Stephen Lowisz: Let's go ahead and get started. I know that there's a few people coming. I just got an email, letting me know that someone else is coming late. So let's go ahead and get started.
First of all, wanted to thank everybody for joining. This is our second Automation Trends Roundtable. First one was a success. Dennis, again, you showed back up so I'm going to assume that it was! But I want to say thank you to everyone joining. That being said, we have it in this format for a reason. It's not a webinar. It's not John or I talking at anybody. It's very much a Roundtable. Come in, give us your thoughts, ask any questions that you think are relevant to the couple of topics that we have, etc.
If you don't mind, I'll have John introduce himself in a second. But I'd like to go around real quick, since there's only a couple of us and that's intentional, we try to keep it between six and ten people. After that, it gets to just be too big, and not as much of a Roundtable type of format. So with that, and a couple more coming, Jess, I'm going to pick on you since you're the first one on my screen. Do you mind introducing yourself?
Participant: Hi, everyone. I'm Jess Lewis, I’m a senior manager in the insurance consulting team at PwC, Canada. So I work with both life and health and PNC clients, particularly on market strategy and operating model design. We do have a broader team that extends more on this kind of technology integration. We do have an analytics team as well, that's involved in the insurance space. But my focus is more kind of the upfront strategic design. And I will be joined by Matt Lawrence momentarily. In case he gets skipped on introductions, he focuses in the life and health space. So excited for today.
Stephen Lowisz: Thanks for killing two birds with one stone. Mike, you're next on my screen.
Participant: I'm Mike Mobley. I'm the VP of operations with pro holding group in South Florida. We're an MGA for non-standard auto for representing two companies in the state have been with the company for 30 years, and we're looking at all any new avenues we can to streamline our process. Excuse my voice everyone.
Stephen Lowisz: No worries. Dan, you’re next.
Participant: Sure. So, I’m Dan McGinnis. I'm the Chief Underwriting Officer and Chief Operating Officer for Cap Specialty. Prior to that, I was at AIG where I crossed paths with John. The responsibility on the Chief Operating Officer side, responsibility for Operations and in IT, and then on the underwriting side, it's under my strategy for the organization as well as actuary and some some P&L responsibilities, while for a couple of our of our underwriting units.
Stephen Lowisz: Cool, glad to have you. Alright, Dennis?
Participant: Yeah, so sorry, I still have to plug my PPC in… Sorry about that. So I'm a senior analyst with Aite Group, Boston based, covering the wealth management marketplace. So, covering trends involving, you know, financial wellness, in the workplace, outside of the workplace, digital advice, but also looking at portfolio and asset management distribution, portfolio creation model portfolios, SME advisory trends and covering a bit on the insurance side as well. So mainly annuities, but looking a little bit at life insurance in the standpoint of wealth transfer, and use for lending purposes. Right, as well.
Stephen Lowisz: Good to see you again, Dennis. Craig, if you want to chime in, if not all good. Craig is on the way to the airport.
Participant: That’s okay. So I’m Craig Hamway, I'm the CEO for a holding that Mike and I work together. As he said, we're a non-standard MGA operating in Florida. And the piece that I would add is that we are you know, just starting our journey with Digital Coworkers with Stephen’s help and the rest of the team at Roots. So excited to launch that, and I'm looking forward to the future together.
Stephen Lowisz: Matt, I think you get a pass because Jess introduced you for you. So I'll, I'll skip over to Rob. How are you? Can you introduce yourself real quick?
Participant: I'm Rob Sarnie. I'm a professor of practice at WPI. And I'm interested in hearing the great things you guys are doing in RPA. Wonderful.
Stephen Lowisz: Stephen Lowisz here, I run Sales and Marketing here at Roots Automation. That’s about all you need to know about me, and I know most of you, so good to see most of you again.
With that being said, we have the pleasure of having John, who is our CTO at Roots Automation, talking about pushing the boundaries of automation in highly regulated, people-centric type environments, which I think constitutes everybody on this call right? Financial Services, banking insurance, etc. John, do you want to give a little bit about your background and and briefly why you're leading this discussion?
