Editor's note: this blog post was published in 2016. For an up-to-date look at artificial intelligence technologies, visit our RPA + AI page.
UiPath recently spoke with Andrew Anderson, CEO of Celaton: the first company to create and apply intelligent automation technology to business processes.
Our conversation reveals what the man whose company is recognized as the industry leader for innovative cognitive solutions thinks about: how cognitive automation differs from RPA; how the two technologies fit into an intelligent automation solution; the current maturity and value of Artificial Intelligence; his vision for 2020 – and much more.
UiPath: You’ve described Celaton’s inSTREAM software as being like “the best knowledge worker you’ve ever hired.” That sounds a lot like robotic process automation. What would you say are the key differences between Celaton technology and RPA software robots?
Andrew Anderson: I realise it’s a broad description but the term knowledge worker refers to those that do many different tasks and have different skills. It’s easy to regard our technology and RPA as being the same and it wasn’t that long ago that we were all considered to be competitors. However, our technologies are very different but complementary to each other.
First let me say that from a descriptive point of view we are talking about robotic automation and cognitive automation.
If you draw a line down the middle of a page and write “robotic automation” on one side and “cognitive automation”, which is what we do, on the other side. Belo
w “robotic automation” write the words “structured data” and under cognitive automation write “unstructured data”.
Structured data is typically generated by systems within an organisation whereas unstructured data is typically generated by people and outside of an organisation. That’s the simplest way to explain the difference.
What makes them complementary is that if you have a solution with the capabilities of both robotic and cognitive automation, then you also have the ability to streamline and automate the handling of both structured and unstructured content.
That’s why we’re working with UiPath, they bring a greater awareness and understanding of automation and together we can deliver true intelligent automation. Cognitive automation is the natural next step from robotic automation.
And so the term knowledge worker in our world refers to those people who are required to deal with unstructured, unpredictable and descriptive content. It’s a very labour intensive job that requires time and effort but is open to errors and omissions. The people who do this have to have the necessary experience to make judgements and decisions based on how they interpret the unstructured data.
In these types of business processes, Celaton and its cognitive automation software can not only understand meaning and intent but then recognise the key data buried within the content that is required for the process. Our output then becomes the input into UiPath. Working together, we’re an incredibly powerful combination.
UiPath: Would claims processing, where experienced claims adjusters handle complex exceptions to standard adjudication rules, be an example of how cognitive automation solutions and robotic automation solutions can work side by side and complement each other?
Andrew Anderson: Indeed, yes, that would be exactly the case. The nature of claims makes them some of the most unstructured and complex transactions that organisations have to deal with. Lots of transactions in different formats received through many media channels.
Organisations have little control over how these transactions originate but they still have to apply lots of resources to deal with them. In simple terms what we are doing is making these unstructured transactions structured; therefore, our ‘output’ is often the ‘input’ for robotic automation.
Some organisations believe that moving claims and customer correspondence handling from paper into email makes it structured. But it doesn't, because the customer is still sending you a description of what they want, no matter how it arrives - by post or email or social media.
The content is still unstructured because it’s descriptive. The sooner an organisation can understand what their customer is saying the better they can respond and serve that customer. Regardless of how much anger or frustration a customer may express in their message, its essential to understand what they mean - then swiftly respond swiftly in a personal and appropriate way.
In an ideal world, consumers would use structured data to complain, claim or request what they want. However, consumers don’t do that because they are people - and people use their own language to describe what they want.
From our experience, there are a thousand ways that people will describe the same thing. Our software has to learn to understand what that means and then make it structured.
When the data is structured, robotic automation can be tasked to take it along to the next step. That might be anything from ‘check this data’, carry out a calculation or create a new record in the case management system.
UiPath: Celaton introduced its technology 12 years ago, whereas even two years ago RPA was largely perceived as a swivel-chair automation solution. What factors do you believe are behind the surge in RPA interest & adoption over the past 18 months, and how have those factors also impacted Celaton?
Andrew Anderson: In the last 18 months I’ve seen a growing appetite amongst ambitious companies for competitive advantage.
That manifests itself in different ways, depending on who you speak with and what industry they serve. Serving customers better is one driver but more obvious is their desire and need to reduce costs. Regulations play a part in the demand for new solutions but it’s also the natural progress of things. If there is a better way to do something you can be sure that someone will find a way and monetise it. Then it’s a case of seeing if there is a market for the new technology. I’d suggest that RPA has been around for many years but the market has not been ready until recently.
The growth in demand for robotic automation is good for us because we’re considered to be the natural next step. That and the ever improving performance of hardware and Internet enables software as a service to gain greater acceptance.
