Indico Data releases industry-first large language model benchmark for document understanding tasks
Learn More
  Everest Group IDP
             PEAK Matrix® 2022  
Indico Named as Major Contender and Star Performer in Everest Group's PEAK Matrix® for Intelligent Document Processing (IDP)
Access the Report

Interview with Automation Center of Excellence expert Maarten Schot

Christopher M. Wells, Ph. D., Indico VP of Research and Development talks with Maarten Schot, an automation leader in the chemical industry, in episode 4 of Unstructured Unlocked. Tune in to discover how enterprise data and automation leaders are solving their most complex unstructured data challenges.

Listen to the full podcast here: Unstructured Unlocked episode 4 with Maarten Schot



Christopher Wells: Welcome to another episode of unstructured unlocked. I’m your host, Chris Wells, VP of R and D at Indico Data. And today, I’m happy to introduce my guest, Maarten Schot, an automation leader in the chemical industry. Martin, welcome to the show.

Maarten Schot: Thank you so much, and really happy to be here.

CW: Yeah, yeah. It’s good to have you. You’ve had a different experience and career path, and some of the guests we’ve had are different in all of the best ways. And I want to make sure we dig into your specialties as we go, but as we get started, could you just tell us about your journey, where you’ve been, and what’s on your mind today as you think about automation?

MS: So I used to be responsible for, like rolling out robotics in a big chemicals and paint company. We’d started implementing there, and it was usually seen as something exotic. And we worked on automating processes mainly back office processes, and it was part of the overall continuous improvement program I was responsible for, for these back office processes. And I really see that as a part of improving. So automation for us was a very, very big part of improving our processes.

CW: Interesting. so this is one of the ways in which you’re different from some of the folks that we’ve talked to talk to me about the broader sort of continuous improvement framework to help our guests maybe understand how automation might fit inside of that.

MS: So, at first it was automation was seen as like the golden egg or the silver bullet. I’m not sure about different, but so, but then we, we started to realize that it should be part of a framework where you should like, look at a process and say, okay, eliminate stuff. So we don’t want to do this anymore in the first place. And then we standardize, we simplify, and then we automate. So it’s part of an overall process instead of just saying, Hey, let’s, let’s throw some automation at a process that’s not so good. And then we sit there and wait for the savings to roll in. Yeah. because they don’t,

CW: They don’t, yeah. Turns out bad processes or bad whether robots are doing them or people.

MS: Yeah. And yes, if you don’t have your process standardized or even it have described well, you spend a lot of time automating it. And then they look at you and say, Hey, why didn’t it take 12 weeks as it said on the books? Right. Yeah.

CW: Right. Yeah. And then you’re mad at vendors, and you’re mad at technologists.

MS: Yeah. And the whole landscape comes crashing down on you. Then everybody’s looking at you

CW: Yeah, yeah, absolutely. I really like that framing and that that’s been a recurring topic on, on the podcast, which is in the last few years, as automation centers of excellence have really started to mature and kind of figure out what they are and how to operate efficiently getting to what’s the right process before you get to what’s the right automation. I think has been a really transformative idea in the space.
MS: Absolutely. And it’s not a very popular opinion because people go to these vendors and, and they, and the vendor will say, oh, well, we can just automate this in 12 weeks. But you have to do some work.

CW: Yeah. Before you have to know what this is, right? Exactly. The vendor, the vendor doesn’t right. They don’t know what you do. No.

MS: So if you know what this is, then, then you can go quickly and, and that’s the whole, so I see it as more as like an integral part of continuing continuous improvement where, where automation is one of the things in the toolbox, and then it could be very cool, and it can be very, very helpful. To give an example, we were, look, maybe I’m running ahead of your question.

CW: No, go for it. Yeah, that’s fine.

MS: No, we’re looking at automation automating our invoice process. So we’d have like couple of hundreds of inboxes where people would send us their invoice. Yep. And then there would, would be loads of folks, you know, reading those invoices and then saying, oh, this, this belongs to Singapore. And this goes to, to the Netherlands and stuff like that. So, yeah, that was one of the things where we’re looking at. And that was quite interesting to be able to start working on automating it, but was hard. It was really hard.

CW: Yeah. I want to circle back to that one and dive a little deeper on it. But before we do that, you have worked in, and run centers of excellence for automation. Talk to me, in your experience, how are those things set up from everything from how do they interact with the business to who are the right people to be on the team to, and you don’t have to tell me the names of the technologies if that’s uncomfortable, but just like, what are the core technologies that you need to have involved?

