internal customer management software every day that may keep you from interacting with customers because its bad, hard to use, or slow. I am sure anybody who has called a customer service agent has experienced frustration with this. It implies, Instead of fixing that software, just hire us to have our robots use it for you, and then your people will have more free time.
Well, its not instead.
Okay.
Its in parallel. What is the alternative?
I would say fixing the software is the alternative.
Yes, but I am saying to do that in parallel. That is very interesting. I had exactly the same discussion with a very big bank in the UK, and they said, We are going to change all our legacy software and contact center. We will standardize on Salesforce or Service Cloud CRM. I asked them how long this project would take maybe three years in total and if it could be replaced in one single step. It is impossible because its not only two systems the agent has to integrate and replace, but 20. If we start placing our automation layer on top of the old system, we make the job of the agent simpler. They will only interact with our software, while we abstract the underlying systems.
Then IT can go and replace those systems at their pace with better testing and better results than just doing it in one step. The CIO of that bank agreed with me and this is what they put in place. All software will continue to be replaced by new software; this is how the industry works. I think its important to point out that our customers are not only very mature businesses that have been through the mainframe era. We also have a new breed of software companies that are only cloud-based as our customers. I can tell you that Snowflake, CrowdStrike, Uber, Spotify, Facebook, Google are our customers, and they dont have legacy software. You know why? They are using us because this approach of emulating people is the only one that works at scale.
I agree with you. I dont think large-scale enterprise software transitions are easy for anyone. I do think one of the interesting components of this is a recognition that people have to use software, and that software might actually be the bottleneck, even though the software is the job.
If my job is to use Excel all day long, a recognition that Excel is the bottleneck is a mind-expanding idea, compared to any of the other bottlenecks you might face in a job, like, Im waiting for someone or, The order hasnt come in. I dont think many people consider, Your job involves repetitive use of software and the software gets in your way.
Software is a tool, Nilay. It can be a better tool, or it can be a shitty tool. If you look at the best-crafted software lets say ERP [enterprise research software] systems, which have been around for 30 years they have a lot of pre-coded processes. They come with the way you should run your business, but cannot include all the interactions between different external systems. You live in an ecosystem, and there will never be one single piece of software that does everything for an enterprise; nobody can put all the optimizations in one single instance.
Companies like SAP are pure examples. You cannot really do a fully automated enterprise on their instance, and people are reluctant to even code into their ERP systems. It is dangerous to record something into the core system every time there is a change to a process, so people dont do this. There is always another layer that sits on top the automation layer. It is easier, cheaper, and less disruptive to the business to put your automations, your operations, on this layer rather than into the core systems.
What do you think the limits of automation are right now?
I think the limits in technology are more around natural language processing. If a process that needs automation requires understanding a lot of natural language to accomplish, it is more difficult to automate. Anything that is repetitive in nature, even if you require intelligence in the process, like reading documents or invoices we can handle. But the moment you need to get to higher cognitive tasks, that is the limit.
A concept that we call humans in the loop is actually embedded in our software, so you can do a big process that involves multiple users and hundreds of tasks. We basically organize tasks like that like a big game of ping-pong between humans and robots. If you send an email or text to one of our robots asking for something and we cannot understand the intent, we parse the request as much as we can. If we completely understand those requests, we go and automate them. If we do not understand them, we will create a test for a human user and they will give us a more structured format. Then that request is passed back to the robots and so on. This is how it goes. Like a game of ping-pong. Back and forth, human to robot to human to robot.
We remove the mundane tasks from peoples daily work while they deal mostly with the exceptions. That makes the job much more enjoyable and the output of the people working in operations higher.
One thing I am really curious about is software products. They are made for their users, right? The product managers of any software product think about who is using the software, and then they write features for that person or they write user stories about that person. Have you seen any software adapt itself to UiPath now that a core user base is robots? That seems like the craziest feedback loop of all time.
Im not sure I get it. You mean that users will be adapted to how the robots operate?
Right. If you are the product manager of Nilays Enterprise Software and you know about 50% of the users of the software are UiPath robots, I had better optimize the software so they can use it more easily. Have you seen that feedback loop?
Well, not really so far. I have seen our clients putting more pressure on their vendors and being more predictive with the changes to the user interface that might break the robots.
