Impossible Things with David Terrar S1 E17
DAVID TERRAR [00:01:07] Hi, and welcome to Episode 17 of Impossible Things for David Terrar. And before I introduce my guest, let me explain that this is called Impossible Things, partly because of AC Clarke’s third law, that any technology that sufficiently advanced seems just like magic and partly beat when the white queen in Alice Through the Looking Glass took it up during six impossible things before breakfast. So we’re into the possible stuff when we’ve got some really interesting topic to talk about this all around, Artificial Intelligence and automation and the modern use of data. I’m delighted to say I’ve got my copy date with me now. I first met Mike when he was at IBM and General Manager of Automation for EMEA there. Now, his Chief Revenue Officer, with AntWorks. Mike, welcome to the show.
MIKE HOBDAY [00:02:00] Well, good to be here. So, yeah. So a little bit of that. Well, let’s just get back a bit and say, you know, I recall when we met, I think you might even see me speaking on stage about “Jaguar Land Rover” or something. When I was in “protest” automation. So a year ago before that, I got three years of implementing robotic process automation with some A.I. across all industries, now as a General Manager for Europe. And you’re talking about impossible things three years ago, I thought the robotic process automation was impossible. And the idea that you could take a bit of code and that it would emulate how humans could operate that software could operate keyboards. And so you could take data and input in the backend systems without having humans do it. And that, to me, is someone who in earlier life done a lot of in a process improvement seemed to be a huge opportunity for cost reduction. And I did that for three years. And that was and then I discovered a new impossible technology because one of my observations of doing RPA was that really robotics was just moving data more quickly. So if the data was rubbish, it just got into your backend systems more quickly. And the impossible technology was when the CTO event works showed me how in what is now my cognitive machine reading solution, which is also known as intelligent document processing, which could take a 40 page document and actually extract all the information a human would want to, might be a contract and make that available for RPA to input into a back in system. And then I realised then that actually leading with RPA was the wrong thing to do. What I needed to do was really stop focussing on automating data at the front end and then deciding actually what I wanted to do, whether it was robotics or integration. And that was an epiphany for me. So that’s what AntWorks does. It reads invoices, contracts. insurance policies, completely unstructured documents and formats a data in a way which you can input it to transaction and apply decision rules or to advanced analytics. Very excited.
DAVID TERRAR [00:04:33] This is fascinating. And I guess is that what triggered you leaving IBM and…
MIKE HOBDAY [00:04:38] Well, “it was” really it was a realisation. I’ve always been somebody who made her this impossible things. I always have always work driven my career to once I’ve done some improved it works and proved I can make it happen. I’m looking for the next thing. It’s always been the way I’ve been. And like I proved robotics, I’d actually improved integration because of Barclays. That helped when I was at IBM “built in” Integration “boss”. So I’d done integration and really wants to get the data that I had previously worked with Watson Group in IBM, mostly around.. Automation round chat bots and conversational analytics and. And I’d not really seen anything beyond OCR “Optical Characteristic…” Which dealt with structured forms, you know, forms where you know exactly where the data is going to be because there’s a box for it. But I’ve never seen anything really as comprehensive that works, that could read pretty much anything that a human eye could read, even handwriting to a degree.
DAVID TERRAR [00:05:47] This is fascinating and obviously for the robotic process, automation point of view, there’s “a lot fairly” standard business processes that can automate and that could do a great job for companies. But this idea are actually opening up to unstructured data and actually making sense of that. That brings that level of automation to an awful lot of workers and styles of jobs, isn’t it?
DAVID TERRAR [00:06:10] Well, exactly, because what occurred to me when I was driving in particular RPA is the obvious place to go with robotics is to do things like reconciliation. So we’ve got human taking some data and doing some reconciliation work or moving data from one source into another because the applications are integrated in any way. Actually, when I thought about it, you know, a lot of the work that we do and do in our lives is because we don’t trust the data because someone else put it into a spreadsheet or you know or if they, you know CRM solution somewhere where you’ve got a whole load of email addresses for potential customers. But many of the aren’t even customers. And we spend a lot of time cleansing and reviewing data right? Now because actually, you know, in the same way the human brain works, A.I. is in relation to the extraction of an ingestion of data. It’s all about patterns. So if you know the way humans work is, we we we think it’s training and being smart. But all we’re doing really is recognising patterns. What sort of email addresses represent our customers vs.. And our competitors are looking at our products, for example. Right. They’re patterns that are very straightforward. You know, in, you know, invoices. Everybody’s invoice is different and everybody talks about statement date. We’ll have it in different formats. But, you know, it may all conform to a standard pattern. And what I can do is Machine Learning is learn that pattern. And deep learning means that if something comes up that’s kind of similar, then it recognised it’s up to. And so what I’m saying is, is that the technology allows you to train, you know, the A.I solution very quickly to emulate what a human is doing in that reading. And some of the ways that we often characterise it, it’s a bit like if you got a graduate joining your business, perhaps you’re in insurance, it’s a trainee underwriter. And in your underwriter does it show them here’s an insurance policy and here’s another insurance policy and here’s another insurance policy and overtime perhaps in a week. This young graduate now recognise the insurance policy, doesn’t matter which insurer or carrier it’s coming from because they’re all different. But there’s something similar about it. And I’ve been told that they’re all insurance policies. And it’s in essence, that’s what we’re training our system to do. You see that. You see the documents we’re not doing templating, as you might do for an OCR solution to recognise structured data. This is completely unstructured data, but I’m recognising that these documents conform to a consistent pattern. And once I recognise the path and I know what documents it is, then I can be trained to extract the data that you want.