There’s also the old saying about “sanity in numbers”. As more companies are seen to be benefiting from use of automation, more will want to use it. Every day there are more and more stories of automation technology creating transformation; and it’s those case studies that spread the word and create the growth.
It’s been a long journey for us. We started developing our technology platform in 2005. Prior to that I had built a software company called RedRock over a period of nine years, I floated it in 2001, sold it in 2002 to Netstore plc - then bought it back in 2004 and renamed it Celaton. While it was under the ownership of Netstore I learned a lot about software as a service, so when I bought RedRock back it seemed natural to develop an ‘as a service’ platform.
I had the idea we could create a technology platform that could do what no other technology could – which was to handle all the unstructured data flowing in from outside of an organization.
Without a technology solution, people had to tackle that content - and it was all very labor intensive, slow and prone to errors.
Our first product was very simple and enabled customers to take in paper documents, OCR them and put them into archives. But our objective was always to develop technology that could understand the meaning of content, then use that understanding to complete the process autonomously.
We thought we’d crack it within a year, but it wasn’t until 2010 that we had something that we now understand as artificially intelligent. Even then, we didn’t get to financial breakeven until late 2011. In 2012 we started the hunt for some growth capital and the following year we raised investment from BGF. In reality, you could say it was in 2013 that the journey really started. It’s taken time and effort to help create the market that we see emerging now.
This happens with every new technology and we’re seeing it in the automation industry. Although there’s a greater awareness of robotic automation, cognitive automation is still very new compared to other technologies that are much more embedded in our daily lives.
I talk with some companies about their plans for automation, and even now their plans for robotic automation are based on potentially implementing it in a few years to come. As I listen I’m thinking, “I can’t believe you’re still sitting on the fence.” They truly consider those companies that have implemented robotic automation to be very pioneering. And there are more organizations using robotic automation technology, with more planning to use it. So we’re seeing what I’d describe as an early surge.
Behind this surge is a growing demand for efficiencies which will enable organizations to be more competitive. That’s the motivation at the core of the surge we’re seeing, and the two most promising technologies for this greater competitiveness are robotic automation and cognitive automation.
That said, I see the awareness of cognitive automation being on a wave about eighteen months behind robotic automation. Robotic automation customers seem to have greater awareness and acceptance of technology and so many more companies are using it. They’re the ones who will adopt cognitive automation as the next step.
It’s easier for those companies to say, “Why don’t we take our first step into this world of intelligent automation by doing something robotic?” And then, once they have trust and confidence, they’re ready to start talking to us about their more complex processes involving unstructured content, the ones that will need cognitive automation.
UiPath: Within the past year Capgemini has entered into technology partnerships with both Celaton and UiPath. From your perspective, how related are these events and what do they say about the relationship between Celaton technology and RPA technology?
Andrew Anderson: I think it’s a great endorsement for both UiPath and Celaton that a world leader like Capgemini has chosen our two technologies.
Over the past eighteen months there has been much more use of the term Intelligent Automation by various analysts in this sector. I think it’s a very appropriate term because it describes the outcome of combining robotic and cognitive automation – which is a more intelligent solution. Companies like Capgemini have proven themselves to be very forward-thinking by partnering with two technology providers with the right robotic automation and cognitive capabilities.
What I see Capgemini doing by partnering with Celaton and UiPath is consolidating these technologies to deliver a customer solution which will effectively trump their competitors. While their solution currently features robotic automation and cognitive automation, it will probably encompass other technologies such as content analytics in the future.
Capgemini assessed other providers, looking for the best technology to address the challenges of their customers, I’m very proud that Capgemini selected us; it’s pretty easy to let your ego enjoy feeling that we’re the best thing since sliced bread, but at the end of the day we’re just a tool that enables those challenges to be solved.
However, to have someone like Capgemini, one of the biggest companies in their industry, recognize our technologies as tools that can be applied to add customer value – that’s absolutely exciting. I’d like to think they picked us because we’re credible, we have many real case studies and they have confidence that we can achieve what we say we can and will add competitive value to their organization.
This brings to mind a good example of how we’re working together with Capgemini to deliver intelligent automation. With one of their customers our technology is dealing with sales orders and accounts payable. It has to recognise different document types, find key data buried in the content, then extract it into a structured data format. Now, after Celaton has extracted and structured it, we give it to a UiPath robot which then feeds into the organization’s ERP systems.
For Celaton, the beauty of this cognitive and robotic collaboration is that it makes for a more rapid implementation - all we have to do is hand over the structured data to a UiPath robot. It’s in the same structure, every time and so we don’t have to be concerned about the format of output data or integrating with complex business systems. UiPath does what it does so well with ERP and enters the data into those systems. The two of us, working together, makes the entire process go so much easier.