MS: So the main one where I worked, we were basically just getting out of the project phase. And then we were really starting up because of, it was a, because of, it was a project and, and the project stopped and we start being a real department. Yeah. All the knowledgeable people went away. So we basically had to start again and then the, it, people say, wow, why are you doing this? You should belong to us. My department said, well, no, you belong to us so it was a huge political issue. And, and, and not really helpful. I think, and if you look at the company’s benefits so that was kind of the landscape. So I spent a lot of time ex explaining to, to our it folks, but also to our business folks who we were, what we were doing, saying, Hey, I’m not here to hack your SAP system, and I’m not going to destroy your 20-year strategy, blah, blah, blah.
Yeah. And to the business that I cannot deliver, you know, wishes in, in like six weeks. So that was kind of a balance to strike. And it was, it, it was quite difficult. And at the end of the day here in the Netherlands, stuff depends a lot on trust. I, I couldn’t just go in and say, Hey, my boss’s boss says you have to open the door and open it nicely. They would just say rude words and not open the door in the first bed. Right. So that, that, that forming of, of this whole structure with people developing relationship and explaining what you were trying to do, that was, that was the tough bit, I think, I, I think the technology, so we were working with a couple of SAP variants. Yep. and and a very well-known automation software, that was not the issue. Yeah. It was getting people on board because they didn’t really know what we were trying to do. Yeah. And if you talk to a solution architect that has like this five year plan and, and, and you’re going basically, you’re going in with a bandaid. Yeah. they say, well, we can, we can fix this. Okay. What’s the cost? Well, it’s 5 million euros. So you have 5 million euros. No, we don’t. Okay. You know, so that was kind of the continuous discussion there. It was in hindsight, it’s quite funny, but at the moment was very frustrating.

CW: Frustrating. Yeah, absolutely. So where in the organization did you know, did the automation team end up?

MS: So we were positioned in our end of the day in the finance column.

CW: Okay. Yeah.

MS: And thathad a quite an interesting dynamic, because there are our people were also in the finance column. Okay. But they ended up at the CIO level. So, but we were part of the shared services function, the global business function kinda.

CW: Yeah. So you were embedded with the business that benefited most from the work that you were doing, I guess is how I would phrase that.

MS: Not always, because in terms of invoice processing, for example, yeah. For doing a good receipt robot the procurement folks were benefiting most because they didn’t have GLI angry suppliers.

CW: Ah,

MS: And that made it even more interesting in the British sense of the word that, that we were doing the, we were making the cost, we were creating the cost, but the real benefit of happy suppliers was not sitting with us.

CW: Huh.

MS: And, and that turned into another, you know, interesting thing and, and okay. How do you create benefit?

CW: And yeah. How do you show the value? Yeah.

MS: Yeah. And if I look at just my little part of the world, you don’t,

CW: I’ve been there, I think for all of the automation, COE leads out there, like this is something that, especially if you’re in early days, you have to be thinking carefully about like, who do I work for? Who sees the benefit? And again, you were talking about, you were sort of in a startup phase like you have to really spend a lot of time communicating well to the rest of the organization, what it is that, you know, that’s going right. And why should care?

MS: We did a round of quite big corporations saying, okay, what did you do with your RPA? Yeah. And, and the end of the day, only one of them could say we saved FTE. We saved cost. Yeah, and in the byline, they said, yeah, but we invested 300 people for three years before that. So now, but to structure the processes right before, before, so eliminate strategy, they put a lot of effort in it. So I have not come across companies that say, oh, we put RPA in, and then we saved whatever XFT

CW: Right. Yeah. We fired 20 people the next day. Right. Yeah. I think there are many good stories they get sort of made fantastic over time. It’s like the fish goes from being this big to it’s this big. Right.

MS: I was gonna say, are you a fisherman because

CW: Then yeah, exactly.

MS: You know what I mean? Then there’s a whole program behind it, which for me, makes a lot of sense.

CW: Yep.

MS: But if it’s sold to an executive, like, okay, we buy automation anywhere, we buy UI path, whatever. And, all your problems are going go away, and it’s going to wash your car. That’s not happening.

CW: That’s right. And it makes the best espresso ever. Yeah. Yeah.