Its interesting, I once had this discussion with a guy from Microsoft that works in their interface group. They are also anticipating a world where enough pressure will come from software robots using the user interface that they will have to treat it almost like an API. You cannot just break the existing things. Right now people are changing user interface much more freely, sometimes stupidly. I dont understand what was in their mind when they changed the interface. Maybe the user interface will become more like a contract, like API is supposed to be.
That or you will have two user interface modes, the robot and the human.
Possible.
There is just a world of software design here that I dont think has yet responded to the rise of automation. I am just dying to see what that looks like, because it seems very unlike anything that has come before.
Thats a great point, Nilay.
Lets talk about UiPath itself now. This is what I think of as the Decoder questions. How many employees does UiPath have?
Its almost 5,000 people.
That is all around the world, right?
Yes. We are very much distributed around the world. In fact, as a business model, we have been one of the very few companies I know that have started their expansion globally from day one. We started in a small country in Europe in Romania but when we got product-market fit, we expanded simultaneously to the US and Asia. Japan was the fastest revenue growth for us in the initial years.
It is a really interesting growth story to expand globally right away. That obviously implies lots of things. How did you structure the company to do that?
Well, I would say I didnt structure the company. In the beginning, I think the key was to offer people a lot of freedom to do what they want. I was working very closely with the people in sales, and one day one of our sales leaders came to me saying, There is an event organized by PWC in Japan. I would rather go there. Maybe its a good market for us. I said, Why not? Just go to Japan. It proved to be an awesome market for us because they were prone to automation. It was not that we had a big strategy. It was more of a test; go, use your best judgment, and then spread.
It was a bit more like Genghis Khan and the Golden Horde go faster than your competition in the richest places. You need to find where those are and then just go and occupy then move on. This is what we did in the early years, and it was very successful.
You are now a public company with 5,000 employees. How is UiPath structured now?
Well, many people in the first years would say that we were becoming a corporation, that we were more bureaucratic and hired a lot of senior people from the external world. I am trying to keep a balance and be a disciplined company that generates predictable results quarter over quarter while keeping the soul of an explorer, so that people have the freedom to try things, to break things, and to find new opportunities.
One of my mantras that I am trying to instill in this company is, Always challenge your boss. I even say we have a no boss culture, and the more-senior people are always taken aback and ask, What do you mean by this? It is very empowering to think, I dont have a boss; I have a partner, and I need to be capable of speaking very freely with them. That is one of the most important things that I am trying to achieve here.
Well, that leads right into what I think of as the classic Decoder question. How do you make decisions?
I listen a lot. In this interview, I feel like I talked a lot, but that is not my style. I am trying to learn more by listening to people. Obviously, I have no idea how to run a big company at this stage because I have never been in a similar situation before, but I am trying to build a very close-knit executive team that relies on each other. Our way of making decisions is to do them very fast if they can be reversed because its always a cost of decision and do them very slowly if they are irreversible.
One of the interesting stories about building the framework of this company was about the culture. We were like a bunch of kids 10 years ago, thinking about how to define the culture. We came up with lots of words like, We are open and transparent, but that ends up diluting the culture. I had this epiphany one day and I said to myself, We need to define the culture by one single word. Lets start there.
Going back to our roots, I found that humility is the word that should define us. Not in the sense of being submissive, obviously, but in the capability to listen and change your mind, your ideas. I dont really like people that believe, Im a big guy. I know what Im doing. I got it. This is not our style. If you cannot prove that you are capable of listening or changing your mind, this is not the company for you.
So let me just add those two things together. You are the CEO, and at some point the people who work for you even though they are not bosses either just want a boss to make a decision. How does that play out in practice for you? You have listened a lot, you want to be slow on irreversible decisions, and you want to have the humility that you dont know everything. At the end of the day though, you have to decide when its time to go public. How do you make decisions like that? How do you empower people underneath you to make those decisions and tell people what to do?
I made the decision of going public the moment I took external funding into the company more than five years ago; our first round of investment was in July 2015. Everyone should subscribe to this idea, the moment you take external funding is when you basically have to decide over some kind of M&A. The better question for me is, Why now?
We went through an accelerated maturing phase in 2019 and 2020, and the market was just ripe I think almost the best ever for an IPO. We knew that the window for an IPO opens and closes, and we were ready from the internal systems and the predictability of our revenue. We had more than 3,000 employees at the time we IPOed, and we wanted to give them a way to cash out after all the years they put in. It was a bit of a no-brainer for us at that moment.
