Get a demo
Select the type of demo you'd like to experience with Indico Data.
Select the type of demo you'd like to experience with Indico Data.
Watch Christopher M. Wells, Ph. D., Indico VP of Research and Development, and Meno Hellis, Business Transformation & Data Strategy Consultant (Former Companies: Manulife, The Citco Group of Companies, IRESS Market Technology), in episode 1 of Unstructured Unlocked. Tune in to discover how enterprise data and automation leaders are solving their most complex unstructured data challenges.
Chris Wells: Welcome to this episode of Unstructured Unlocked. I’m your host Chris Wells, VP of Research and Development at Indico Data, and today I’m really excited to be joined by Meno Hellis. Meno, how are you?
Meno Hellis: I’m well thanks for having me, Chris. I’m really excited to share some of my experiences and thoughts on data and unstructured data.
CW: Great, thanks for taking out the time. Let’s start with a few, who are you, what do you work on, types of questions. Does that sound good?
MH: Sure, so I’m a digital transformation expert. I started my career with a fintech company called RS Market Technology, which is kind of the equivalent of Bloomberg in Australia and New Zealand. But then moved on to spend the biggest chunk of my career in fund administration within the domain of data management, digital transformation, business process management, and then it moved into manual life to work on a similar capacity focusing on digitalizations for operations. And I continue to work as an advisor and a consultant to financial services that are trying to really get around some of the data challenges, specifically document management and business process management.
CW: Great so you’ve seen the data analytics and automation space from all sides it sounds like.
MH: Right, probably more than I would like to but, yeah.
CW: That’s good. I want to dig into a lot of that of course but let’s start really high level and talk to me a little bit about what you’re seeing in terms of, what I would call the mandate, for the automation center of excellence in the industry today.
MH: So I mean I’m glad you mentioned center of excellence because that is a step zero when it comes to any kind of automation because I think in my experience I’ve seen various arrangements, but the simplest and most successful is to start with a center of excellence. Bringing together a group of cross-functional talent and giving them a sense of purpose, and rallying them, and as a driving force to do something and solve a problem within the organization. Such as for example, an automation, process automation, digital transformation. And it has to be kind of layered within an overall strategy as far as, if you think about it digital transformation, business architecture, and then you have this capability that’s called the center presence of the process automation. So the key step, or step zero is to create that center of excellence.
CW: Yep, you mentioned digital transformation which is one of those buzzwords. And it’s been a buzzword for, I don’t know, probably about a decade now at least. Where do you think, especially in the financial services vertical, where do you think we are on that industry journey to digital transformation?
MH: So I think there’s been a lot of progress that has been made over the last 10 years. But then I think what’s happening, there’s a lot more challenges and a lot more scope is being defined. Especially in space where I am, wealth management, both retail and institutional. A lot of things are being disrupted, there’s changes in demographics, there’s changes in technology, there’s changes in customer needs, and that’s opening up more challenges and opportunities for digital transformation. But that’s it, I think in parallel the advance of technology has done well, and therefore there’s tremendous growth happening in digital transformation. But where I see it right now I think it’s a core strategy of any company and any organization that’s trying to be client centric. You have to have it as part of your corporate strategy. You have to have it as part of your operational strategy, and it has to be part of the culture for it to be really solid and be foundational.
CW: Yeah, no that makes a ton of sense. Rising interest rates and inflation would have nothing to do with any of that I guess?
MH: No, it probably will make the cost of talent, because this is an area where digital transformation, especially if you’re looking for let’s say analytics data scientists, software engineers, those are going to cost more because there’ll be wage pressure. But overall it has nothing to do with it, it seems. But obviously it does impact, because whichever program you’re running, whether it be a regulatory or digital transformation, it’s the people that make the difference, and those people are in depression now, yeah.
CW: Yeah, okay. You mentioned challenges and I definitely want to hit that but, for a moment let’s drill down to the COE a little bit more. You talked about business strategy, and then architecture, and then the COE sits under that. And that often means relationships with a lot of departments throughout the company for the COE to manage. What have you seen work really well, and work really poorly in terms of the way the COE interfaces with business units?
MH: That’s a great question. I think setting up the COE is the easy part. Maintaining it, setting up a proper governance for it, and a roadmap and clearly defining the roles and responsibilities is the tricky part. And evolving it so that it’s a continuous improvement, because at the end of the day this particular shared service, or COE, is working with emerging technologies. It’s also working with emerging problems that are being defined. We’re not trying to solve accounting problems that have been around for a long time. Where we’ve had decades to figure out the standard operating procedures. Most of the problems are being brought forward to us are problems that have not been solved before or not been solved for a long time and therefore, there needs to be a a culture and a mindset of learning so that as the COE does one project or does one process automation, there’s reflection, it can learn how to do this better next time, faster, maybe more cost effectively. So that’s an important element of that COE, because you can set up the structure, but then the structure is very static, it’s not learning, it’s not really evolving to adapt with the changing business needs. Especially, for example, in spaces that are changing rapidly like let’s say retail, media, or wealth management, or something. It’s regulatory, changing quite fast. The COE needs to evolve so that it can align its practices, processes and frameworks to continue to manage those changes.