MS: Yeah. So it’s been four or five years since I tried to buy an RPA platform, but I remember I was working in finance at the time, and I remember asking the question like audit and governance are really important to us. Like we’re gonna spin up all these bots, and you’ve got all these citizen developers running around, making things, how do we keep track of it? And the answer was I don’t know. I hope you do good luck. And to their credit, in the intervening time, I think especially the biggest RPA platforms have started to add features and functionality that make governance at least easier. But it is a problem. It’s not just like you have bots and they use CPU time, and you pay a license fee. You have to have a framework in people that are, that are making sure, just like you have HR for the human workers. You have to have governance and audit for the robotic workers.

MS: Yeah. So, that again was was quite a, a journey that our legal department didn’t really get what a, what a robot was. I frowned as well. I’d explain, and I’m like, okay, but if we have a, a third company, like one of the big shared service companies in the world, they develop a robot who owns the intellectual property of that. And we had to really start from scratch on that. That was really, and as a byproduct, we also had to start on scratch from, from scratch on the process that they were doing manually, like with people. So we didn’t own that IP either. So all those processes were not ours, so that was one of the things, and we really had to start, okay, what’s this thing doing?
What’s what server is it hitting? What processes is it hitting? How does this really work? And I had to explain all the time and, when people think of robots, I dunno, Arnold, Schwartzenegger taking over the world. Right. and you go like, well, that’s like a human, but one that doesn’t sleep. Yeah. Ah, okay. Okay. That was, that whole governance bit was, was also almost nonexistent. And so we had to build it. I wouldn’t say it was perfect when I left this role, bu we had something. Yeah. and we had, and the thing is we were also working with a third party. So that was also developing complicatedly.

CW: Yeah.

MS: That makes it more fun. Yeah.

CW: Yeah. More fun. You’re very optimistic. I love that. Yeah.

MS: Yeah. I changed roles, so.

CW: There you go. Yeah. You raise a good point, though, which is even if the governance wasn’t perfect, you were thinking about it, and you got started on it. And I think one of the ways that this can go wrong is that you build a bunch of bots. You deploy them, you have the process out there, and then you have like that moment where you wake up in the middle of the night and you’re like, oh no. What if the bots break? And, you know, some data gets where it doesn’t belong. And then you’re trying to, then you’re trying to do governance from scratch when there’s pressure. So even if it’s not perfect, I think it’s wise, it’s put something in place.
MS: I did maybe something stupid or wise. I don’t know. Well, we’ll see in 10 years, but I asked for an audit.

CW: Yeah.

MS: And that was not a nice thing that audit report. Wasn’t a very smiley, smiley document, but it gave me leverage to, to, you know, to bring in the, it, people to bring in the finance people and to really make something out of it, instead of just being like a pet project or hobby project, people started to realize, Hey, this is not going away. This is not somebody doing a macro and, you know, and then it’s all fine. Yeah. This circles back to that whole citizen development thing. But maybe we’ll talk about that later.

CW: Yeah. We’ll make sure to circle back to that. I wanted to close the loop on the structure of the COE. So we talked about where it sits in the organization. We talked about the tech stack and some of the, you know, some of the bigger picture, things like governance the, the people on the team, what roles, what skills did you find to be really important? And if, if anything was missing, what was missing?

MS: So the team consisted of me being in the Netherlands as like an intermediary between process people. And we had a team in, in India that was developing things, and they had real, like hard coders, like real tech people. And they had a business development person as well there. It was a very small team. So basically, it sounds really grassroots,, and it probably is, but I meet I’d meet people or I’d talk to people and, and they say, oh, we have an issue here. And I say, well, we can help you with RPA. And then we’d set up meetings with the people in India, with the team in India to say, okay, they would go really deep into the detail, like, okay, when you say, you know, you post an invoice, what do you mean?

CW: Ah.

MS: I post an invoice. No, no, what’s the SAP trend? What’s the SAP system? What’s the transaction? And what’s it look like, but so they would go down deep on, on that if I’d say what was missing, basically this whole structure. To a process owner person. Okay. What does, yeah. What does the process look like in your head? What should it be?

CW: Okay.

MS: Because we had for example, invoice posting, we had 70 versions of that.

CW: Oh, okay.

MS: And then somebody has to say, this is the way we post invoices.

CW: Yeah.

MS: Yeah. So it can’t be me, right?

