31 July 2018

Data Ninjas meets Andrew Anderson

Posted in Industry

5CEOAndrew, amazing to meet you, could you give us an introduction to yourself?
I'm the CEO of Celaton which I co-founded with Gary Grant in 2005. I've spent the last 30 years building software companies and as a result, I've gained lots of experience in creating, buying, funding and selling start-ups. The thing I love about Software is that you can literally create magic if you bring the right people together. Whilst I'm happy to take the credit for what we have achieved, the reality is that I simply create the vision and motivate the best people I can find to execute on the strategy. I like solving problems that others can't and with the right people, we can be a real disruptor. Before I began my entrepreneurial journey in the software industry I spent seven years in the Parachute Regiment. Quite a different career path but if nothing else it taught me to value my team and be resilient.

Tell us how Celaton came about?
I think of Celaton as my 3rd start-up. It began in 2005 but its history goes all the way back to 1993 when I founded my 2nd start-up called Redrock Software. The headline is that I sold Redrock in 2002, bought it back in 2004 then renamed it Celaton. To give you some background behind the headlines, Redrock was focused on desktop messaging applications for SME's and we distributed our products all over the world. That was at the time when software was sold on a CD and shrink wrapped in a box with a big user manual. It was a great business and we IPO'd in London in 2001, then sold the business in 2002 to Netstore plc. I didn't realise it at that time but Netstore played an important

part in my journey to creating Celaton. They were the original ASP (Application Service Provider) which is now more commonly known as software as a service or SaaS. They were real pioneers in that space and acquired my business to enable them to deliver Redrock software as a service for a monthly subscription. Fantastic idea, and very early in the market, there was little appetite for SaaS at that time.

After the Redrock acquisition, I joined Netstore and went on to acquire 5 other companies that enabled us to deliver a broad range of technologies as a service. I led product development at Netstore for about 18 months which was longer than I expected. It was a change of environment for me and also a frustrating period in my life, it made me realise how much I loved building my own ventures. However, I learned a lot in a short time, primarily about the power of contract recurring revenue, software as a service and about business value in large organisations.

All those things prepared me for creating Celaton, but the key driver was the realisation that despite all the technology we had access too, we could not solve the problem of dealing with unstructured data. And that became the mission, to build a company with a platform that we could deliver as a service to solve the problem of dealing with unstructured data.

In the process of leaving Netstore, I met up again with a great friend of mine, Gary Grant, who was building his own startup called DG Tech and I agreed to join him and invest in that business. At the same time, I had the opportunity to buy back Redrock software. So, faced with the opportunity to buy back Redrock or invest in DG Tech, I chose both! I acquired Redrock back from Netstore through an MBO and we acquired DG Tech - relocated both organisations back to Redrocks previous headquarters in Milton Keynes and renamed the combined operations to Celaton.

And the name Celaton, where did that come from?
It's a little-known story, but when I was working with Netstore I had a daily commute of about 150 miles which gave me plenty of time to think about what I was going to do for my next venture. Whatever that venture was it needed a name, if, for no other reason, I wanted to secure domain names for it.

During those long commutes, the name Celaton came into my thoughts from when I was a boy. My mother was, and still is, very health conscious and she took food nutrition supplements and vitamins every day. Most of the tablets came in small brown glass jars with cotton wool stoppers but one tablet came in a rather innovative 'Smint' like container that I enjoyed dispensing. That tablet was called Celaton. My mother always said that they made her feel great and so that seemed like a great name for my new venture. I spent months getting around to checking if the domain was available, fortunately for me, it was, and I registered it.

When I brought the DGTech and Redrock organisations together we needed a different name because this was a new venture with a new mission. Having the Celaton domains registered made the choice easy, so that's why our company is called Celaton.

How is Celaton utilising Machine Learning/AI?
In many different ways. One of the main reasons I bought back Redrock was the great commercial class development team we had created. It was essential that I got to bring that team with me to Celaton, they are responsible for developing the technology we have today. We initially thought it would take about a year to create a platform we could go to market with. In reality, it took about 6 years and it was all self-funded, it was a high-risk approach.

We focused initially on handling high volumes of semi-structured, machine-generated data such as invoices. In 2011 we got to show our solution to analysts and they saw how our software recognises and extracts key data by monitoring human intervention. It was then that they remarked: “So your software is learning, it's artificial intelligence!"  And that's exactly what it was doing and that's when artificial intelligence came into our vocabulary for the first time.

