Making your digital transition seamless and trouble-free

Boris Krumrey, Chief Robotics Officer, UiPath 


As the transition to digital races on, robotic process automation (RPA) and artificial intelligence (AI) technologies are converging and will gradually permeate almost every aspect of a business. RPA is already creating a lot of operational value – processes accelerated by up to 80 per cent, and executed with 100 per cent accuracy – and can offer a payback period of only a few months. AI, on the other hand, is a more sophisticated business solution and therefore more complex and costly, so requires a lot of advanced planning. We have customers generating more than a million hours of capacity in one year with RPA while boosting many of their processes with AI, but not all organisations are prepared to go big. They need a strategy in place, they need change-management and they need a skilled team with clearly defined roles. Luckily the industry’s guardians are many and of great value, from advisors, implementation partners and AI partners to developers that are professionally trained into specific RPA roles.

It’s easier to start small and grow fast, and because RPA is, intrinsically, a technology that connects other technologies (working at the user interface level without affecting underlying applications and systems), it will probably establish itself as a gateway for applying AI to the organisation. RPA can be the backbone for further AI investment, a platform that connects systems and applications with AI functionality, evolving together in ways that make sense for the business at each step of its digital evolution – gradually and without disruption. There are use cases already in place that demonstrate the successful fusion between the efficiency of RPA and the sophistication of AI.

To capitalise on the latest advancements, leaders must understand that they are dealing with moving targets. They must keep an open mind and move fast.


Open to try, open to learn. Try UiPath for free now – true enterprise RPA, built for business and IT.


Video Transcript:

Hello, and welcome to Business Reporter's UK 2030 campaign. I'm Alastair Greener. And today, I'm talking to Boris Krumrey from UiPath.

Good morning.

Good morning, Alastair.

We hear a lot about process automation and AI. Perhaps you can help us to start off with by explaining the relationship between the two of them.

Process automation, and in particular, robotic process automation, because what we are doing in that particular field is we are mimicking the user what the user would be doing on the desktop. So being able to do that, we can operate in any kind of environment. Now, the interesting part is if you combine that with AI or any kind of machine-learning, deep-learning technology, you're adding to it another feature, which means you're adding to it something that can make a decision, make a prediction. It's like you're creating a coworker who works with you on your own desktop or workplace. And that's where the vision is.

It's exciting technology. But perhaps you could help explain it by maybe giving us some examples of typical applications where you could use this.

So a typical application for robotic process automation, to start with, would be around claims processing. So just imagine you're a big insurance company. You'll get a kind of a filled form. And first thing, you have to be able to extract the data from that form.

So this requires computer vision. So you need to be able to identify where the data fields and objects you need to extract. And then you have the process automation part that helps actually then automating, putting it immediately into action.

So you're already using here AI technology, in a sense. You're using computer vision. You're using maybe a machine-learning technique to identify objects on a particular kind of form.

So claims processing, also in financial services-- a lot of activities you'll do in that particular area that fall under particular rules. And whenever you need a certain kind of capability of making a decision or a recommendation or prediction, that's when you'll embed custom-built machine-learning models. And you use all these cognitive technologies.

Obviously, lots of people are embracing it, as you've said. There is already a demand there. But for those who are still thinking about it, what are the typical concerns that maybe the C-suite have about incorporating RPA and AI?

That's an interesting-- it's actually two different paths here, because AI is the more difficult animal to bring in or to kind of master here, because AI really depends on the data, on the level of digitization you have brought to your company. So if you don't have the data available digital, and if you don't have data scientists who have mastered the art of labelling the data, in a sense, so that they actually can solve a machine-learning or artificial-intelligence challenge, then you cannot actually use or unlock that capability, whereas an RPA is-- so like our software, you can download it. You can immediately employ it. You can record yourself. You generate the automation, and off you can go.

Still, the bigger challenge for all of them is when it comes to scale. So if you're rolling out process automation at large scale to thousands of employees, you have to do change management. You have to prepare the organisation. You have to communicate.

We say, you know, don't fear. This is not about replacing you. This is about making the whole process, the whole operation of our company far more effective and productive. So that is something that definitely will always apply when you look into kind of bringing it to scale.

And what about once you've incorporated AI into your organisation? What about some of the pitfalls that can happen during its use? So over a period of time, what are the kind of things that the C-suite should be aware of?

Well, the biggest challenges-- again, if you want to start the track on AI path very early, and you haven't even got the data to substantiate this entire journey, you haven't even thought about how to digitise data that is currently not digitised, then you are far too early to start. So I think applying robotic process automation first is probably a first good head start, because you're covering all the rule-based automation. So that's the first one.

The second one is when you do process automation, start a power track on the analytics part. So this is where you do kind of deal with big data, because when you start that, you start thinking about data. You start employing people who have kind of more the data science and data analytical skills.

And then you have to later on expand your team to having real data scientists and data science engineers. So the difference is almost like if you think of a building, the architect is like a data scientist. And the engineers are the ones who are then actually putting it all into place and then into action.

Judging from what you're saying, AI is something to be embraced, not to be worried about. So maybe to help people understand some of the biggest advantages of AI, you could give us some examples of where it's really been transformational.