CW: That’s right. No exactly. And if the business is somehow, if the only sort of implicitly or TA tacitly, assuming that it’s you, then something has gone wrong. You’re, that’s not your, that’s not your job. No.

MS: The detail level is very important. And I found that at, at, in this instance process owners were not really interested in detail on that detail on a transactional basis. And that’s what you want.

CW: That’s right. Yeah. I mean, you’re trying to, you want to know exactly what the subject matter expert is, right? What they’re doing, like as they, as they process these transactions and to your earlier point, you don’t necessarily wanna replicate every single step of that, but you need to know the what, so you can ask why, and then say, should we keep doing this right? Yes. so would you characterize that missing function as sort of like a business analyst type of role? Or is there something, something else that you would characterize it as?

MS: At the same time, we were also building this whole what we call GPO structure. So global process owners, and then, you know, and, and, and maybe in hindsight, these people were not equipped yet. You know, that they, there was so much stuff going on. We were coming from a, like a 200 companies conglomerate towards one way of working. Wow.

CW: Wow. That’s big.

MS: Yeah. That was, and again, in hindsight, it’s fun, but it wasn’t cool at the moment. So, so you need somebody to say, this is how we post an invoice, and this is what an invoice looks like, and it has a letterhead top left, and it has our invoice statement, whatever bottom. Right. And if you don’t do that, then we will not pay you.

CW: Yeah. Okay.

MS: And we were not there. So, it’s not an excuse, but it doesn’t help.

CW: It doesn’t help. Yeah. And it’s, it’s one of the ways that it is one of the ways these things go sideways is not having all of that stuff captured. And the business thinks that the automation COE is gonna do it. And the automation COE saying, no, that’s, you have to help. And not meeting the middle,

MS: The automation COE can do it, but then you need to adjust your, expectation level of, speed. I think is the thing with this whole ESSA thinking on, on, on that, that it’s, it’s not a silver bullet. It needs to be part of a program. But there are 12 weeks, as it says in the books.

CW: Nope. Not unless your process is so simple that you probably aren’t gonna gain that much benefit from automating.

MS: And we wouldn’t need.

CW: Yeah, exactly. Yeah. Let’s I wanna circle back now that we’ve talked about the people and the team. I wanna talk a little bit about the citizen developer. I am of the firm opinion that the concept of a citizen developer has been dramatically oversold and is also a little dangerous. And so I wanted to, I wanted to get your thoughts on citizen developers in that framing for automation.

MS: Yeah. I, I haven’t figured it out yet. We had a COE, so we had like a couple of people that were doing all the robots and, and they were responsible for it and for the government’s governance and for maintaining them and for everything that went wrong. So that was our model. Yeah. we did have some vendors that would say, okay, you know, everybody can be an RPA specialist. And I’m like, okay, but what if John leaves and the robot goes wrong? Yep. so I, I, I wasn’t happy with that thing. On the other hand, I think people on the floor, on the GABA, and on the shop floor really know what’s going on. Yes. So how do we capture that knowledge? So I was kind of, I hadn’t decided yet. So I, I was leaning towards the governance part because, yeah. If people leave and back in the day, I had a guy that wrote a, macro and, then the millennium turns, and the macro stopped working. Right. So, this is old school, right?

CW: It really that’s great.

MS: But I don’t know. I think that the governance part is important. The maintenance part is important. I, I really like everybody on the shop floor to be thinking about how to improve.

CW: Yeah.

MS: And, and be able to do something about it, because I think that’s even more important that they have the, that they have the tools to, to, to improve your daily life. My concern is how do we keep this in check? How do we make sure that stuff keeps running after five years after, three years, or even next week?

CW: Right. Yeah.

MS: So it’s not a yes or no answer, sorry.

CW: No, it doesn’t have to be. And I, one of the things that you highlighted was that having the right people with the right knowledge, build the automation makes it much quicker to do now, you know, those people could leave or, you know, they could stop being interested in that role and what, or whatever it is. Right. Yeah. And then the danger is that the automation they built is good, and everyone likes it has become business critical. And that’s the place where, so I think it’s okay that you gave a yes and no answer. I think it’s actually both. And like in it’s in a lot of ways you want to have the freedom for people to start automating, especially that low-hanging fruit, but at the point where it does become business critical, everyone has to be eyes wide open and say like, okay, this now needs to fit into a framework. It needs to be resilient. We need to have a dev reach and a broad region, whatever it is. We have to eventually, you know if it really is that important, you just have to treat it in a grown up way.