John Cottongim: Sure. And yes, let's make it a discussion, certainly. So John Cottongim, as Stephen mentioned, CTO and a Co-founder at Roots Automation, as some other folks on the call know. My initial career started over at AIG, and actually, my background is not the typical coder background. So I actually have a pretty deep finance background. And technical analysis and insurance underwriting is really where I grew up, I have a CFA as well, or I guess I am a CFA.
I did a few things at AIG and including, of course that underwriting technical piece across different commercial and M&A type businesses, but then I also took part in heading up a lean management practice at the company. So really deep diving into the operations. I think a hobby of mine was always process improvement, even in the underwriting world, which was a natural fit, moving over into the lean side of things. And that was the natural path as well, leading over to robotic process automation, AIG was probably customer number three for Blue Prism as they came over to the States. And so we were very early on in that journey.
So the combination of process improvement and technology coming together, I ended up heading up an automation Center of Excellence at AIG, which means when it's that early in this cycle, really getting all the scars, for the first time. In creating a team that both identified and develop those solutions really end-to-end, so I have a very practical understanding of how the sausage is made here. And then, after that, actually moved over to a different industry for a little while over to Mars, everybody's hopefully favorites -- there's really only two -- but probably favorite candymaker. And pet care company.
So that was a very interesting experience. It was one of the best companies to work with and it was a great place to work at. But it was for me -- and why I bring that forward -- It was an exercise in scale. So when people talk about pushing an automation plan or digital transformation across something like 125,000 people across the globe, what does that look like? What's different about that, compared to at a midsize or a smaller company? That also goes into these conversations. I always love to touch on here in terms of build versus buy. Those equations really differ depending on the environment that you're in.
But I headed the practice over at Mars, I started the practice for them and handed it off to a great person I helped recruit. And as I was on my way out to start Roots Automation, and connecting back with Chaz Perera, our CEO (from AIG as well), to start this. I would say the impetus for starting Roots Automation was really trying to accelerate the coming together of this technology we know of as RPA, as well as the AI and Machine Learning components where to do that completely within, let's say, the corporate space. And being just between us here on this call -- you understand the corporate space --you almost have to do corporate takeover internally. To be able to gain all the resources that you need in order to do all of the things you want to do in a shorter period of time, it really requires digital transformation at scale. It's about how do you get access to all those pieces.
Chaz and I had enough. We said, “Let's do it ourselves, help other companies accelerate that process, minimize the integration that's required, maximize the execution of the solutions. And focus on the end customer.” Being that guy or gal on the processing team, who's actually side by side with these automations, leveraging them as the benefit in their day, how are we thinking about them in the solution? So this is really what we bring together: the process improvements, speed to execution, and also that customer centricity. I think that’s the great thing about working in a startup, we get to focus on what matters to us and, and that's what matters to us.
Stephen Lowisz: Great introduction. Question for John. And then I'm actually going to flip it on whoever's willing to participate here. But John, I think everybody on this call has already had the conversation around “Why should we automate?” Well, John, where do you think we are in the lifecycle of digital transformation, if you will? We’ll start there for a moment.
John Cottongim: Yeah. So I think everybody's in the, let's say, broadly financial service space or insurance space here. Yeah, I think that right now we're in between this utopia that everybody talks about in terms of, “AI is going to help us solve every problem.” And we're also not in vaporware period anymore, either. So it's really about how do you leverage great AI and ML toolkits and platforms to be able to get at true execution. So I'll take just a topical example today. Across my desk, some of the team members said, “Hey, Google put out this this notification today that said they have this AI document processing. And it's the best thing since sliced bread in processing invoices, and some maybe receipts as well.”
Then you go over onto their site. And, literally, the example they show on their site is missing about a third of the words on the page for these receipts and whatnot. And, you know, if you think about it, people are kind of pounding their chest, talking about how this is as good as the solutions are today and how excited they are about it. Yet, it's only extracting, let's say 70% or 80% of the data that's there. Then we know, as practitioners, we have to then take that 70% or 80% of data and translate that to, “How much of our actual process can we transact? How much can we complete without a complete data set?” And therein lies the core statement, if you will, of where we exist today within the digital transformation space. We are a lot further along than we used to be. But we also have to be very practical and understanding, even with the greatest tool sets out there today. What can we automate at what percentage based on what's available?