So I think the most common example is if you think of driverless cars, right? And I think Tim Cook, when he says it is the mother of all artificial intelligence, he's not wrong because that's the first time we actually hand over trust. We lean back and let the computer drive.

We're not realising actually that this is constantly happening already when we are stepping on a plane and we are flying to some other place, because the whole plane is a robot. And the whole plane can land and start, fly without any human intervention. And you're just there, kind of observing.

And in many cases, of course, the computer is much faster than a human. So in the same case here, the biggest kind of convenience we can see-- you don't need to drive anymore. You can just put in your destination, and it's all been taken care of.

Let's say you've won me over, and I'm in a position to incorporate AI. However, I'm still aware there might be some pitfalls out there that I need to be careful of. So what are sort of some of the things that I need to do to make sure that I avoid them?

I think one of the key points is once you've successfully at least found that a challenge-- a business challenge-- can be solved with a machine-learning or deep-learning solution, and you have moved on to saying, OK, here's my deep-learning model-- it could be a model for giving you details around which buyers are the best buyers for a particular product, or there are so many applications how you can define the model. But then you need to keep the model, in a sense, maintained. And this is where the challenge is from an enterprise level.

If you think of banking and where risk is a very important factor, we embrace AI and machine-learning technologies. And then on the other hand, but we are not too happy about an unpredictable outcome, right? So because the machine learning, with more deep learning even more, would in fact learn and modify its outcome based on the data.

So it's not just an algorithm. It's also the data. And it could even be that with the addition of new data, it could even be becoming worse than before. So the maintenance and control around this particular in the enterprise is a challenge.

And what about teaching AI? What are the alternative methods of doing that?

Today, we need to have a very, very good mathematical understanding of this. In future, we would have kind of ready-made sets that you connect, like LEGO sets or like, oh, if you like, an electronic construction set, where you just say, I'll take this element. I'll take this element.

And then out here comes my yes and no signature recognition algorithm, for example. This is how the future will be going. So then it would be far easier for normal people, or even just people who are kind of good programmers and so forth, to just plug those pieces together and say, this is how it works, right.

Different countries are developing and using AI at different rates. Which countries are doing a particularly good job and which countries are really embracing it? And how are they embracing it?

Well, we are seeing clearly Japan is driving this initiative and particularly strong, not just AI but in combination with robotics process automation. And it's not unsurprising, because they were the country that achieved the breakthrough of industrial robots. It's also interesting to see interesting developments also in Finland.

America is coming also. There's a huge demand in America. But I have to say the countries that embrace it with their whole kind of cultural setup is Japan, and we see some Nordic countries, yeah, as well in Europe.

Tell us a little bit more about your platform, the UiPath's RPA platform and how it works.

Right. So UiPath's platform consists of three key components. So the first component is the robot. And you have to think of it, it's like a runtime execution thing. It's just a process-- sorry, a programme.

And you can load any kind of process on top of it, any kind of process flow description or whatever. And it will translate it into keystrokes and activities that a human would be doing on the PC. So that's the robot runtime.

And then we have the orchestrator, and in the same way like an orchestra, where you have multiple musicians, and then each musician is basically the robot. And the scores would be basically this process definition file. Then the conductor would basically, you know, bring it all together and make sure that everything plays with each other in the correct way. So if you think of large scale, if you want to make very powerful, big music, then you would need kind of multiple robots. And you need one orchestrator kind of to coordinate that.

And the third component is a studio. It's basically where we create those simple workflows, where you can record your activity you're doing on the PC. And then it visualises the whole process in a very simple way so that non-technical people can even understand what's happening.

We've seen rapid change in this area. So what about the future? Where do you see AI and its self-learning capabilities develop further over, say, the next five years or so?

We will have enormous, powerful tools that will help us to do our job on a super-human level. So the knowledge worker of the future will be much more empowered. Best example-- if you think about what will come very soon is that all the call centre solutions will advance significantly.

We will not need any more humans to cover all the standard cases. If you call, it will be a robot that will kind of answer and cover all your standard transactions. And only the exceptions will kind of switch over to a human to help out.

So you could think, ooh, what will this mean for all the contact centres and call centres? Will these people lose their jobs? No, because now you're commoditizing and making it kind of so much more affordable to have a real customer service hotline.

If you could have three key takeaways that AI decision-makers could have, what would they be?

So I think the first one is, you know, understand your data. Understand, you know, what's your vision around the data? How do you digitise data? So data is the key point for when you think about where you want to go with AI.

Secondly, think about where do you want to go with your automations? Or what can you automate? How can that be done? And applying technologies like RPA is easy now because you can then combine those two.

And the third one is think about how the human-- how will your employee connect to all these technologies? How will they work together? So it's the orchestration. It's the human in the loop, as we say in this area. How is that connected?

And when you're able to put all these three together-- and I'll say even a fourth factor is think about scalability, because you have to then bring those to the scale. Then you're covering kind of the important answer of how do you get best benefit out of those technologies in a fast way, you could say.

And we're going to see a huge amount of this as time progresses. It's been fascinating finding out more about technology, about AI, and also about RPA as well. Boris Krumrey from UiPath, thank you very much indeed.

Thank you very much, Alastair.