MS: Yeah. I haven’t figured out how to do it yet. I mean, I am a very strong believer that people come to the shop floor or the workspace. And they use probably 40% of their skills instead of 80%, or 90%. And, and they’re the financial chairman of their local football or soccer or rugby club. Yeah. And, and now they come in here to make whatever paint or toothpaste or what have you. So I, I, I think it’s really powerful to utilize that, that skill. But again, how do you make sure it works in a couple of years and indeed andis business critical and it’s John’s macro, that’s running and John’s won the lottery. Hmm. Okay. Problem.

CW: The macro’s less interesting now.

MS: Well, yeah, exactly. Exactly. I think if you can do it with a good governance structure, it would be really, really powerful. And we had some ex examples in our company where we’d had a center that was running, it was macros, it was small automations. People were really happy about it. They were really transparent about it. And when it became big it became part of the it infrastructure.

CW: Nice. That’s, that’s a really happy path for that. That’s good. You were talking about John and running the local football club when he is not at work. I think is one of the intangible benefits of automation, is you’ve got knowledge workers, and a lot of what they’re doing doesn’t actually require knowledge. Right. It’s just kind of mechanical. So talk to me about, again, it’s less tangible but talk to me about some of the wins that you’ve seen in that regard.

MS:Well, you have a finance background, right? And, if I look at, at, at, and so do I, so if you look at the mountains, closing, you see many repetitive things going on. Are you back? So, if you look at the month end, there’s loads of maybe it’s language, but I call ’em silly processes. Right. That you, you do repetitive tasks that, that makes make no sense. And then there’s like 10% of those, those tasks that are important and which require you to have like a university level education, stuff like that. Yes. That’s. And we’ve been trying to take a lot of that. So posting journals, for example, right? We, we had, we came across business unit we’re posting journals in Mont end for like four euros per journal.

CW: Okay.

MS: Yeah, exactly. So that’s not what you want to do when you run a 700 million Euro business. So, that was part of eliminate, you know, you don’t want to do this at all, but if you really need to do it, let’s not do it or by a master science person at, at nine o’clock at night, on a Friday. Yeah. So, we ran a robot for that kind of okay. And, because we were also an internship company, you’d have all these exchange rate issues with, with this kind of stuff, you had to adjust for that. And, that was also a reran. So if, if in some countries, I mean, well, nowadays it’s almost in every country, but the rates go up and down all the time. Yeah. Quite significantly. So that was one of the robots we built, because can you, I’d like the financial people to look at things and think about things instead of just doing SAP transactions and making manual adjustments for, for nothing.

CW: Yeah, absolutely.

MS: Does that answer your question?

CW: That is one great concrete example. Two, I have a lot of friends who are accountants, and I’ve worked on trying to automate accounting processes. And I think accounting is one field that generates many of what you call silly processes. I think it’s sort of native to that work, unfortunately.

MS: Well, and it touches upon if you come in and say, I’m going to take this away. They, they will, they, they might be thinking, Hey, are you taking away? My job, which I think, in general, is in robotics a theme.

CW: It is.

MS: And, also, in continuous improvement is the thing, you’re going to fire me.

CW: Yeah. Right. No, but to your earlier point, I think there are very few horror stories like that. And I think I think RPA and automation, professionals and vendors and consultants should be speaking more about the fact that you’re not doing that for Euro journal entry at Friday, you know, nine o’clock on a Friday night. You’re doing the stuff that requires the training and expertise you’ve built, and your mental energy is not being taken by things that don’t add a lot of value. And it could just be automated away

MS: If you hire somebody that went to university and studied finance, and then you say SAP transaction B, C, D. That’s not value for money.

CW: That’s not.

MS: And, Dutch people are very keen on value for money. So that’s, and, but it’s still happening everywhere. It’s like, okay, this is just the way it is. And, and then you take out that stuff, and then you take out of all these processes and then, and then you, you require people to really think about what are you actually posting and why is this happening and why is this country always coming with this for Euro journal entry at the latest moment?

CW: Yeah. Yeah.

MS: And that, that’s the discussion I’d like to have.

CW: Absolutely. And automation, COE doing things the right way and capturing the process, the right way really creates space and gives permission to have those conversations, which is important.