Now there's, there's ways to get industry specific or domain specific, and improve on what a big box brand like a Google or Amazon would provide. Because we know that they have to serve every customer that shows up. And that's a very hard task. Because one day it's Mars. One day, it's a Walmart coming to them. And today, they have to serve that customer base. We can get domain specific, we can improve upon that base, but it takes it takes focus. So what I would say just in general, to kick this off, is that where we are today is that you still have to work hard. You still have to work hard with -- and have -- smart people on the team to be able to push things over. And to be able to leverage them and move forward and put them into those types of tools like RPA, where now we have decent digital data. And we're going to put it in a very straightforward linear process. I think that's a good analogy for where we are. There's lots of promise that you can execute, but it still takes a lot of effort. There's no Rosie the Robot to come in and just sweep up the apartment just yet.
Stephen Lowisz: I appreciate the in-depth answer. I know that we have a couple of different consulting companies on here, and corporate -- and even academic -- backgrounds. So I certainly appreciate the diversity of perspectives. How far along are you, or your customers, when it comes to automation and digital transformation? Are they still at the – or are you – still at the email inbox phase? Or is it like cognitive automation or even somewhere in between?
Participant: You know, covering wealth management firms and the financial advisors that work under them, they're further along than they thought they were in 2019, though not by choice. But you know, I still see that a lot of struggle, right. And John, to your point, I think a lot of them saying this is where we want to go, but the data isn't put together. It's not clean. There's not a consensus of what they want to do. They're sort of evaluating providers. So I think there's so bigger firms that I think have always had that digitalization strategy and they’re building things out and moving forward. But then you drop down another tier. And you have firms saying, “Hey, we kept afloat, we're further along than we thought, but we're not sure we've got the right infrastructure in place, or even the data to really move out.” So I think they're almost taking a step back and saying, “What is it that we're looking to automate? Where do we need to build out capabilities? Where do we need to scale?” And a part of it, too, is just prioritization. We have some clients saying, “Look, I've got all these other priorities. And so automation in certain areas are just number 24 on the list, right.” So it may be a 2021 priority, but it may be something that's going to be a 2022 effort that they have to get to. I think it's just a matter of knowing where it stands in their prioritization.
But it's clear, the volume is picking up. We're seeing a lot more business development. They're going to run into something -- whether it's onboarding clients or a number of factors, right -- they're going to start to run into areas where capacity just gets too tight, and problems can arise.
John Cottongim: I suspect, Dennis, that you'd agree with this, but do you see in your space that we all thought 10 year ago that paper was not going to be our biggest problem, but it still is. Did you see that in your space as well?
Participant: You know, interestingly enough, when COVID hit, we went out to the market globally, and the US in particular came out with COVID fairly well. I think a lot of practices and wealth management already had e-signature in place. They were moving into that paperless environment, even though the advisors don't believe it was true, and still having all this paper in their office. But I do think when things went remote, it wasn't that big of a deal. There were some, especially the early days, like in March 2020 that were a bit unstable. Most found that it settled down quite heavily. And we fared better than Japan, and fared a bit better than Europe as well. I think we're in better shape than we thought, at least enough to keep going. But I don't think it's optimally where everyone wants to be with it. And now they don't have a choice. I think two firms maybe had a longer term view of what they had to do and how they want to look at automation. And all of a sudden, their three year plan is now an 18 months or 12 months initiative.
Stephen Lowisz: Anyone else want to chime in here?
Participant: I’d say we were a little further along than just the email thing that you mentioned just from what I have seen. Stephen, we had conversation about this a while ago, but we have a lot of other things going on in the organization, a lot of you know IT and transformational things going on. And one of the challenges that I find with with something like this is where do you start? And how do you really insinuate it into the organization without it becoming yet another massive project? Because it's probably not a sort of second skin to most people in the organization.
When I'm talking to my operations manager about her business and how they're doing things and asking her to give more thought to this and get involved in it, there's just so much on a day to day that's going on. It's not familiar to them. And so you can't fault anybody for feeling like, “Well, this is a major project and how do we even start.” I feel like where we've had success on this, or any other kind of major transformational things, has been when we've had a specific issue that has come up, and we needed to solve for it right away. And so we found a solution and did it. I guess what I look for is, how can we kind of find or create some of those opportunities, and use that as leverage to create momentum to get more stuff done, as opposed to saying corporately, “We're going to embrace this and put together a major transformation plan,” and then get behind it and assign capital and things like that. That’s a completely different animal. I think the probability of getting off the ground is lower, in that case.