Public markets are a brutal stage but I think its very simple to explain why. As much as we tell ourselves and our employees that we are here for the long term, the fact of the matter is that every day you see a movement in the stock price and it affects us because its tight. Many people I work with are leveraged on the UiPath, so it has a direct impact on them. In a private company, maybe they raised money in August last year with high valuations, but right now they are not forced to adjust their internal valuations. Even when they hire people, they can say, This is my valuation. This is not real money in a sense, so they are less affected by these movements of the stock.
In our discussions with investors, its a different level, having private investors and having public investors. In a way we have been forced to mature in a condensed amount of time. I felt that makes us a bit more competitive and more resilient. So it has benefits, but its not easy. This is what Im trying to say.
You went public on April 21, last year. What is the biggest lesson you have learned in that year of being a public company CEO for the first time?
The biggest lesson for me was never lose the grip on the company. I think all of us felt a bit too relaxed after a very successful IPO and stock pricing going through the ceiling, and we were maybe a bit drunk with our own success. We didnt make big mistakes, but I felt that we started to drink our own Kool-Aid and could have done better, honestly. We are working in an amazing market in its early innings. We could have captured more of this market if we had stayed laser-focused. You can see this lesson across many recent IPOs.
Lets talk about that market. I think that is a really interesting thing to focus on.
You have a product, people like it, and you are selling it to a lot of big companies. UiPath has a ferocious product development cycle. The other side of that is sales. You could grow your market by selling the product you have to people who might not know about automation, you could apply it to more places with your existing customers, or you could develop new products to attack new segments of the automation market. Whats the split there for you? How do you decide on making new products for new uses, versus do you invest in sales to get new customers, or sell more to existing customers?
I think the only way is to do both and simultaneously. When we saw the big market expansion, which started seven years ago, we had two big competitors. One of them was the most advanced at that time in terms of the overall technology, but they stopped investing in that and moved to invest more into go-to-market.
It was still early in the market, so they got the lead. They got a very early IPO button and more penny markets than on the big market. That happened in the UK, and I think they completely stopped the innovation in the company. If you look at their product seven years ago, you can see big differences. We invested hugely into product development, because we had brilliance.
This is part of what we call computer vision. You look at the screen and you understand whats on the screen. Which one is the button? What are the radio buttons, the selections, et cetera? We understand very well. That was our secret power, and we are much better than everyone else on the planet. Starting from this brilliant thing that we had and developed over time, we invested hugely in building the wide spaces around the product. We went from being a distant number three in the market in 2015, to number one by 2018. Since then, the lead is increasing. After we got product-market fit at the beginning of 2017, we had a kind of overarching product capable of automating end to end. Then we started to invest massively into the go-to-market and then it continued.
I think this is the right go-to-product-market fit. If you invest faster than this you will be bored, and you may not be capable of hiring the best sales people out there. The moment you get product-market fit, its kind of a self-fulfilled prophecy. The product sells itself. Sales people smell it. You get better and better people that make bigger and bigger deals.
Its amazing how the best sales people always gravitate towards the product that sells itself.
Exactly.
Theres something there. It would take another hour of Decoder just on that idea.
I am curious about challenges. You do have some competitors that are in different stages. Many years ago when I was with Microsoft I saw [Microsoft CEO] Satya Nadella, and he had just discovered robotic process automation. He was glowing about it, telling a room full of reporters and analysts that robotic process automation was a big new market for Microsoft.
Obviously, I think one of the key pieces of software that UiPath automates is Microsoft Office, specifically Excel. How do you see that relationship with a company like Microsoft or a Salesforce? You are using computer vision, so you dont necessarily need permission from them, right? Your software can see their user interface, can move the mouse, can click the buttons. You can do that in a repetitive way. Do you need to work with them? How does that work?
Yeah, we have a good relationship with Microsoft. I used to work for Microsoft, five years before starting this company. I even met Satya, and he is really an outstanding person. We gave him a demo of our product. To us, Microsoft getting into robotic process automation created more awareness. It was a good thing in a way.
I think its also clear to the markets where Microsoft plays in relation to the RPA right now. They have this tool called Power Automate, and because its more API-based, they also acquired a small company out of Greece to do the computer vision part and combined it. But Microsoft is best when it comes to personal productivity.