DAVID TERRAR [00:09:27] This is fascinating. One of the things I’m hearing really interested by in this context is the fact that in particularly in bigger enterprises, this data problem where you’ve got systems that are integrated and data has been moved from one place to another via spreadsheets, which inherently mean they’re mistakes. You asked amount of wasted time, you know, you know, as you say, cringing and worrying about what they should be in manipulation to get it from one format to another. And this kind of this kind of software is going to be able to take away a lot of that grunt work, know that people have to do.
MIKE HOBDAY [00:10:03] It’s transformational, really. Which is which is why in this space, because you they need them. The large, larger clients that I personally spend my time talking to are looking at automating data ingestion across the whole organisation. So they’ll take the artwork’s platform as an enterprise platform. So, you know, the single platform can be trained to support the H.R. Director, recognising a CV and extracting key data, for example, or for the finance function, extracting invoices or for the legal department, reading contracts. And so the driver here is to start to. Ties. Data at the point where it enters your organisation. So the first rule, of course, is try and get your customers and your suppliers to digitise it for you. But clearly, there’s going to be a long tail of people are still sending images of documents to you. And therefore, a platform to be able to train and read that and know critically, no getting anything between 85, 95% accuracy, which, you know, may, ou know, people may be disappointed, but actually that’s a lot better than they’re getting with humans, doing repetitive reading of documents. And also, what’s great about these solutions is if they’ve got low confidence in the quality data they actually call an expert eye human to go. This is what I found. Is that right? And that’s “how…” learning. And once that’s in place. What happens is the number have the volume that is rejected gets smaller and smaller, and hence you get into the 90s. But if you have that confidence in the veracity of your data, it means downstream you’re not having the contract, the checks and controls. So you previously had to capture the mistakes. And those are exactly the processes where everyone’s been driving robotic process automation. And their processes, which really aren’t required. If you get the data right at the point, it enters your company.
DAVID TERRAR [00:12:15] Interesting. Now, there’s a quote this occurred to me from a guy called John Mercer. I think that’s.. I’ve got his name right from “Programmable..”. He said that in the 2000s you needed a website. From 2010 onwards, you needed to be mobile first. In 2020, you need to think about “APIs” and data. And this is all about creating you from “data from it”
MIKE HOBDAY [00:12:41] Yeah, absolutely. I mean, the way I characterise it, this will be my quote. I guess it’s kind of like the 20th century. And by that I mean everything up to about 2020 has all been about process automation. So and if you think about it and go back to the early part of last century, you know, you just had unstructured data. People would write manuscripts and type letters. Right. And and then we invented the P.C. and therefore we invented a spreadsheet. And then we invented ERP systems. And then we invented a bit of robotic process automation in capturing the history of the 20th century shorthand. But, you know, in that what we did was we were automating routine, but we were only doing it in a way in which humans could do more and more routine. That’s all that happened. We weren’t making anymore decisions. We were ending our time using technology to improve productivity and productivity should be banished. It’s a 20th century term about people doing stuff, not people making decisions. So as we move into 2020, and it’s been exacerbated by the recent pandemic. “Is the…” data. so we move into data automation where everyone’s looking at how do I get all of the unstructured data? You know, we all know that. We’ve all say in the last two years, more data is being created than in the history of mankind. And 98% of that is is unstructured. And what our solution. And others coming are doing it, actually structuring that data up. And once that data is structured up, we can drive an analysis into it, and not just, you know, reduce costs by reduced downstream processing, but access untapped sources of data for new revenue streams. And and so that everyone should be focussing on data not processed, because process is about linking people and systems together.
DAVID TERRAR [00:14:49] And the value in that data, it could be few internal systems. It could be a value to your customers, but it could be something different that might not be for your organisation. But maybe you could sell it to someone else.