John Cottongim: That’s the old “create the crisis” I heard in there. I don't entirely disagree. I certainly think tying ropes to the anchor is helpful. And when I've seen the senior leaders make upfront commitments, you know -- maybe a little bit blind, not knowing exactly what they're going to need to get themselves into or how they're going to solve it -- but know that the they're going to commit to it.
It certainly, certainly helps drive. And these teams -- let's say whether you have your own automation team or you're going to use somebody externally for some of these -- these things can very much get looked at as ‘The Others,’ right? They can get looked at as they have a hammer and they're looking for nails. And I think if you could, as best you can, flip that around and say “This is a solutions team, and there's many solutions that we have at the company that are available to you. But by the way, this is your problem, and you have to go solve it.” I think if you focus on that first, everybody will be in a bit more collegial environment that way?
Stephen Lowisz: I'd like to go down the path of what are the real limitations in financial services, automation. And I think one of the big ones that tends to come up is underwriting, right? Trying to make big decisions and mitigate risk with automation can be very daunting. And I think it depends on the product line, right? So you have standard P&C, which I think most would agree you can automate in large part. Go online and get a quote from Geico, instantly. And then you go all the way towards – I think of like Marine shipping and so forth, these policies on billion dollar ships. But you go from automated to very artistic handholding and personalized take on underwriting. Where does automation stop?
John Cottongim: Yeah, so there's a couple of points to hit there. But in terms of, let's say, complexity, and where we can make the biggest impact, maybe the first point to ask ourselves is, ‘where unstructured becomes too unstructured to take that.’ And I think in the world of brokers sending emails to underwriters asking for things, and describing things, you are in the world of probably still too unstructured, even the latest AI models. GPT-3 is something that's out there in the news that you'll hear about from a natural language understanding and generation.
Even these models today with billions of parameters, don't understand natural language, right? They can read, they could kind of parrot, but they don't really have cognition yet. So in the world of email, unless you're asking people to templatized emails through this process, it's very challenging to work with that.
What is doable is semi structured data today. So let’s take the commercial underwriting side of things. If we're in the world of casualty insurance, and there might be 20 different applications that come in from Chubb and AIG, and they have completely different formats. There's going to be some natural language that goes into those, but they're what we think of as semi structured, in that we can identify common keywords, we can build these lexicons, we can build these dictionaries of what insurance terms mean, and then we can simplify how we understand, and how we can extract meaningful data from these things.
We do not need to templatize each of those forms, right. That, I think, is in the past. You don't have to actually sit there and say, “This form is going to come in this way. And I know where this information is.” Instead, we're generally looking for anchors, we're looking for keywords, we're looking for sections. So the AI is that intelligence.
And then of course, if anything is repetitive, right? That's our distinction between AI and ML. Machine learning is the repetitive things that I can probably explain to somebody how to code, but it's too voluminous. You wouldn't want to do it. AI is that semi-structured piece where you know how to do it, but you really have a hard time explaining how you do it. And I think we're there today. Emails are just a whole other thing. And, you know, not quite yet in an unstructured way I'd say.
Has anybody on the call done a little bit better than that, maybe on the unstructured piece? Or does anybody have a success in the semi structured data that they want to share?
Participant: Not really. I mean, we've gotten better at getting the data, but you know, in terms of actually synthesizing it into some kind of an underwriting decision, that's still in the mind of the underwriter.
John Cottongim: We talked with the company recently and they asked us, essentially, “What can we extract from these documents? They're 300 pages long, and this goes into environmental insurance.” And we said, yeah, we could extract and understand. But I think our question back was, “Can you tell me what the underwriting decision is going to be, if we have a document here?” And even with that document, they're kind of like, “Well, even if you pull out all the information, and even if you can say, ‘yes, it has chemical risk,’ and ‘No, it's not a parking garage,’ even then there's art to it.”