MS: And that’s why I, I found it really good that we were part of the RPA was part of the continued improvement stuff. It’s not just an, it department with a little, nice gadget, but just an overall process. So that, that I found that, that really interesting and a good, good way to set it up.

CW: Yeah. By the way, we, the American people, also like value for money. That’s why McDonald’s burgers are so big. I think

MS: Fair enough.

CW: And with that little one-liner, we’ve been sort of implicitly talking about it, but the main topic we always like to cover on the podcast is unstructured data. You were talking about invoices, which are a form of more on the semi-structured side, but still documents generally. When you’re, thinking about automation, or thinking about a new, you know, sort of process documentation, when someone says unstructured data, what things come to mind for you? What does that mean?

MS: Well, maybe it’s closer to my mind. Yeah. Still, the whole invoice processing and all the documents that came into our company and probably also went out of our company, and you’d still have people delivering 40 tons. I don’t know what that means in us but like a big truck full of stuff. And they’d have, have a paper of a piece of paper and that needs to be processed somewhere. Yeah. so that, that Springs into my mind, it, it might not be the wi Wikipedia definition, but that’s, that’s, that’s what comes into and, and how do you deal with that? Because it is such a document, a bit of lading or whatever implies a lot of knowledge because you first need to know how to hold it. What’s, what’s the top, what’s the bottom?

CW: Yeah.

MS: Stupid stuff like that. Right. It’s it. Yeah. and then how, how do you teach a robot to do that? And how do you teach the operator not to fold into 70,000 Les? So, so the OCR can’t read it anymore. Yeah. And, how can we teach the supplier not to give us a piece of paper but send it preferably in a format, in some form?

CW: Yeah. Yeah. So you highlighted two of the difficulties with working with unstructured data. You highlighted two difficulties with unstructured data; one is data quality, which is an issue with all forms of data everywhere for all time. But you also highlighted that you need intelligence to understand it, you know, just the orientation of the page, right? That we sort of take it for granted, but it’s a form of intelligence. What are some of the other challenges in working with unstructured data, especially as you’re thinking about process automation?

MS: Hmm. Well, to say to your vendors or suppliers or whatever to say, okay, we want this format, but that, I think, is a sign of weakness. We should be able to deal with that

CW: Agreed.

MS: When we had an issue in China, for example, there was loads of handwritten stuff coming in, and you can’t go say to China, look, you need to do this differently. I mean, they will just, you know, so, so the stuff you can’t change, so you need to be better. Yeah. And, and then you, we, we hit like the edges of technology O on this particular case in China, we, we, we couldn’t really do stuff about it. We could border it a bit; we could say I’m just thinking about your question it’s that you can only control so much. That’s the thing. Yeah. And, and then for, for me, it came back to managing the expectation in, in, in my own company where you could say, okay, we can automate, we cannot automate anything over everything. Yeah, that was the biggest struggle. And, it sounds a bit silly, I think, but managing the expectations inside the company was sometimes a bigger, bigger thing than

CW: Oh, yeah. Then working with technology. Yeah. A hundred percent. You’re not alone, Martin in that. Thank you. I can assure you one of them, one of the things that are really exciting to me both in my role, my, my day to day role and in talking to folks like yourself that have such deep experience is that I think we’re right at the point where the technologies are, are getting good enough in the sense that, okay, we’ve had OCR for decades. Right. But OCR is just; it just tells you what letters are on the page. And only if they’re, you know, legible by the machine. Right. So, maybe it tells you where they are on the page, but what is the page? And what’s the context of the page, and why do I care about it? And how does it match my mental model of what this data is? I think there are now we’re now starting to see that there are platforms that can start to do these things. And, I think what’s really exciting to me looking at the next couple of years in the future is as professionals like yourselves start to incorporate those platforms into that automation fabric and really push it to its limits. I think we’re going to see an explosion of exciting future developments.

MS: I hope so. It’s been, it’s been sold to us for like years and years and years.

CW: It has

MS: Especially on the OCR as well, but also, okay. What are you gonna do if the OCR reads it correctly, what are you gonna do? Because is it, is it, is it an invoice, or is it, is it you’re doing those pizza books, right? That you’re reading what? Who knows

CW: That’s right. Yeah, absolutely. No, that’s great, that’s a great distinction. Yeah, go ahead.

MS: No, no.