MIKE HOBDAY [00:15:02] Well, exactly right. And I think we had a chat about this earlier. But the example of the mortgage industry and everyone knows that this has been around for a long time, that 80, 90% of valuable data in organisations is locked away and only used. And that’s because when we do an application and we might have contracts which are multi-page and gather other data, when we’ve done the deal, we only puts the high level data into our systems of record. Everything else gets locked away. And the mortgage industry is a classic point. Right. Because if you think about it, when a bank, you might have a bank that has 10, 15% of the UK property market. Well, when they do a mortgage application and take security over the property, they’ll know what the energy rating is for that property. They also are postcode, but they they know if using our solution, they can ingest all of the valuable data in all of those forms and make it available for analytics then I’m sure even though energy rating and postcode is of no value to a bank, it’s certainly valuable to someone. And that’s the kind of simple example of, you know, looking at the data, the, you know, the real value that sits in your data stores, in banks and insurance companies. And for anyone watching who has a business and they keep past documents, they will know there’s value in there. If any had to have time to go in and, you know, extract the right data quickly and at a low cost and then stick it into a spreadsheet if you must.
DAVID TERRAR [00:16:39] If you must. I mean, I love all this and I love your the idea of servicing more data and getting more value out of it. All of this has huge implications for the future of work. And that’s also “a sort of things”
MIKE HOBDAY [00:16:55] Yeah. It’s been it’s been around a little while, but the it is really important to think about this. So if you think that the process automation century was really about organising and making sure people. Did they work on time and in a in a controlled manner. Right. Very rules-based. Right. That work, that routine is removed. Right. And progressively, it’s not gonna disappear overnight. But as we capture data at the front end and put it into order, as I said earlier, less of the checking and controls and data cleansing that’s going on downstream. And if you’ve got 95% confidence in your data, you know, I’d expect to see transactions happening end to end automated without human intervention. But what it does mean is you get the you know, the new roles, new jobs will be about quality assurance. They’ll be about dealing with exceptions. And, of course, they’re going to be managing the automated environment. But above all, it’s going to move to the real roles are going to move towards customer engagement, empathy, conversation, decision making. All of that in innovation is of another term that gets bandied around about the future work. But, you know, I was talking to some clients yesterday and we were going into it a little bit later on saying, well, you know, if you look at your whole environment, your whole business and the company I was talking to have 40000 employees is what you know, if you’re on the board, what does it mean? And there was a cartoon recently where the Head of Transformation comes in and says to the CEO, I think we need to reskill all of our staff. And the CEO goes but they’ll all leave. And I said, if we don’t reskill and they may all stay… because the reality is there are going to be new skills. Now, first thing in business skills don’t go out of the door “is you” have about the ability to synthesise data, to make decisions will still be relevant. Right. And what happens is the information you need at the point you need to make it will be provided to you. Right. And you’re the one who will own the technology systems that are doing the extraction and looking them are running teams, running the queue ways “for the rest of time”. So that remains. But it won’t be about managing lots of routine done by people. The second area is technology. You know, I don’t think people need to think about it in a new way. So technologists talk about low code, no code “set not solution”. You know, you don’t need to know any algorithms to run also to build our solution. It’s a user interface. It’s highly configurable, and a business, you can configure the application to find stuff in documents. And what they don’t know and they don’t see and they don’t need to know is behind the scenes, loads of algorithms are configured to go and look for the patterns that they’ve defined. So it isn’t the case that everybody have to be, you know, a Python expert, although I would recommend that anyone starting off in a career or in early stage, a career do learn to code is a good idea. It’s just another language. And then I think the third element we won’t forget, it shouldn’t forget is that most organisations have huge, huge volumes of legacy technology, which isn’t going to disappear overnight. So legacy technologies, often people working codes which are older than us or are getting around for the next 30 years and we’re not. And that’s going to be a lot. “..which I think” draw upon is “…the idea” of data science. So the extraction of data and using “but” people often think A.I. is a very complicated thing. I like to think of it in two ways. One, using Artificial Intelligence to find stuff in documents to look at a photograph of a damaged “many” and compare it with the insurance claim in an automated way. Determine whether the two things match is a good example. Right. Or taking a financial instrument documenting in financial services and either repapering it or reconfiguring for the “labour” transition, which is going on. All of those things are about they’re not about applying A.I. as an analytical tool. As such, it’s about extraction and organising unstructured data into structured data. And then the second element is where the data scientists fit. So all of that structured data in suppressed data or in any other form. Right. Is it then available “to try it” you know, all of the heavy Artificial Intelligence and thinking it needs to go into analysing that data and finding new revenue opportunities and understanding how the organisation could operate more efficiently. So there’s a lot in there. And I think is really exciting. And I think the kind of work “we’re waiting” for which undoubtedly will enable more flexible working and working from home because, you know, you don’t have to go to the office to read a contract anymore. The data you need comes home and doesn’t consume your whole Wi-Fi. But also the nature and style of work is much more attuned to not going to sound old how younger the people today.