And I think this herein lies the problem. If we can't get a ground truth out of data, and sort of hold ourselves to it, then we really don't leave ourselves the ability to use Big Data, right? It's an art. You know, if you hang a piece of art in a gallery wall, it's not going to be assessed in ones and zeros yet. If that's the case, so be it. But I do think that's a great gate if you're talking to your teams. You really need to get them to give you the input data, to understand how are we going to get a final score on that data. And if we can't do that in theory, you really shouldn’t start the Big Data project.
Stephen Lowisz: I got permission to pick on the two PwC folks here. So thank you for that Jess. I'm curious Jess, as a senior manager at PwC, where do you see organizations focusing their efforts on automation and AI, first of all?
Participant: So it's interesting, because if we were to look back a couple of years ago, everyone was talking about the intake of forms, and being able to look at what information is submitted, and to be able to match it into their backend operating systems more easily. And I feel like we haven't moved that much farther past it.
There's still kind of – as, John, you've alluded to -- that reliance of “Will there should be a human intervention here? We should make sure that the decision is made by an individual.” I find that even as we're building out these more slick front end systems, ways to communicate with the customers, even having that chatbot functionality, they're still kind of keeping that control of the decision making happening with human intervention.
And I kind of see that on the customer side. I think when we look at what's happening just in how data is managed internally -- and I'm speaking of both the P&C and the life and health companies that we work with -- they've done a really good job of building trust in insights and analytics, but they're still trying to get to self-serve functionality, with the employees being able to self-serve versus asking for reports from what would be kind of the CLE teams that are running their analytics strategy. So I think that there's still a ways to go, both in how they consume and what the user experience looks like, but also how they just start to build an environment of people. People are more comfortable using data themselves. People are more, I guess, skilled in their level of data analysis. And I think that’s a big impediment in moving it further along. Because you don't have a broad skill set of data, understanding data visualization, being comfortable with kind of moving beyond just reading reports, but actually generating insights from it. Matt, I'm not sure if you caught any of that. But if there's anything you wanted to add.
Stephen Lowisz: The question was essentially, where do you see your customers focusing on advanced automation at PwC? What does that look like?
Participant: I mean, it's still early days. I think in terms of advanced -- and we work with some of the larger insurers in Canada, right -- I would say they're further behind than the banks especially. I think the life and health players focused on some specific use cases like, for example, one out of the US has recently done some automation in leveraging AI in the underwriting process. And they've been able to reduce their underwriting time by about 80%. So, I think there are specific use cases; it's use case focused. I would say it's not at scale.
And still in the relatively early days, and moving up the maturity scale, for sure. But I would say other potential use cases that I've seen, aside from underwriting, you would probably see it within claims, and adjudication as well. Those are some other pretty cool use cases.
I was having a conversation with the CIO at one of the big life companies a couple of weeks ago, and she mentioned, “Look, everyone talks about straight through processing and that's never going to happen across the board because we're going to prioritize, and we can't do it all.” And I think, John, that kind of resonates with some of your other comments. It'll be focused, and they’ll be prioritized on specific areas based upon the strategic objectives they have at that period of time.
So that's what we're seeing, at least today. Where that might go, you know, it may continue to advance and mature, but those are just some of the early, early things that we're seeing today.
John Cottongim: And yeah, I think if you ask those same folks, if the humans on their team are processing end to end in theory, I think the answer is probably no. The claims First Notice of Loss team, and the tech team is parsing through documents and uploading them, somebody else's adjudicating and, so I think end to end, if we're being kind to ourselves as sort of delivery folks, you know, that that's looking at somebody’s end to end job. Can you essentially give them a junior on the team that would take care of the happy path? Right. And that's what we're looking at, mostly. And if there's one area in the insurance space that's just rife with possibilities, it is claims. We see tons of people parsing documents, and that is a nearly solved problem at this point. So document identification, categorization and parsing, that's essentially solved. The data extraction and the understanding the extraction is closer. Understanding is one step away again.
Stephen Lowisz: Again, we have PwC, we have a gentleman from Mphasis, I'm curious from the consulting side first: how do your customers usually judge the success of automation. Is it FTE reduction, is it capacity increase? What is it?