There are other companies who also partner with UiPath and together we provide intelligent automation solutions for their customers. You’ve come across Genfour in the U.K. and the situation is exactly the same. Genfour doesn’t create such a big story because they’re nowhere near the size and scale of Capgemini, but they’re expert and agile and they were working with UiPath and Celaton long before Capgemini.
UiPath: Celaton has a software-as-a-service model (SaaS) while the RPA industry is dominated by the software license model. What makes SaaS such a good fit for Celaton and do you foresee RPA providers moving in that direction?
Andrew Anderson: SaaS is perfect for Celaton because we are focused on the outcome we achieve for our customer - and that includes the rapid return on investment. I realise that software as a service is not everyone’s cup of tea, although it is fast becoming the standard as more organisations adopt it. It works for us because we can charge on a transactional basis and we can also take responsibility for the infrastructure that’s required to achieve the outcome.
Some partners are starting to provide RPA as a service, so it won’t be long before it becomes the standard approach. To me it’s a much simpler approach, but we still talk with some companies who say, “the solution has to be on our premises or it has to be on our side of the firewall.” Of course we can address this issue by installing our software in other data centres, but inSTREAM is not designed to be installed on a single PC in the corner of an office.
My background is a factor in this approach. I sold my previous company to a very early pioneer in the SaaS world, but at the time it wasn’t called SaaS, it was called ASP - application service provider. Those were very early days in an emerging market. During the eighteen months I was with them I could see - and I learned - the power of the SaaS model; because when you talk with a potential customer you are talking about the outcomes to be achieved for a monthly fee, not capital intensive technology.
At Celaton everything is designed around delivering our technology from ours or our partner’s data centers. So it’s all about the service levels, and the thing our customers find very attractive is that there’s no capital outlay; they simply pay for the outcome on a transactional basis.
The challenge we have is that SaaS is still viewed by some with skepticism. Some people like to have the software installed on their premises, under their control. But I think that’s changing rapidly now. I think SaaS is becoming much, much, more acceptable, certainly more recently and as a result of more high profile organisation choosing that route.
The question we’re hearing more and more is, “Can you give us assurance with regards to security and resilience?”; because some of the content we’re handling is very, very, sensitive and no unauthorized access can be allowed. As a result, there are organizations that cannot or will not adopt the SaaS approach, but for many others it delivers massive benefits.
Behind the scenes, our cognitive automation software incorporates machine learning. And whilst that runs in a dark room within a data center, it requires an awful lot of horsepower to make things happen. So we adopted the SaaS model because we can take on the problem of delivering the horsepower needed for customers to meet their SLAs.
In the early days, sometimes we got it wrong and we did some deals where we didn’t make any money. But those deals also gave us essential credibility. Today we serve some of the UK’s most famous brands and automate over one hundred and fifty critical services. All these are served from our two data centers in London.
The challenge we have, of course, is that’s it’s all very well for European companies - but when you do business with North American companies, they want their data to remain on North American territory. To solve this issue, we’ll be implementing our technology in new geographical locations.
Andrew Anderson: These companies are all creating intelligent applications but artificial intelligence is a huge subject that grabs headlines, stirs debate, and also creates confusion and misconceptions. That said, AI has also attracted some negative publicity that doesn’t help us because of our associations with it.
I think this is to do with people’s lack of understanding. In reality we’re all in a very narrow part of AI, one that’s called cognitive learning technology or machine learning – we’re not a threat to the human race!
It’s probably helpful to describe these products as being AI-enabled. But it’s also worth mentioning that there’s a lot of smart technologies around which aren’t artificially intelligent - despite the fact those companies claim they are. AI may be used in a headline to create awareness - but if it doesn’t learn, it’s not artificially intelligent, and trust me, it’s really as simple as that.
You also have to realise this industry sometimes gets ahead of itself and requires a reality check every now and again. That reality check comes in the form of customer case studies, because they really tell you how these technologies are being used and the value they create. The market for AI-enabled products is still in its infancy; there’s a lack of real case studies, so each new case study supports the growth of this industry.
You see a lot of technologies that appear smart, but they’re configured to deal with every possible permutation. And there’s a big difference between configuring a technology to do something and a software actually learning how to do it. It’s the difference between teaching and learning; they are not the same.
You can apply machine learning in many ways. Some of the companies you mentioned have machine learning and they apply it to different things such as content analytics.
With Amelia, IPSoft’s venture with Accenture, they’re trying replace the human presence and the model relies on pretending to the customer they’re actually dealing with a person. We’re focused on offline communication such as email, fax, post, paper and social media. We’re definitely not trying to fool consumers into believing they’re talking to a human. In reality, we’re simply a tool that is making people more productive.