Our inSTREAM software applies many different techniques that are considered artificially intelligent, it all depends on the task. Our AI is not a single general-purpose capability but a collection of specialist AI’s designed to carry out individual tasks. We call them 'skills' because they're a bit like the unique capabilities you find in different people. When you bring those skills together you can streamline and automate entire processes.

It is really interesting, at Data Ninjas we meet a lot of incredible companies who were born out of the new technologies available, or that were perhaps inspired by the AI boom if you like, but Celaton seem to have been utilising this tech prior to it becoming normal to do so! How has the AI landscape changed for you in that time?
In our first five years, we weren’t using terms like artificial intelligence, we simply focussed on solving customer problems. When we started using the term AI we put ourselves into another category, one that created interest but not necessarily new business. The pace of change has certainly accelerated and there is more noise. Now, even the most basic software applications can be labelled with the term AI. They may be smart but they're not artificially intelligent.

Regardless of the technology, the most important thing is to focus on and describe the problem you're trying to solve. If you can't articulate the problem you solve, you are unlikely to sell a solution. The problem we solve is one of productivity.

Fundamentality, people take a long time to do things, and they make mistakes which is not only very costly but limits how much an organisation can achieve growth and scale. The companies we work with receive lots of unstructured data which is very labour intensive to process.

They would otherwise have to deal with it themselves by increasing their workforce or outsource it to a third party, but that often shifts the problem without solving it. So, our software consumes this plethora of information that flows in every day and streamlines the processing and reduces manual effort required. In simple terms, we take in unstructured data and output structured data which is what line of business systems need.

Whilst we started our journey by solving the processing of invoices, inSTREAM is far more capable now and can handle much more unstructured, human-generated data such as enquiries, complaints and claims. It's a much more complex environment and one in which our customers rely on our software even more so it’s essential to get it right because if you mess up a complaint or enquiry, it can be very costly to our customer's reputation!

There are a lot of companies working on artificial intelligence at the moment, what is special about your approach.
We focus on the problem we solve and not Ai for the sake of it. Our software learns through the natural consequence of processing every transaction which means customers don't need to 'fix' the vast silos of data they have accumulated over many years before they get started. More importantly, inSTREAM learns from the actions and decisions that ordinary human operators make when they interact with the process which means inSTREAM learns on the job. It's more accurate to say that we are augmenting human intelligence.

In our world the pattern in the data is important. And that’s what inSTREAM is looking to understand when it is trying to make a judgement about meaning and intent or recognise and extract key data. In handling invoices and sales orders our competitors rely on templates or key words. That's fine if the variation in content is low but even when dealing with invoices things change constantly and so templates or key words cannot cope, and a new template or key word configuration needs to be configured by a technically qualified person.

Whereas inSTREAM understands the textual pattern in the content, it can learn to cope with infinite variety and therefore can deal with a hugely complex environment. It doesn't need reconfiguring when a new supplier is added or the expected invoice content changes, it simply refers the 'exception' to a human operator who can use 'point and click' to answer relevant questions and select the key data on their screen. What the human is doing, is teaching the software which improves inSTREAM's confidence and accuracy and reduces the dependency on human operators to deal with exceptions.

So, humans are still important with your solution?
Absolutely. Much of the noise around AI gives rise to the perception that humans are surplus to requirements, but nothing could be further from the truth. 'Human in the loop' is such an important part of our solution. It's the human that teaches the software and will continue to teach the software with every transaction. It's a process of continuous optimisation.

I hear 3 words, used in our industry, that distort reality and are often used irresponsibly. These are: “Artificial Intelligence, Automation, and unstructured”. This word ‘automation’ often sets the expectation that you don’t need a human in the process. But that is not the case. It is the same with AI, people often use AI to mean “artificial intelligence”, when actually they are talking about “augmented intelligence”. Because we are augmenting human intelligence, we improve their productivity. That is the reality. So, if we try and apply Artificial Intelligence to replace the human, 2 things will happen. 1, we will fail. And 2 we will meet resistance in companies.

Because despite what the media says, companies value their workforce. They don’t want to let them go just because a new technology comes along, but they do want to use that technology to grow. And that's also one of the reasons we work with large, ambitious companies who want to deliver better service and achieve competitive advantage. They recognise that their workforce is important.