CW: Yeah I think that’s a really good point, is that RPA has been around for, I don’t know, almost a decade now? And that’s one obviously important tool in a COE for automation, but the tools are still growing and changing and there are lots of new entrants to the market all the time.
MM: For sure, yeah. The organizational setup is important, and you bring up a very good aspect of it. I think organization, like I said, it needs to be open for learning, experimentation. It has to have the right mindset and attitude. But also from a technology, like you mentioned, there’s always a flavor of the month. It could be RPA now, it could be analytics visualization. It’s important to have a tech stack that is continuously being examined. Is this the right tech stack? Look for tech stack that’s open architecture that allows you to scale and switch if you see that there’s a different and a better technology out there that can better, better match your needs.
CW: Yeah I, a lot of the platforms out there sell themselves as, oh this is it, you just need this tool, and you said tech stack which makes me really happy. Because you really do need a tech stack for automation. There’s no one tool that does everything.
MH: 100%, yeah.
CW: Great, so looking at it from the other side now in terms of the business units, what do you see business units do wrong most often in the way that they interface with the COE?
MH: So I think a lot of the mistakes that I’ve seen and the challenge is that, they expect that this is something that can be achieved fairly quickly, and that there’s a lot of tools out there. And I think they’ve seen a lot of buzzwords and talks about technology can do a lot of things, and I’ve worked with platforms that are described as low code and no code, and they make a lot of promises as far as, well, you don’t really need technology, you don’t really need software engineers this is just a point and click and drag and drop. You can get your operation balance to build end-to-end applications and or enterprise applications. and that actually, and unfortunately, has created some kind of false hopes. So the business leaders now expect that things can happen fairly quickly, and expect that things can be developed, application delivery is going to cost them a fraction of what you used to, because apparently everybody is an app developer these days right? And in my experience that is a very dangerous assumption to make, because if you want to build an enterprise application, if you want to solve for automation end-to-end, you need technology partnerships. Especially for critical elements like security, performance. There is a portion of application delivery that can, no doubt be delivered by what we call the citizen developer. Which is your operation analysts that have been upskilled to do the UI, to do maybe the business rules. But the biggest chunk of the work, it will continue to be delivered by professionally trained technical expertise. So that’s one aspect, the cost and what it takes to develop an application or process automation. The other aspect of course is especially in the space of document management and unstructured data. And you’re dealing with volumes of forms that have paper forms. It’s that the solutions out there are amazing but they need to be trained. They’re no different than your processes. Just like when you’re onboarding an employee you need to train them, and you need to summarize them, and you need to give them good examples, and bad examples so that they know what to do the next time. The technology is no different. It needs time to learn what it means to do. And I think that hasn’t gone down well. It hasn’t been explained well I should say to the business. I think people expect that from deployment day that just knows how to do everything
CW: Yeah, I’ll say for my own part at Indico, we work with unstructured data and automations and, for a long time, our marketing looked a lot like the RPA vendors’ marketing. And so part of that was because we wanted to sound familiar, but it was a blessing and a curse in that we’re not RPA and they’re very different things.
MH: Right, and then an RPA is a great example where I think a lot have been misled. In the sense that I’ve seen many organizations and operations heads where, for a long time, they measured their success when it came to RPAs, well how many bots are out there actively work? Yeah, but at the end of the day what are they doing? Are they fully utilized? If you have a bot that’s running an end of day process, or start of day process that just runs one hour at the start of the day and one hour at the end of the day, it isn’t really an efficient bot. You have to think of them as a virtual agent or a digital worker. You’re paying for it, and from my experience it’s not much less than an offshore resource. And you have to ask yourself, does it really matter if you’re running 2,000 bots, if each one of them is running at a capacity of 10% or 5%? So the utilization is more important. And similar errors or similar nuances are happening in other technologies, whether it be on document management or OCR or ICR. I think people can easily make the wrong assumptions or will use the wrong metrics. So as a business leader, you really have to know, what are you trying to accomplish? And have your objectives clear from the start and metrics clear from the start, so they know whether you have a strong business case or not.
CW: Yeah that’s a great point. I see it all the time both in business units and in COEs where there isn’t a clearly defined return on investment that they’re trying to get out of their bots or their automations, or workflows or whatever you want to call it. And so they default, like you mentioned, it’s easy to measure the number of bots in production so you default to stuff that’s easy to measure like accuracy. But accuracy doesn’t tell you how much money you’re saving or how much capacity you’ve created. It’s table stakes. You have to be accurate. So that raises, I was gonna save this hot potato for closer to the end but I’ll just ask it. Has RPA in your opinion, delivered on its promises? You mentioned the citizen developer thing, and of course there are armies of RPA developers now, has RPA delivered?