Participant: Yeah, I would say both of those things, but they're not very good at tracking benefits. Because that’s the biggest challenge, at least in the clients that we worked with. And actually, a lot of the work that we've been doing is helping our clients actually start to realize and track benefits, because it's important.
So but I think they would look at both of those, but they just aren't really good at following through in terms of being able to critically articulate the benefits of the driving. At least that's what we've seen.
Participant: It also seriously hinders scaling beyond the first automation, because they'll go, “Okay, well, that pilot was notionally a success, we saw value in it. And now, we're going to continue exploring other pilots and looking at other areas that we want to optimize.” But they're not fully investing in transforming a process, even if there can be benefits achieved. And I think part of that is just because there isn't the metrics and reporting to fully show the value of it all the time.
Participant: I just think it takes a while for it for those benefits to show up and quantify, right? And there's a bit of impatience at that C suite level, right? Like, “We spent this much on this. So what's the ROI? What's the uptake? And what's the ratio of client-advisor? What's the uptake for onboarding, or for cutting back?”
I think it's going in the right direction, but again, when everyone's looking to allocate money, you know, some of these efforts have been years in pushing it. It takes a while to really get everything cleaned up and working well. So I think it's a hard job for a lot of our clients, you know, in that level, trying to sell automation in certain areas in keeping the senior management on the hook for the two years or so that it takes to really get it up and running. And in asset management too, by the way, the asset managers spend a lot of money on distribution and trying to automate different processes, and trying to make it more effective in your distribution efforts. And, it hasn't necessarily shown up somewhere as a as a key number.
John Cottongim: And anecdotally here, I've been involved in a couple hundred projects, and there's probably been about three of them that would be worth all of the other projects. Those three always have to deal with outward facing revenue generation. And if you can get to revenue generation, particularly if you're a midsize company that will dwarf any FTE savings that you're going to get to. So sometimes you have to pay the price to get in to the show. And, often, that is back office FTE reduction. But really, the biggest impact on the firm shouldn't be getting after new customers, and enhancing products, if we can, to generate revenue. Those sorts of things
Participant: Or linkage to other benefits, right? It doesn't just have to be efficiency, right? It could also be improved client experience; there are other net benefits. And I think being able to crisply articulate the business case and the purpose is also important.
John Cottongim: As a finance person at heart, if not by training, I do think, essentially, at the end of the day, if you can think about it as enterprise value. Even your Automation COE, if you have one -- you should think of that as, “How do I value this as a little mini enterprise inside of my company?” And so, if you're thinking about benefit cases for that team or if you think about what KPIs we move, how does that impact enterprise value? Because I could guarantee you the marketing folks have figured this out. They've figured it out on their spends and what they move the needle on, and how that affects enterprise value without seeing heads move. And so, ultimately, if you could line that up in the same way, I think that's the best way to do it.
Stephen Lowisz: So I don't think any automation conversation is worth having if you’re not mentioning security, right? And seeing as Financial Services is the most impacted industry, likely out of any, when it comes to security with any company that you work with. Whether it's an outside provider, or you're going to the cloud and moving towards cloud based services, they have access to your PII, PHI, etc. John, what do you recommend that organizations -- and especially non-technical individuals -- should look at when thinking about security?
John Cottongim: Yeah, obviously you could say all the platitudes. It has to be upfront in the conversation. And reputationally, this has a huge impact or addition to how you think about developing solutions. All in all, though, I would say that the number one thing that's missed is basic hygiene of an organization, and developing and maintaining and cleaning and keeping what is a poorer proliferation of data and PHI and PII in the back end, clean. I would say policing and auditing solutions.
In a way, that's helping people realize their gaps. It’s probably quite important, especially in large organizations that can afford to take the time and investment to do this, that this needs to be done. Obviously, consultatively, not in a in a way that's going to slap people on the wrist, but that's going to help them make better decisions and have cleaner data sets. And, in the event that there is challenges in the future, you're well prepared, and you've done as best you can to eliminate the need. Let's say for PII in training these solutions, if at all possible.