Given you weren’t using AI, or Machine Learning in your vocabulary until around 2011, do you think that if you had, your customers would have been ready for it? Or understood the value of it? My concern now would be that companies want to be involved in AI because it is ‘fashionable’. Would you agree with that, would it have put customers off?
I don't believe that using those terms in the early days would have gained us any advantage. We may have created some noise, but customers are more interested in hearing about the problem we solve than what we call our technology. It takes more than one company to create awareness and demand for this type of technology and that has taken many years to emerge. What we are seeing now is that there is 'sanity in numbers' - the more organisations that use this technology, the more it creates awareness and eventually we reach a point in the lifecycle when organisations are aware of it and understand something of the benefit it can help them achieve. Eventually, it will be considered irresponsible for organisations not to use this type of technology within their operations.

Until recently, and I mean a few years ago, it was a continuous process of education to potential customers and analysts. Now there is lots of noise and the facts are being lost amongst the hype. Look around the industry, and every tech company is an AI company, and if they are not, they risk not being included in the conversation. The reality is that AI is the reason why we can solve problems now that previously couldn’t be solved. That may get you into a conversation, but you still must explain the problem you solve.

So, would you suggest then that companies now understand the value that sits within their data? Are you having to do much in the way of education?
Organisations know that they have accumulated huge quantities of data over many years, but they haven't given much thought to how that data will be utilised because they didn't prepare for AI. Only now, when AI technology is emerging do they realise that the data is not useful. It may be data but it’s not learnable data.
There are two things that we have to deal with and they are education and expectation. We talk to people every day who have different types of expectations. Some are sceptical, some have expectations beyond reality. It is much easier to deal with the sceptic because we can educate them, and they will become a believer, it does depend on what they've read, or heard. But what we try and do is educate everyone we talk to.

It's becoming less of an education now because of a growing awareness but in the early years we had to educate everyone about reality, terminology, capability and the industry, but now people are much more aware, and not only understanding but more accepting of it. There are fewer risks and dangers too and that's because there are more case studies around.

Our focus from day one has been to engage with well-known brands and create a great case study that helps us to educate the industry and secure the next customer. We feel like we are helping to advance the industry through that education and those case studies are essential. But we are still climbing that wave of awareness, it is a long journey.

Are we doing enough as an industry to promote the good stuff? And thus, helping to educate. I think we are, but we also have to realise that this means dealing with the hype. There is still a lack of case studies that help people understand what it means in real terms. The industry will always get ahead of itself, the hype will create a wave that people and organisations will ride to exploit it for their commercial gain. Whenever I speak at conferences I always refer to case studies and try to avoid using the term “artificial intelligence”.

Technology advancement comes in waves but with AI there is a lot of hype. The reality is that we've been using AI in our everyday lives for many years. We don't think of it as AI because it's emerging as step by step improvements. What was once innovative, and a novelty soon becomes the new normal. Things like autonomous cars. Some voices are opposed to autonomous cars, but we are using elements of it every day, but they're called driving aids. Whilst some are resistant to change there are always examples where people could hugely benefit from it, which might include people with disabilities. They might otherwise be unable to drive, but this technology has the potential to drive for them. So that helps bring a use case to life. And so, it is talking about the reality, not the technology that helps to educate people.

What is the best thing about working in the AI/Machine Learning sector?
Contrary to what doomsayers say about the impact on the workforce I think the most exciting thing is the potential it creates for people and work. Not only is it exciting to work in an industry where so much attention is focused, this all requires people to develop skills for roles that didn't previously exist.

I may express frustration about the hype and noise but it's generating so much awareness, you can't help but be infected by it, not only in what's happening now but its potential to improve the world we live in. This new capability has the potential to impact so many industries. That creates opportunities for new ventures, investment and jobs. Regardless of what we call this new capability, it enables us to solve more problems and deliver efficiencies that were previously out of reach.

What does 2018 look like for Celaton?
We've spent many years getting the technology and commercial model right, building the team and developing capability on our inSTREAM platform. The focus now is growth and scale. That means we're making each of our solutions more productised so that they can be deployed faster and via an international partner network. We've seen this growth rate accelerate over the last 18 months and that rate is only going to increase.

If one looks into Celaton, one thing they will notice is the number of awards you have won for your product. What do you put that down to?
It's always a great feeling to win awards. I think they add something to our credibility, which is good for business but also acknowledge the hard work of our team. I put that down to three things, Reputation, Capability and People. People are at the heart of everything we do, they create the capability which enables us to build a reputation. They really are critical. However, when we're going for an award we simply put the customer at the front. It's always about what we do for them that counts. As I keep saying, you need to get good at describing the problem you solve and not the technology.