MH: I mean, I think it’s one of those questions where it really depends on where you’ve deployed your RPA and what is your tech strategy. So like you pointed out, you cannot, there aren’t many platforms that are just one place or one-stop shop where you can do everything. So if you’re deploying RPA as a part of a combination or a tech stack that you have for, let’s say to address the here and now, because that’s where RPA originally came. It really came about because there is a long time to wait and for the long term solution. So to give an example from the experiences, the problems that I’ve solved, we might be building a digital pool so that we have the first digital intake for all of our account maintenance and services. But that’s going to take a long time because once you’re dealing with external users, then you have to engage risk marketing, customer experience, and a multitude of stakeholders. And it’s a larger project that’s gonna take a long time. But then what do you do with the here and now? How do you address the people that are wallowing in pain, addressing the problems that are coming out of the current solution which is, it’s legacy, it’s manual, it’s laborious. So that’s how RPA came about. It solved that problem because it mimics the actions of human beings going into an application, keeping system clicking the buttons and so on, as a tactical solution until the strategic solution is being built. And I’ve seen RPA gone wrong where people have used RPA for a strategic solution, or they’ve used it in tactical ways where it didn’t make complete sense. Or they’ve used it as, to do everything from A to Z, and you really need to use it to more of a specific pointed problem and to address what I just described instead of just, well I can use RPA for visualization, I can use it for reporting, and I can use it for reconciliation. That doesn’t make sense. You have to have really specific purpose for what you’re using your RPA. You have to have proper governance to make sure that before you’re provisioning any bots, you really have an idea how they’re going to be utilized. Processes that will be doing. You have to have the proper risk controls and balances in place. Because I’ve been sit in situations where bots were transferring data for the bank and insurance for example. Or transferring data from Canada to the U.S. and there’s data residency regulations. So you kind of have to really be careful about all these things, and make sure that it’s deployed in the correct and in the proper way. So it has a lot of good promises. It can yield benefits for sure if used correctly.
CW: I really like the point you make about governance. It’s not a surprise that you’re making it coming out of insurance and banking. But I think that’s one of the biggest hidden costs of automation platforms, is when folks get into this and they say, oh well I’m just gonna, I’ll build a bot and it’ll do A, B, and C. Well who’s gonna make sure down the road nine months from now that it did A, B, and C the right way on this day at this time? It has to be auditable.
MH: Yeah you have to treat, especially when it comes to RPA, maybe other technology is not so much, but because the nature of our RPA the mimic of the human actions it’s just another user that you’re on boarding. So you have to go to the user account, you have to provision them what they can access, what they can’t access now they are no different at the end of the day.
CW: That’s a great point, yeah, should think about them the same way. Alright we’ve covered some really broad stuff. Let’s zoom in on you a little bit, Meno. And why don’t you tell me about your personal philosophy or process for finding the best candidates for process automation.
MH: Sure so, I think when you’re dealing with, so once you have a COE which is important, so the organizational setup is step zero. I think the talent that you bring in, the mindset, the attitude is also important. Then you go into the execution part. Which is how do we find those good opportunities that can yield, get good value. What you need to accomplish that is, more of a standardized approach and methodology that can help your organization evaluate and prioritize those opportunities. So when you have this in place, you’ll be able to quickly establish and utilize, let’s say a defined set of criteria, that you can use to determine which processes are good candidates, which for automation. And then you can evaluate those candidates to measure their potential benefits.
What I’ve used in the past is a simple, and I think keeping it simple would help for practical reasons and for efficiency purposes. To find the fit you have to recognize that not all processes are created equal. And good candidates share some common characteristics. So when you identify those opportunities for automation, typically you look for ones that require manual interaction with one or more IT systems or applications. That they are repetitive in nature, that they occur with significant frequency, for example. And that as much as possible they include logic and rule-based. So there isn’t really a lot of human judgment and maybe, although that again, the technology is advancing. But if you want to maximize the benefit you definitely want to look for something that’s logic and rule based. And that it’s prone to human error, so if there’s a lot of dexterity in that industry, you know what I mean. And if it can be performed after hours, if there’s a flexibility when it can be performed these are additional passes. And it’s time consuming to perform. So those are typically criteria that you’d look for. And when you’re qualifying the opportunity, so once you look at all these kind of sort of artifacts that you’ve collected, you can map in, well I used to have some kind of an automation heat map, process automation heat map. And we use kind of similar quadrant if you’re familiar with the Gartner four quadrants. So we have the bottom left quadrant is the lowest priority candidates. And those are ones high in complexity and low in value. Then the opposite quadrant, which is the top right quadrant, is the highest quality candidate. So those are high in value and low in complexity. So if you have, for example let’s say, an invoice processing process that is, it’s significant in terms of volume, it’s significant in terms of frequency, yet it’s valid because if you don’t attempt your invoices then you’re basically leaving money on the table. That should be in your top right quadrant because that’s one of your highest priorities. And in the other two quadrants that are the top on the on the left and the bottom on the right these are the medium priority, either they are high value but then high complexity, or low value complexity. So if you just kind of sort of collect these artifacts and then map them into the various quadrants that gives a kind of visual way to immediately figure out which ones you need to hit first.