For an example, we look at medical ID cards. The individuals name is John or Jane, right? That's not something we're going to need to train on. I need to know that it’s a name, but I don't need the name. So, throwing that piece of information out while retaining that it's a name, that's a critical idea for minimizing PII risk. How are your teams thinking about this? How are vendors thinking about it? What tools do they leverage for this? These are critical questions that you should be asking at the start of projects to get there, because in the middle and at the end, it's already too late. The data structure is there. The digital exhaust is all over the place at this point and it's harder to clean up.
Stephen Lowisz: I think there's an interesting conversation around what’s a bigger security risk, which is automation does what you tell it to. Or you have people, who can wander off and make errors. But they have intelligence. How do you manage that balance? That a question for anyone, including John.
John Cottongim: I do think that technical solutions, if done well by a technical team, are going to be more secure, more robust, and have less challenges than even the human teams that are working with them. What would keep me up at night occasionally, in the past running COEs, in having trained other teams to do some of their own automations and wondering, with what rigor they get done, in particular, to some of the sensitive data that the companies have.
And this idea of ‘everybody gets to automate, everybody gets to play with the company's jewels of data.’ We're very early days there, in trying to figure out how to enable that great federated pace of delivering solutions, while maintaining great centralized control of that data and risk. It's essentially not there probably today. There'll be some great startups that figure out how to help us with that, but it's still early days.
Stephen Lowisz: Does anyone want to add something? And furthermore, how do your organizations perceive automation, innovation, etc? Is it inherently risky? Or is it inherently a safer option? What is the perspective that your teams have?
Participant: Most look at automation as ‘you're taking the human error out, you're providing more structured process.’ But nevertheless, we have a separate team that looks at cyber security and other aspects of it. One of the cautionary we give to a lot of advisors is that just because a firm has a client portal doesn't mean that you're not putting in your information. It's maybe automated but it doesn't mean that it's not putting your information or your clients at risk.
So I think there’s firms where I'm surprised the venture hasn't had bigger blow ups in that. But I do think there's a bit of caution where everything is being pushed pretty rapidly, everyone's trying to build portals and API's, and everything else going on, that there's a lot of information out there. Being a little bit more selective and cautious might be, you know, not working with every firm or plugging in different things. Pick your battles. But I think by and large, at least internally, firms look at automation as a way to avoid those potential threats and conflicts, right.
Participant: John, you've brought up an interesting point about the operating model implications. A couple of years ago, we were seeing a centralized group that's building out everything for the organization more and more, and they're trying to be more federated and decentralized. They're even dropping that kind of ‘dotted line’ reporting mentality, and just saying, “The business units need to figure this out for themselves.” But I don't think we often talk about the inherent risk in doing that. And I'm kind of surprised by that. The Canadian market is quite risk averse.
But I don't know if there is kind of the right balance, or if there is one model that makes sense. I think it depends on the size, the scale, the pace of change that the organization is looking to achieve, and the level of capability that they're trying to develop. If it’s that you are innovating your service offering, or if you're automating the processes that your employees are taking on. We collect a lot of case studies as we go of different model options in that regard of COE functions, if they're building out an analytics function, or robotics, or automation. I don't think that there's one that’s a silver bullet solution for everyone.
John Cottongim: Yeah, I think that's right. And where I've seen federated models work the best are in highly technical companies, places like IBM or Verizon -- these companies that have engineers to spare and they're made of engineers, basically. Where they have that mindset and they're thinking about structure and good hygiene in these solutions. And where I think it falls down is almost everywhere else, where it's a services company or it's industrial, where people don't have that mindset, and it's difficult to put them in there.
I think in the phase we're in, especially in the automation space, we're still automating a lot of low hanging fruit, where it is pretty linear in terms of how to get at that. And I do think that as the solutions become more complex, because what needs to get done inherently has more complex data and structure to it, I do think we will see a drift away from the self serve model in this space. I think you'll continue to see a drift towards data analytic self serve, because it's inherently less risky and it's inwardly facing with the solutions. And I hope that the younger generations coming in are more data savvy, particularly with statistics. My one wish of the education system, and I'll say the US because Canada probably better, is that Statistics needs to replace Calculus in school. We need to get an understanding of just simple things like causation and correlation and really get people using things like Tableau and Power BI and all those great tools that are available.