CW: So what do you, so a lot of organizations have we talked, we talked, we’ve talked a lot about the sort of flow of unstructured data. So invoice comes in, you know, bill of lading comes in a lot of organizations, you know, we talked about finance, shared background here. They just have stacks of documentation sitting in, you know, blob stores or network filed drives or people’s computers. What do you think are some of the risks that are inherent in just having that much data untapped sitting around in an organization?

MS: I was just thinking first, do you need that data in the first place?

CW: Yeah. Great question.

MS: I haven’t figured it out. Okay. But and, and that’s, of course, that’s regulatory stuff going on. But I do remember a project in a country in Europe, and I went to six sites, and they said, okay. I said like, what do you store? Oh, this, this document, that document, this document, they all stored different documents. Why do you do that? Yeah, we have to do this because of the law. So I, I, I phone, I, I, I called the lawyer, and he says, well, that that’s just nonsense. And this was like rooms full of stuff, paper, and they didn’t need to store it. I mean, a digital copy would be okay. And that’s still storing stuff. I mean, that still costs money, but I keep going. I keep going back to this eliminated, simplified, standardized, automated model. Do you need these things in the first place? Yeah. And I think, and in a lot of cases, the answer is no. Yeah. And, if you need that, what exactly do you need? Do you need the letterhead? Do you really need the VT number or, or text number, whatever you, or do you just need the, whatever the article number, and if you do so, I mean, you can, but then I think people should look at these processes more I think the word is holistically, right? Yeah.What am I trying to achieve here?

CW: Yeah.
MS: Instead of just gathering files, be it digital, be it, be it physical, and then sending ’em off to iron mountain for 10 years to be destroyed in a very expensive way. But then you’re going across departments.

CW: Right.

MS: Who’s getting the benefit?

CW: Good point. You highlighted another element I wanted to bring up, which is it, a lot of times, the unstructured data that’s sitting around, it’s full of unknown unknowns. Like we keep it what’s in there. I was working with an organization a number of years ago, and they were interested. They had thousands of client relationships with your sort of documents that governed those relationships. And over the years, as people started to rely more on data and cloud, the sort of data governance clauses in those contracts changed and got bigger, or whatever it was. And they wanted to know like, what have we promised and to whom? And so I asked the question, well, we have some tools that could help with this. Can we put these sorts of terms and conditions in the cloud and run them through the tools? And one, I had to explain what the cloud was to our attorneys or to their attorneys, which was not, not a fun experience. And two, they ultimately said, you know what? We know so little about our own data governance relationships that we’re not even sure we can put the terms and conditions in the cloud. And I was like, I’m so sorry. I can’t help you.

MS: Wow.

CW: You just don’t know what’s in there. Right. So it’s, it can be very scary.

MS: But okay. You triggered me with that one. I mean, yeah. One of my previous roles was with the customer service department, and we would, we would go to people that bought our, our stuff and they made a service ticket, and they write on it with a pen

CW: Yeah.

MS: To the younger audience, that’s a writing device.

CW: Yeah. People used to scratch on paper with sticks. Exactly.

MS: And that would be like, okay, this is wrong. We need to fix a component B, C, D.

CW: Yep.

MS: I don’t see it getting more unstructured than that. Right.

CW: Yeah, absolutely.

MS: That would greatly impact our quality department, guarantee costs, and everything. And somebody had to interpret that.

CW: Yeah.

MS: And, as long as we keep doing that kind of stuff, I dunno.

CW: Oh yeah. And maybe someday we’re not doing that type of stuff, but the next few decades, I don’t, you know, paper and pen going all the way.

MS: Well,

CW: Yeah. And even then maybe it, you know, maybe it’s not actual paper, but it’s gonna be on a tablet, and it’s still an image. Right. Someone’s still actually working. Yeah.

MS: No. Okay. So we were working on tablets still.

CW: Okay. Yeah.

MS: But, still.

CW: But still, it takes intelligence to interpret it. Right.

MS: And then Cindy was interpreting it all day and, but then she got sick and then some, and Wendy came in and then, you know, all sorts of stuff happened.

CW: Yeah. That we’re coming up on the top of the hour. Yeah.