CW: You’ve got your road map right, work from the top right and then hit the other quadrants until you get to nothing left worth automating, yeah.
MH: Yeah exactly. You don’t want to complicate it and go beyond that kind of simplicity. I’ve seen different versions of automation heat maps and you can have different criteria, it doesn’t have to be this. But as long as it’s something that you know for sure that can guarantee that you’re selecting the top opportunities first, and it’s something that’s repeatable, procedural. Then that can function right. But if you complicate it too much when you’re spending so much time to try and figure out which processes then that’s time, valuable time that’s being lost.
CW: Yeah that analysis is probably the most expensive part of the process, in getting there. One thing you mentioned, very specific I want to drill into why do you see it as an additional value that the process can run after hours?
MH: Well because it, what I meant is that it can run any time. So sometimes the process is run during the time where you have capital markets running processes and things like it can easily get in the way. And if you can run it after hours, then that means if the process runs into exceptions it can be attended by locals or offshore team in a different location. I think what I meant is flexibility when it comes, not necessarily that it can run after hours. If you, if your process must run let’s say at 1 pm, which a lot of them for example, oh 9 a.m. or 8:30 or 8 a.m. for a start a day or report for example. Then you’re going to have less flexibility, for example, than if you can run this process any time. There’s also cost savings because if all your processes, for example, have to run at eight o’clock in the morning, as an example. Well then, then you’re bot army if, let’s say you’re using RPA, they’re not gonna be multi-threaded, you’re gonna have to provision another bot. whereas if you have flexibility right, then you can just use the same bot, but now you’re getting better utilization from that same bot.
CW: That’s right, so cost reduction and de-risking essentially.
MH: Exactly.
CW: Yeah that makes total sense, that’s great. This is going to sound like a job interview question, it’s not but, as you think back over your career both being hands on keyboard and then in a leadership role in automation and automation COEs, tell me about the project that was maybe, most satisfying that you’re most proud of, in terms of what you automated and what came out of it and the impact of the business.
MH: Sure, yeah. I think the, I mean the one that meant the most to me because I’ve seen the biggest impact was a workflow solution that we built for a large accounting team, well over a thousand. And we ended up yielding 120 head count reduction out of that and with much better scale than 300% scale. So this is securities industries where you had the trade life cycle and if you’re familiar with the securities industries, they really work as part of a life cycle. You have the trade capture, so this is the trade from the let’s say the broker or the counterparties. So the trade capture, the trades are captured and they go to the book of records and then there’s a pricing team that has to price those trades and positions. And then there’s a reconciliation team which is a team that is going to reconcile what the client said that they’ve traded, this is what the broker said that they’ve traded on their behalf. So matching the two different stories and looking at the exceptions to see, are these legitimate exceptions or just data anomalies. Then the accounting team, once everything is priced and reconciled, comes in and has to do the net asset evaluation before handing it off to the investor reporting, who report to the investors for those funds. It’s an orchestration of activities across what, back then was a very siloed setup, where every organization has had its own technologies, had its own processes and then when it finished its task it relied on emails and chats to let the other group know that, hey I’m done for this particular fund, I’m done for this particular group and so on. And it was very client-centered. So for example, the accounting team for this fund did all the accounting for this particular fund. So the first thing that we did was to actually reorganize the workforce and functionalize the team such that, you, to no longer be in a client centric team. Not, shouldn’t say client centric client, focus or set up team in the sense that there’s no reason for a team to do A to Z for client A. Instead, why don’t we just have a team that does all the listed instruments or all the, for example, the Morgan Stanley trades. And another team that does all the Goldman Sachs trades for all the clients. So instead of you doing A to Z for one client, now you just do everything to do with Morgan Stanley for all the clients. So that’s functionalization, and this is more of a different part of my mandate that I’ve managed, which is more of a process design, process simplification, introducing process engineering and lean, before we do any automation or transformation. Because you can leave a lot of benefits by simplifying the process before you automate. Once we’ve done that, what we did was we created an intelligent workflow, that was driven by the movement of the underlying data. Such that it detected from the systems if a specific trade or position has been priced, it would automatically let the next group in the activity that, hey your task is now up and it’s ready for you to be done. And it intelligently manages the workflow across the global team of 1,000 plus accountants and operation analysts in multiple regions. And when I say intelligently, because what it did is, it figured out who’s the next best person to work on the next task based on visibility, based on the type of data and so on. And that transformation was phenomenal. Because you no longer have people huddling in, in rooms to figure out, okay what are we gonna do today what’s your resourcing strategy capacity. And even more so the highlight of how impactful is that, whenever we’ve had typhoons in Manila we’ve had, for example, resource center offshore. It would be more of a disaster recovery we’re trying to figure out all the directions, how do we manage the workload and transfer it back to the other locations whether it be in Europe or North America. What that workflow intelligence automation solution did is that, it automatically figured out based on who’s online, how to reallocate those tasks to various members in other locations, again based on eligibility, based on skill set and so on. So all that headache and overhead management that used to be in place to figure out how to respond to specific BCP situation business continuity situations, is gone. And this is just the icing on the cake, on top of the cost reductions, the scale, and the process of improvement.