Stephen Lowisz: We started the conversation focused on where the limitations of automation in financial services is. We talked about underwriting and the fact that even GPT-3 isn't up to snuff quite yet. In fact, there's a lot of cool tools out there that you can try based on GPT-3, and it comes out with gibberish yet it's supposed to be the most advanced language modeling out there. So that's one limitation, what are some of the other big limitations that banking, financial services, insurance, etc., really need to be focused on over the next 12 to 24 months, John?
John Cottongim: Well, I think the number one enabler that everybody could spend more time on is not say, data cleanliness, but just the categorization of your data and grouping it into the usable places and repositories. It's going to be messy, there's going to be paper, there's going to be different sources. But I do think preparing for future use cases on your data is probably step one. If you don't have that data repository -- I'm not going to say data lake, I'm not going to say data warehouse -- but just these data repositories where you can then leverage them, where you can bring in teams, you can sort of point to, “There it is, there's my problem, how do we get at it?” I think that's step number one.
If I were a financial institution, I'd be making sure that my teams have their broad brush data groupings in order. And then obviously, I think you have to just start. You have to start. You have to learn. You have to understand PLCs are in the RPA space. I think, please don't do POCs, there's plenty of people who have done it for you. But in the AI and ML space, I do think you need to know how challenging your data is. And it's a space where you might not have the technical know-how inside to know how challenged your data is until you get professionals in. Get the data in a place, get professionals in who could look at that data, and then you'll have a sense of where your challenges really lie.
Stephen Lowisz: What other specific challenges are your organizations facing right now? And I guess the question previous to that would be, how is the data hygiene of the organizations that you're in and/or work with? Is it up to snuff?
Participant: Question for you, John. Do you see firms sitting back and saying, “I don't want to start because the data is not clean?” And the months tick by, the years tick by, the firms are wanting that good hygiene and they’re just waiting for the perfect time? Or are there other firms where it's not perfect, but we’ve got to start somewhere.
John Cottongim: I think it depends on the executive leadership and how much hubris they have, especially on the IT side. And I say that because I do think that there's certain IT leadership that says, “We're going to make this perfect data structure, I'm going to spend two years and make a panacea of a data lake that's going to have everything connected up perfectly. And then we're going to start, and it's going to be amazing.” And I think the reality is those tend to be -- or they can be -- blackhole projects that require tons of effort and time. And as soon as you shovel some dirt in, more forms on the outside, and it’s pushing the boulder up the hill. I think that is not a great way to go about it, trying to get your data perfect. I also think that you could have some naysayers, possibly, who say, “Our data isn't good enough. And we should just get on with it.” That tends to be maybe on the operations side, where they've done it with people in the past. And that's why I think it's the leadership that's going to drive the company culture there.
I love this idea of CEOs and CTOs being the same person. Or, those lines essentially becoming completely blurred. Because ultimately, we need people who understand the process and realize that technology is what fuels our processes nowadays. So we need to be practical, we need to move ahead, we need to execute. We have data challenges, but what can we do practically? And so it just needs to be this nice balance. We're never going to be in the world we want to live in. Here we are now. By the way, we have to make progress before time is up. Otherwise, somebody will show up to out-innovate us and move ahead.
I don't know all the ins and outs of companies like Palantir, or the companies that say, “We’ll make sensitive your data in situ, without those data lake projects and without these fancy things.” I think there might be something there. But it's the early days on that one. But that's something to watch out for: can we -- without that two, three year lift -- start to make progress in a better, more agile way?
Stephen Lowisz: I want to be cognizant of the time here. Any other comments or questions that you'd like to ask John, or frankly, anyone else on in this conversation?
I do want to take a second to again, thank everybody for coming into the second Roundtable. And I want to highly encourage you to participate in the future Roundtables, we'd love to have you. And if you know of anybody, or even would like to be a speaker, we are absolutely looking for a diverse set of speakers, meaning the next one will not be a Roots Automation Co-founder. So Chaz Perera is ruled out on the next one.
I appreciate everybody joining. It was a pleasure to have you all in one long, great afternoon. And please, feel free to reach out directly as well, if anyone wants to follow up on some conversations.
John Cottongim: We'll be glad to! I love talking about this stuff. All the pleasure, thank you so much, everybody.