DAVID TERRAR [00:22:56] Millennials and Gen Z?
MIKE HOBDAY [00:22:56] “Trying to” avoid that. But you got that.
DAVID TERRAR [00:23:02] But this is interesting. So. Well, although you mentioned coding, what you’re what you’re saying is where these constellations that are often low code or no code. So it’s actually in the business analytical skills that we need more than the coding.
MIKE HOBDAY [00:23:15] Yeah, well, I think you just said exactly. That’s exactly right. Because, you know, I think the key message is we don’t all need to be data scientists. Right. Most of us couldn’t possibly be, right. So. And that’s fine. Data science, if it’s your bag and data and in advance maths is your bag, then there’s a great career. There’ll be huge demand for people with those skills. But it doesn’t mean your whole organisation for most organisations need to be consumed by data scientists. I imagine they’ll be very introverted places.
DAVID TERRAR [00:23:56] The other kind of kind of warning bell for workers. From what you’re saying, is that if I’m doing some sort of job, lots about lots of repetitive stuff, that’s most likely to be to get automated and try to use the software as more.
MIKE HOBDAY [00:24:11] Yeah, I think that’s exactly right. And so, so intensely inevitable future. And, you know, the great thing about low code, no code is which we find is that people actually doing the work today are helping build the solution of the future. They’re getting to understand the technology and that themselves is reskilling them. And, of course, you know, with these new skills, they can start to do other technology related or associated business activity within the companies that they’re working. And I think the other thing we should remember is every country in the world is racing for the A.I. digital environment. This is now every race. And, you know, for your viewers in the U.K., I’d say, well, our outlook for a recession is not that great. I think if I were sitting in government, I’m glad I’m not, “I think” it is that we should be doing everything we can to accelerate the use of A.I. digitisation, not just to allow people to work wherever they want to work, but also the organisations and therefore economies that move first that get to, you know, digitised a lower cost base, get to the value in their data and start to really leverage to that. Those economies are going to grow significantly. So jobs will grow. So we shouldn’t hold back. We should all be doing all we can to retrain, to educate people in these new tools and take the opportunity ahead.
DAVID TERRAR [00:26:04] That’s actually a great adviceor for what people should be thinking about in this “next time, when we come out of” lock down. Any other kind of advice for people as organisations come out of lockdown. And of course, an organisation like yours is very sensible. You don’t have offices…
MIKE HOBDAY [00:26:21] I mean, in my me and we worked and decided actually know at least to come to an end, I think, in May, I think so. Actually, you know, I run the Sales Team globally from my dining room. This has been my office the last six month, so actually it’s not. I don’t actually think it’s been more effective, but… So my advice, my advice, your question is advice to running businesses. Right. Think about data. Right. So that, you know, there’s a truism. I was going to talk to someone recently and said, you know, if you think about your career, you know, and if you’ve been working 30 years or 40 years, how much time was spent reading stuff? How much time was spent reading emails? How much time were spent typing e-mails in response to emails? How much time was spent listening to speeches or colleagues really only to gather data? And I’d be very surprised if someone was honest that that wouldn’t be 70%, 80% of their career. Now they’ve learnt from it and that earning allowed them to support the 20% of activity, five percent of which might have been making decisions, and 15% telling them, telling everybody else, especially their boss, about the veracity of the decision they’ve made. So. Everyone has the opportunity now to take the technologies that they’re pretty cheap. I mean, and if I talk about it, think about my our own solution, then, you know, a typical project might be three weeks or six weeks. So if you’re pointing the solutions because you’re using Machine Learning, you’re not templating every documents to try and get those yellow stations, for example, to read a form. And therefore, you know, why not? Because know, grasping this technology, you can pay for it in the new financial year. And above all, if you need to start now because you need to learn what works and doesn’t work in your organisation, don’t wait for your competitors to succeed. You need to go. Yes, actually, I’m going to be a fast follower. “Because he wen to fast, you know”.
DAVID TERRAR [00:28:45] That’s a great way to end this conversation, Mike. That’s been fantastic. Unfortunately, we’ve run out of time, but that’s a really nice way to end it. It sounds like a really exciting new world. And thanks for telling us all about it.
MIKE HOBDAY [00:28:59] Pleasure. Thank you for inviting me.
DAVID TERRAR [00:29:04] And well, that’s my day. And all about data automation and some really interesting stuff there. If you want more content like this, then follow @DT that’s me on Twitter, @DisruptiveLIVE that’s the guys that run the programme for us and look out for impossiblethings.fyi on the web. And next Wednesday will be episode 18. Not sure what that one’s going to be about, but come along and see. Thanks very much.