CW: That sounds like an incredible project, that must have been a lot of fun to work on.
MH: Oh yeah, I mean it was challenging. I think it was quite difficult to absorb, because I’m not a fan of boil the ocean. And this seems like you’re trying to boil the ocean. And it’s very, it impacted many stakeholders, many, many functional groups. But we managed to get it off the ground because we kind of used agile practices incrementally building piece by piece, incorporating one more function, one more function at a time. But at the end when you step back and look at it, we did kind of boil the ocean from an impact to that organization.
CW: Absolutely, no that’s incredible. I wanted to pull a couple of threads. One, I really love that you mentioned the fact that you, sort of slowed down and said are we automating the right process? And you took the opportunity to re-engineer the process before memorializing it in an army of bots and software. I see people not doing that constantly, and it’s a real missed opportunity.
MH: 100% yeah. I think out there there’s kind of a philosophical debate where there is, do you simplify it and fix it first before you automate it, or do you just go ahead and automate it? I mean I’ve met many colleagues who say, oh wow why waste time, incorporating lean and six sigma where you can just automate the hell out of it? And get over and done with it? I tend to be more procedural when it comes to this. I cannot bring myself to automate something if I have a gut feeling that some of these activities are not even necessary to begin with. I just like to follow kind of a method where I look at the process into it, simplify it, make sure that it’s relevant, make sure that all of it adds value. Then you can go ahead and automate.
CW: Yeah you’re gonna have far fewer headaches down the road managing the bots. To your point the bots are just workers. They happen to be electronic workers but they’re workers. And if you have a bad process you’re going to have a bad management experience with the bots. it’s going to happen for sure.
MH: 100%, yeah.
CW: The other thread I wanted to pull is that one thing that we see over and over again when we’ll build an unstructured data based workflow, and often these workflows still have a human in the loop somewhere. What we see over and over again is just the increase in job satisfaction for the human worker, because they’re not doing the laborious stuff as 95% of their job anymore. They’re doing error handling and decision making and more strategic thinking. How did that play out in the sort of the big workflow process that you automated, that you discussed.
MH: Yeah I mean you bring up a very important point, and I would say that this alone is actually competing with customer experience, which is employee experience. I’ve seen a lot of trends happening in COEs was, what are the drivers behind digital transformation, why are we doing this? And where it started with me is always about focused on revenue generation, customer. But more recently I think it’s now almost equal, customer experience and employee experience. Because if your job as a CEO or the head of operations is to take care of your own employees, they then in return take care of your own clients.
CW: Yeah, that’s right
MH: It kind of works that way, and the more satisfied they are, the more engaged they are, and the more that they are willing to spend more valuable time curating your clients and helping them look for opportunities, upselling and so on. So that is very crucial. And, in my experience, It does tend to increase productivity, loyalty to the company. And, I think overall the company’s output is improved. In my experience with that workflow solution, enterprise workflows solutions that we built, we’ve seen turnover rates drastically reduced. We’ve seen people that have improved their skill set. Never mind that they’re now have more time to direct it to, like you said, to more strategic decision making and high value work, but some of them are now able to leave on time to go to their son’s soccer match or daughter’s basketball match. These are also wins. How they feel about their job, that we’ve improved their life, not just their action, what they do at job. But they can actually, because a lot of them were staying over time. If you’re doing stuff manually, if you’re doing a lot of processing, you’re tired, you’re exhausted, it does impact your work-life balance. It does impact you, how you think about your career and so on. So I’ve seen that transformation happen first hand, and I can tell you that it does make a lot of difference.
CW: Yeah, that’s great to hear. I think there’s been a lot of fear in the market about the bots replacing humans. But the bots really are just replacing humans doing things that humans probably don’t need to be doing, that’s been my experience.