Other companies like WorkFusion focus on process automation and claim to have both robotic and cognitive capabilities. In trying to differentiate ourselves at this level the conversation needs to be about learning, not just if it learns but how it learns. These applications can be configured or taught how to handle content and what decisions to make. However, that’s not the same as learning. It’s a bit like teaching a child how to brush their teeth – you tell them what to do, but that doesn’t mean they have learned.
So you can spend a lot of time configuring your software to try and anticipate all the outcomes you think it might encounter, or you could let it learn – and have it become increasingly confident.
Who’s to say one is better than the other? Essentially both are examples of learning. In one example you configure the software for every imaginable outcome, whereas with Celaton’s technology, you start with a blank sheet of paper. It’s never seen a transaction before, so the first one it sees it handles with very low confidence, looks to a person to make a decision and then learns from that interaction.
Going forward, Celaton technology applies artificial intelligence to learn from every interaction. As it becomes more confident, it starts suggesting the correct action. When the suggestions are consistently correct, the next step is to say, “Listen, you’re getting it right every time – why do you keep asking me?” It’s a bit like any new member of staff, as you task them with some work they will ask questions. With every answer they become more confident, and then they start making suggestions – until you finally say, “You’re always right, just get on with it, I trust you.”
When considering artificial intelligence (AI), forget the word artificial and just focus on intelligence- because it’s a relative term. Think back; twenty years ago your iPhone would have been considered artificial intelligence, but today it’s the new normal and everyone’s got one. In fact, people are likely to say, “What’s special about it?”
So AI is always going to be perceived to be about tomorrow’s technology. It’s an area to look forward to and discuss, one you might write about and describe all its potential. Just keep in mind that it may not be relevant today. But, when it does come, it’s certain to become the new normal – and then you’ll look forward to the next new thing.
This is why the topic of artificial intelligence will always be relative as well. In reality, we’re surrounded by artificial intelligence. There are so many things we interact with every day that are artificially intelligent, but we don’t think anything of it, it’s just normal. So we find ourselves in the position of talking a lot about the potential of artificial intelligence even though we’re surrounded by it.
There is much written about the world of AI; there are summits and conferences that attract large audiences. I’ve been to some at which artificial intelligence experts stand on a stage and talk about what they’re doing in this field, though not all of them have a live product. Now, does it have relevance for what the world needs today? No, but it’s really exciting and this guy’s very clever and engaging. And I think to myself, pray you don’t run out of money before the world needs your technology.
UiPath: Your point about artificial intelligence - how it will always be the new normal once it becomes part of our lives – is a good segue into this final ‘crystal ball’ question: what do think will be considered the new automation ‘normal’ in 2020?
Andrew Anderson: That is a long way into the future and so it’s a difficult prediction to make but easy to imagine.
The rate at which technology is emerging is getting faster. Whist there are some incredible inventions and innovations emerging, the limiting factor will be our own ability to embrace them and find value in them. So the limiting factor is the human being. We’re always trying to develop new capabilities that enable our customers to deliver better service faster with fewer people, but that’s not unique to Celaton. I feel sure that the performance of everything we are amazed at today will be accelerated and built into any device that we care to use.
We’ll certainly extend our capabilities from offline to real-time content. For example, we’re having this conversation now, and you’re recording it. I anticipate our technology will be able to take our conversation, while its being recorded, and understand what it’s about. You can imagine technology handling the translation of content in real-time to the point where you don’t have to worry about learning a new language.
Humans are the limiting factor but they are also the most relevant part of this question. I’m frequently asked about where our technology is headed. I believe, as I always have, the key to learning the answer to this is listening to customers and discovering their challenges. That’s where it all starts.
One thing I do see happening in the automation world is a convergence of technologies. Take robotic automation, cognitive automation, content analytics – all of which you could describe as being part of that convergence.
I see them converging into solutions that will address all these areas for the customer. And I think - certainly in our world - the consumer will be able to communicate in any way they want. And they will be able to get real-time feedback and answers, as a result of the speed and accuracy with which technology like ours can help organisations deal with that information.
I hope we’ll see AI-enabled devices emerging that add real value - and see fewer examples of hype and scare stories that undermine the great work that many people are doing in this cause.
UiPath: Andrew, this has been an illuminating and thought-provoking conversation for me, and I’m certain it will be the same for our audience. Thank for being so generous with your time.
This is a companion discussion topic for the original entry at https://www.uipath.com/blog/rpa/uipath-rpa-innovator-interview-series-andrew-anderson-of-celaton