MS: I think that’s, that’s as unstructured as it is as

CW: It kinda isn’t of isn’t structured as it is. We’re coming up on the top of the hour, and I wanted to highlight one element: unstructured requires intelligence. And also, you pointed out the fact that even is in something that seems as simple as invoices you might have. You were talking about how something that is seemingly simple as an invoice process might really be 70 different sub-processes. Right? Yeah. And the thing about unstructured processes and processes that are driven by unstructured data is that that interpretation is not always the same, depending on who looks at that piece of paper. And I, I think that’s a really good point that you made. And I just wanted to highlight that for the audience.

MS: And I’ve seen that in manufacturing as well. Right. So if we had recipes for Bob, just to paint and, and the recipe said mixed to a yogurt-like consistency.

CW: Okay.

MS: And, but the guy that did it knew what was meant with, so he didn’t really look at the recipe, but then I know, well, you, there’s loads of, of types of yogurt. Right. So is it thick? Is it thin? Is it that flakes in it, whatever. Yeah. So when that process got transferred to a different site consistency of the paint wasn’t cool anymore. It wasn’t good anymore. Yeah. And even in, like that kind of manufacturing processes, there’s so much knowledge in people’s heads that you really have to be very careful before you say, okay, this is how we do stuff around here.

CW: Yep, exactly. Yeah. And that’s one of the places where I think the automation vendors have gone wrong in the last few years is like promising folks straight through processing for complicated documents. And invoices tend to be pretty complicated in that, like, you don’t want that to be straight through. There are important pieces of data on there that a human should have eyes on. Otherwise, you end up with paint that’s the wrong consistency.

MS: I mean, if everything is fine, then you’d have a perfect RPA process, but you need to fix what’s coming in. And of course, it’s not really their business of, of the vendors kinda is, but, you know, it’s, if you have a, a really nice process, then, you know, you can do stuff with RPA. That’s magical, but you need to fix the process as well, and you need to invest. And that’s, that’s quite a costly process. Yes. And, if you don’t do that, then don’t expect miracles from the RPA folks. That’s, that’s my kinda takeaway

CW: It’s technology, not magic, right? Yeah.

MS: Yeah.

CW: Good. Well, as we close out here, I want to give you a chance to speak to your colleagues in the automation COEs in the world. What’s your number one best piece of advice for folks? And let’s do it in two ways. One early-stage automation, COE, just getting started, and then mature COE, trying to figure out what the next thing to do is. What pieces of advice would you give to those folks?

MS: Yeah, you’re getting me in teacher mode now. Let’s let’s go. No, I would combine it. I mean, it’s not one super magic bullet. It’s, it’s, it’s part of a program. And, if you don’t see it as a part of a program, you’re not going to succeed because people will expect savings of you that, you will not be able to deliver. So if you, but if you see it as a part of the business and part of evaluating your processes, robotics can be really, really powerful if you think, oh, we’re just gonna throw in these guys, and after six weeks, everything’s gonna be fine.

CW: Yeah.

MS: You’re going to be disappointed after six weeks.

CW: Yeah. Great advice.

MS: I dunno. Hang in there.

CW: You’ll get there. Yeah.

MS: Well, it’s cool technology. It’s but because it’s making you think about your processes instead of putting a bandaid on them, it’s making you think about your processes. And I think that’s the most powerful part.

CW: It is the most powerful part. One of the things that I’ve discovered as I’ve been interviewing folks for this podcast and just over my career consulting with people is that automation is a shockingly human endeavor. You know, it’s supposed to be like to your point, right? It’s supposed to be Arnold Schwartzenegger and the robots taking over the world, but it’s not; it’s really people figuring out what they do and how to make it better.

MS: Why do you drink coffee with your right hand? Right.

CW: That’s a great question.

MS: Yeah. And that’s kind of the level where you need to go to. Cheers. If you don’t want to go there, then don’t, don’t try it.

CW: Try it. Yeah, exactly. All right, Martin. It has been an absolute pleasure. This has been another episode of unstructured unlocked, and best of luck out there as you continue to automate the world and make people’s lives better.

MS: Thanks so much for your time. Take care.

Check out the full Unstructured Unlocked podcast on your favorite platform, including:

To learn more subscribe to our LinkedIn newsletter.

Subscribe to our LinkedIn newsletter.

Get started with Indico

1-1 Demo



Gain insights from experts in automation, data, machine learning, and digital transformation.

Unstructured Unlocked

Enterprise leaders discuss how to unlock value from unstructured data.

YouTube Channel

Check out our YouTube channel to see clips from our podcast and more.