MH: Or can’t keep up with. Because I feel if you look at the reality is, there’s more and more data being pumped right into the systems. And we just cannot keep up. I mean, if you look at unstructured data as an example which requires a lot of work to access, never mind process, just to access. There’s just not a, not enough human resources to do that right. And it’s increasing, because previously we traditionally looked at documents only, but any kind of wealth management firm, any kind of financial service firm, is now trying to look at the 360 view of their customers. Whether it be retail institutions which includes things like email, video, text. All sorts of information needs to be captured. I just don’t see how you can have your human resources manage all that complexity.
CW: Absolutely no, no you can’t. You’re right. This is a great transition. You mentioned unstructured data, we’ve talked about a few times, give me your personal, the Meno Hellis definition of unstructured data. How do you explain this to people?
MH: So I mean the way I see it is that it’s qualitative. It’s data that cannot be processed or analyzed by conventional data tools. It’s stuff like I mentioned, like emails, social media, web pages, or even customer feedback from chatbots and so on. It can be images, and when I say images it can be PDF scanned documents, for example, these are images, audio, video. And those are very challenging because this type of data is rough, it’s very unorganized. It cannot be processed with your conventional technology or tools. And in most cases it ends up being inaccessible. It just gets stored in its raw format. And from a regulatory point of view, for risk are very nervous about this because they don’t know what kind of data is in there.
CW: What’s in there? Pandora’s box.
MH: And the regulations, most of the regulatory bodies, don’t really care how the format of your data is, the regulations apply to every data that you get and it’s kind of your own problem to figure out. How to access it and figure out, does it have classified information? How do I figure out, who can access it? Who should access this data, if I don’t know what the data is? I can’t read it, I can’t process it. So that’s kind of what we mean by unstructured data. It’s the format and the way that it gets stored, and the challenges that it brings to organizations in terms of provisioning, access, security risk. And to a lesser extent analytics, because data is a currency, it’s insights, most people use data for process improvement for identifying new opportunities for the company. You can’t use that data in its format. You have to have specialized tools to digitize it and then you can then look at it.
CW: Yeah you can’t put PDFs in an Excel column and then run a V-lookup on them can you. The accessibility is one that I’ve seen over and over again where it’s, I asked questions of a client like, how many of these documents do they have? And they couldn’t even answer that question, because they were stored here and they were stored here, and someone else owned this repository, and who knows which ones were duplicates, because things get copied around.
MH: No that’s pretty cool this is why actually in many organizations they’re afraid to unstructure data as dark, because it’s dark you don’t really see what it is, you don’t know how much of it you have.
CW: Yeah and and the easy ways of measuring it, like how big is the file, that doesn’t really tell you anything, it could be full of meaningless tables.
MH: I mean it might be helpful for infrastructure sizing, but not necessarily from an operations management.
CW: Right, absolutely. Where have you seen success with unstructured data?
MH: That’s a good question, so it’s not all hopeless. So there are situations where you can employ emerging technologies to access the data and make sense of it, give it meaning and lessen the time it takes to process and for folks to, let’s say, get to it. The successes that I’ve seen is people employing what we call OCR or ICR. I’ve had experiences where I’ve led groups that employed ICR, intelligent character recognition, to process documents for account maintenance services that were basically transactions, whether it be change of address, change of beneficiary. And the solution worked really well as far as classifying what kind of document they’re working with, extracting that data, and then making it available to either an API or RPA. To then manipulate it and feed it into a record keeping system or downstream system. So it can be successful. It does require a very well thought out mapping of the different technologies at the different stages, so that they can all function and work together to get to that kind of success and solution. But it’s possible, and I would say not only it’s possible, it’s becoming popular. Because a lot of organizations are realizing that, although the assumption that there is that, hey the demographic is changing and people becoming more digital savvy, they’re ditching paper forms and using self-serve portals and so on. That’s true, but also at the same time, it’s taking a long time, and you have other forms of unstructured data. Because you’re trying to interact with your clients at every touch point and capture the data because that’s valid information whether it be emails, whether it be on youtube, whether it be video, or any kind of touch point that you have in different channels. That too is unstructured. so the percentage of unstructured data from the overall data that you were getting is actually higher. But the traditional one, which is the documents, machine type PDFs, even handwritten. And sometimes I’ve seen success with cursive writing, you can you can still get quite a good solution built, and significant reduction in terms of effort from your processing team, if you have have set it up right, you’ve spent the time to to train it, to supervise it, and you’ve used, I would say good technology, and you partnered with somebody who is familiar with that technology that can help you jumpstart this.
CW: You raise a number of good points there. One of them that I wanted to highlight was, you highlighted some of the unique challenges of unstructured data, and I think one of them is training. Unlike a column in a spreadsheet which sort of tells you what’s in the column, unstructured data, you have to bring your own intelligence to it. This is what I care about in this document, or this audio file, or whatever it is. And so as you’ve undertaken unstructured data projects, how have you pitched the need for training and building that intelligence into the business?