They can do really light, simple use cases with their approach to RPA. Particularly those where subject matter experts can do it themselves. This is where Microsoft plays. Now, going into enterprise automation when you have to take an entire process like procure-to-pay and order-to-cash this is not where we are seeing them. When we were seeing them, we always won in the enterprise automation market. At the same time, our cloud software is based on Azure. We have a very good relationship with the Azure side of Microsoft. So its more of a coopetition between us and Microsoft.
Thats a great word. One of the issues with being in coopetition with the big players, especially Microsoft and Enterprise, is that youre Slack. One day Microsoft is going to say, Well, now Teams is free, and they are going to crush you out of the market until you have to sell yourself to Salesforce. Is that a thing they can do with Power Automate?
Can they just bundle it into their product, or go to their consulting firm partners I know UiPath sells a lot through the big consulting firms and say, Look, were going to start bundling these capabilities into everyones Office suites. You can build your business with a tool that your customers already have. Is that a worry you have in that competition?
Theyve done it already for two years now. They deliver a free version with Windows, but we are not seeing competitive pressures on what were doing best. It is the medium to complex processes really getting the return on investment done. This is why this is a particularly different proposition than Slack. A good-enough product that you can deploy for everybody across the enterprise is not going to make a huge difference to any return on investment.
Slack is not a return on investment game. Its a different type of game. We are more a tool that customers are buying. I use this crude metaphor even with my sales people. If you order food and they give you a very nice steak, are you going to use the plastic forks they send with the food? You get my point.
I do. I have never heard anyone describe Microsoft Power Automate or Slack as plastic forks and knives before, but I appreciate the metaphor. Do you think the Salesforces and the Microsofts are the SAPs of the world? Outside of the business reasons, is there any technical reason that could prevent UiPath from operating? It doesnt seem like they can stop you from operating through computer vision and KVM, or operating the mouse and keyboard. Is there any risk that they might find a way to do it?
I have seen attempts to stop customers from using this approach by charging higher licenses. I have seen all of this, but in the end, they have to do what is in the interest of the customers. For instance, we are pretty big in healthcare and there are a few ERP systems there, but they are not particularly pleased with the way that we read their screens. But when you have big customers like Cleveland Clinic and Mayo Clinic, and they say this is how we are doing it, what can you say? Were not competing. This is the situation with SAP as well and with Salesforce. There is only so much you can do within your system, and you have to recognize that there are very few companies that will base all their operations on one system.
The moment you get out of one system, you need more of an independent player that is capable of operating equally well on all the systems. I want to also make sure that you understand. We started from the UI piece from the user interface piece using mouse and keyboard, but we equally have an API approach. Today, its about combining and doing whats easiest and more reliable from both worlds.
I will ask you this broad question, but I dont know how far back I want this history to go. I would say the history of depending on other peoples APIs to build your business outside of the operating system context is not a history full of happy companies that thrive. You want to build on the Facebook API, your business is going to go away. You want to build on the Twitter API, your business is already gone. You want to build on Microsofts API, they might be able to take that away from you.
The reason I have been asking about computer vision is because that is the thing they cannot take away from you. It seems like a more reliable place to live, rather than trusting that API access will persist, or be dependable, or not subject to business pressures that actually turn it into leverage over us.
This is a point that I never reflected on really. That is actually true, they cannot take this away because they have to make the software usable to human users. They need to have a human-readable interface. The moment you know how to operate this human-readable interface, you have absolute freedom.
To us, what was the key? It was not necessarily that people are afraid of using APIs, I am not going this route. I am saying its easier to build automation on top of human interfaces rather than APIs. APIs dont match in a one-to-one to what you see on the screen. To do an operation like creating a new opportunity, its involved differently. It is kind of a complex business to understand the corresponding API, but anyone can understand the human interface. Just emulating how I do its much easier and requires a lot less knowledge.
We are way more cost-effective in implementing automation at scale because we reuse the workflow. With API, most of the time you have to change your existing operations. Every change requires intensive testing, reeducation of the people, and getting more project managers that understand. When I take the existing operation and replicate it one-to-one in software, I can use the same people with less volume of work. It is the only way you can get to the long tail of manual processes. This is where nobody can actually compete with us.
Take Microsoft. If you have 1,000 processes, they can maybe automate 10%. I can show instantly that using us for 1,000 processes, the return on investment is much higher than 100. And even for those 100 processes, my tool is sharper and I cut way more slices of bread than with the plastic knife. Even if they pay you to do this, its still not worth it.
Visit link:
UiPath CEO Daniel Dines thinks automation can fight the great resignation - The Verge