MH: That’s a good question I think because that’s often missed as I pointed out earlier. The nature of the technology does require training for it to be effective and highly successful. But people assume that it’s just another technology, and most technologies don’t need training. You just deploy it and it does specific activities. So if you have, for example, a network monitoring, you just deploy it and it starts monitoring your network. That’s what it does. That’s different for the document management, and OCR, and ICR. There is an expectation that you need to give it samples and train it right. And the way we’ve approached this is that, we’ve explained to them that even the human eye can make mistakes. When I look at a document, especially if it’s handwritten, I can’t tell somebody’s, I can’t tell sometimes my own wife’s handwriting. What did you say here? If that is the reality, why should we expect the machines to know everything? You really need to point out to them this name is Smith, for example, this is John. That way they know what the word John looks like next time it sees it.
The technology is getting better, but it’s in your own interest to train it so that you can customize it. Maybe the quality of the PDFs that you’re getting because every case is going to be different. You cannot take a solution that has been trained by different organizations. You have to train it based on the data that you get, based on the quality that you have to deal with. Maybe the quality is determined by your own client profiles. if you’re dealing with many seniors maybe then the quality is going to be handwritten perhaps. Versus if you’re dealing with most of your investors as young investors, there might be machine type PDFs. It varies significantly. That’s why I say that it has to be unique, it has to be done at your organization. It cannot be acquired somewhere else. It’s no different than culture. It’s no different than if you bring in somebody on board, you train them. There’s specific skills they expect them to have. They’re asking for somebody to know Python, for example, yeah they should know Python. Doesn’t matter where they come from. But if you want them to know how things get done here, whether how to navigate the organization, how to do governance, that’s very specific to your organization.
CW: Yeah and your people have to learn it, so of course the bots have to learn it. You have to teach them well. Yeah it’s funny, I was thinking about, I have a 14 year old daughter, and I imagine at some point when she’s talking to an investment manager it’s going to be Snapchat. Unstructured data is only going to proliferate, to your own point.
MH: And maybe it’s there’s no chat it’s just emojis, who knows?
CW: Gosh I hadn’t thought of that. Unstructured data processing for memes. That’s probably what’s coming.
MH: The industry has to evolve to meet the needs of folks, so I wouldn’t be surprised. But I think I’m impressed with the technology out there. I think it’s catching up. There’s a lot of smart people that are trying to solve problems, and I’ve worked with organizations such as Mass Challenge, I’m not sure if you’ve heard of them, yeah, it’s a startup accelerator. And I’ve seen a lot of solutions out there that trying to solve very unique problems for organizations. I think multi-nationals are now more comfortable dealing with smaller startups and boutique vendors. That’s a big change from, let’s say, 10 years ago when much of the solutions built, were built from a select number of large vendors.
CW: Yeah 10 years ago no one gets fired for hiring like IBM Watson, it’s a household name. But that’s changed a bit. You also raise a good point in that the tools are catching up, and I think the market is catching up to the fact that, just like with RPA and more conventional automations, there is still a place for the human in the process. Straight through processing on every document is not realistic right now. Well, maybe Tesla will eventually build fully automated cars, I think humans are going to be behind the wheel for quite a while. I think the market’s realizing that, with business critical processes.
MH: For sure I think if anything, what the technology I think is doing, or is going to do, is to really make the human in the loop interactions richer. To give an example, what I mean by that is that if you’re a wealth advisor, then and your distributor or your dealer is using technology to manage your investors, then you have this 360 view of your client, you can easily get an alert as the advisor that your client just had a baby for example, or their family circumstances have changed. Then in response, either automatically or you, can decide to offer them different investments now that they have a baby. Maybe a family education just as an example. You couldn’t do that before the technology advances. If you have this 360 view of your client capturing every interaction, capturing every change in their profile, whether they’ve changed their address, and they moved to a different location, whether their family has changed, and then have a newborn, you can take advantage of that information that’s now being immediately, in real time, curated and given to the advisor to take advantage of. And use it as an opportunity to further develop that relationship with that one. So I think the technology is making the human in the loop interaction much more valuable and much more richer than it was before.
Another example for like where I’ve seen, it’s probably not happening but it’s being talked a lot about. If for example, that every time the market goes down, it’s been going down significantly, a certain number of clients start checking their portfolio. You can automate a message to comfort them. So let’s say you have hundreds of them, now you can be more personal. You’re delivering more personalized services to your clients that you weren’t able to do previously. You can send a personalized message saying, look I’ve checked your portfolio, don’t worry, don’t make any changes, we’ll stay put and then the market will catch up or something like this. Whichever matches it is, but that is one of several things that you can do using technology
CW: It could say hey, hey dum-dum don’t put all of your retirement in crypto for example. This has been great, we’re coming up in about five minutes left and I don’t want to take too much of your time, but I want to ask you a couple of more provocative questions to close it out.
So there’s scary data out there that would suggest that less than 10% of all AI and automation initiatives get from sort of like the lab bench or development environment to production. Why do you think that is?
MH: I think there’s a lot of skepticism unfortunately. I think there’s a lot of, I wouldn’t say fear, maybe just doubt that the technology isn’t there. And I think where that is true is in organizations that have had rigid structures. What I mean by that is that there isn’t really space or capacity for experimentation. They’ve not set up digital factories or incubation teams that can experiment with what the emerging technologies out there can do.
Those organizations I think are missing out on bigger potentials. Because I think there’s been perhaps false promises in terms of AI and what it can do. There’s always going to be a group of people that’s promising that technology will do more than what currently does. That’s true for everything, not just technology, anything. Whether it be investments, there are investment advisors who will say that your portfolio will be up 15% every year, and that’s not going to be the case, or something like this.
But that’s not to take away from the hard fact that machine learning, NLP, AI, all these technologies have a lot to deliver. If used correctly and if used wisely, it can actually transform the business, it can make a significant improvement. I think the key factors, like we talked earlier, is to really start small, experiment with something that you’re comfortable with in terms of failure, and be ready to course correct and step out if it doesn’t work right. And work with an advisor or a partner. I think if you go it alone, especially if you’re on boarding a technology that your in-house experts have not had any exposure to, or an emerging technology that you’re not familiar with, it’s important that you partner with an advisor and somebody who has done this before, so that they can bring in lessons learned, and they can bring in experiences from other companies, how they’ve done. That way you’re increasing the likelihood of you being successful. So that skepticism, unfortunately, is because it’s just that a lot of people are doing it and they’re trying to dabble and experiment, but then when it fails they don’t say, we didn’t do it right, they say the technology sucks.
CW: The idea of working with a partner is just great advice. I think maybe five years ago you probably had to go it alone, especially if you were trying to tackle unstructured data, but now you don’t. You should be talking to experts, they’re out there. And some of them are right here.
MH: There are many experts out there and many advisors. I think what’s important is that you team up with somebody that can guide you throughout the process. It doesn’t doesn’t have to be… I’m a big fan of, if you’re signing up with an emerging technology, let’s say a vendor, that you sign up with their professional services to jumpstart. I think that’s helpful. A lot of folks feel that, oh they’re being taken advantage of, when they’re signing let’s say, a statement of work. We thought we’re going to be buying the software, now we’re buying the professional services. It makes sense. Once you’re comfortable you have matured, you have your COE set up and everything, you can step away but don’t do it alone if you’re not comfortable.
CW: Returning to one of your earlier points, you should have a mindset that you’re building a stack, and right tool for the right job, and if you don’t have the expertise, be honest about it. Most of the vendors out there, including Indico Data, we want to use our professional services to enable you. Because ultimately we want you to have the skill set so that you can do more on the platform with the tools.
MH: I think that is in line with my experience, a lot of these SaaS-based companies understood they make the profit from the subscription based on revenue. They’re not really interested in the professional services, it’s just that for the SaaS subscription to be successful. For you to use the product correctly there has to be some kind of professional services to get you started, get you off the ground.
CW: Final question, I’m just going to open it up to you, what is your number one piece of advice for the COE leader out there?
MH: The number one advice I would say is that it’s really to be… keep an open mind about the various technologies out there. Make sure that whatever the COE does is aligned with an overall organizational strategy, it cannot function on its own. That means it has to trickle from the top. The top has to have a digital transformation strategy that is supported by a business architecture unit, and then one of those capabilities is the COE that is supporting those objectives. if you don’t have that sponsorship, if you don’t have that alignment, I think no matter what tools you bring in, no matter what talent you’re bringing in, you’re bound to fail. So I can’t stress enough the alignment with the organizational strategy. And making sure that you have a clear purpose and vision of what you’re trying to do, and it’s aligned with the leaders that you’re trying to serve, whether it be the operational leaders, or the chief strategy officer or the chief digital officer, there has to be perfect alignment. It cannot operate as some kind of under the desk business solutions team, or some kind of process automation team that they’re basically dabbling in python. I think those can create a lot of risks, they can create a lot of problems, they can easily become disintegrated and distracted as they get pulled into different projects. They can have, or face, funding issues so all those problems can happen. But if you have that alignment with the organization so that it’s part of an enterprise strategy, then I think it’s more successful. So it’s important to keep that in mind.
CW: Wise words from a wise man. This has been Unstructured Unlocked, I’ve been your host Chris Wells and once again I want to thank my guest Meno Hellis, thank you so much.
MH: You’re welcome thanks Chris for having me. Enjoy the rest of the day.
CW: All right YouTube, best of luck out there automating everything.
Check out the full Unstructured Unlocked podcast on your favorite platform, including:
