We chat with Tom Kelshaw, managing partner, operations, and technology at Wavemaker, a global agency under the WPP umbrella. Tom walks us through his automation practice and how he eradicates dumb work. This is an interesting application of RPA outside of the usual domain.
Interview with Tom Kelshaw of Wavemaker
automation, rpa, agencies, bots, people, data, media, teams, reorg, support, run, roi, business, pretty, systems, driven, development, invoices, build, orchestrator
Mark Percival, Brent Sanders, Tom Kelshaw
Brent Sanders 00:03
On this episode of the podcast, we talked to Tom Kelshaw, Managing Partner operations and technology at Wavemaker, a global agency under the WPP umbrella. Tom walks us through His automation practice and how he eradicate dumb work. This is an interesting application of RPA outside the usual domain makes for really interesting listen. So stay tuned. Thanks for listening.
Mark Percival 00:25
First off, Thanks, Tom for joining us for another formulate automation podcast. Maybe for our listeners, it'd be great to get kind of a background of where you're working now and start there.
Tom Kelshaw 00:37
Okay, great. Well, thanks very much for having me on, guys. Um, I am the managing partner in charge of operations and technology for Wavemaker, which is a global media planning and buying agency part of WP pay, which is the world's largest kind of media marketing advertising holding company. So wavemaker is one of their media buying agencies. In short, we use data and a bit of kind of industry know-how to figure out where the best place to put ads is to reach an audience and drive growth for clients. So if you see a Superbowl ad, there's, I don't know, one in three chance we put it there. If you see a Google ad, there's a you know, one in five chance, wow, you see a Time Square Billboard. And again, one in four chance that we had a hand in our group, at least, we have four or five operating companies on the media side. Servicing you know, 25 to 35% of America and the world's biggest brands. So it's um, media as the game. It's halfway between a madman you're a creative agency, but more like a kind of KPMG era consultancy, you know, we are very data driven. Well, lingua franca is not kind of artboards like in the advertising creative agency side at Excel.
Mark Percival 02:07
So no, it's smoking.
Tom Kelshaw 02:08
No, no, for a long while, and certainly not in America. Maybe in some of the far flung regions of the world, you can still get away with it, but certainly not in America. No, no smoking. So I've been working in tech for over 20 years, I started actually working in advertising and technology in high school, making annoying banner ads, you know, those flashing things? I'm purely because I purely because I knew how to and yeah, uh, well before flash, okay, even flash playoffs. So I learned to code, um, by basically building animated banner ads for cash, in high school, took that into learning about PHP in order to create CMSs to build websites for kind of, you know, my friend's parents for a bit of cash back in the dotcom, boom days. And so, no, no, whilst I did a digital media communications degree, no formal computer science degree, all self taught. So I've carried the dangers and the bad habits of that you might say, for the last 20 years into a kind of into the enterprise environment. So I've worked in predominantly advertising, and technology roles with a couple of startups thrown in, but they always had a tech startups always had either a marketing or a marketing data Ben, that's actually how I got into automation was arm scraping and spidering in order to build competitive intelligence and competitive data, both in like social media listening about 1015 years ago, that was pretty hard, um, competitive intelligence, who's running digital ads, where and how much. So you know, there are big agencies that do this now. But back in the day, we really had no way of driving forward the growth of digital media without showing up showing our clients what everyone else was doing. And it was a little harder. They used to be these books of, you know, every newspaper ad that everyone was running, because it was only one out in one square inch of newspaper, right. But on the web, it became very hard. So I learned a lot about, you know, spoofing, spidering and crawling the web and scraping the web. And that got me into automation. Now, I run the or system in working group running the automation projects for Wavemaker and a large group across the shared services and back office finance nature, into what we call our digital services or platforms. That is all of the work required. To put ads on Google, Facebook, Amazon, upload them, download them and get the data from that. So they started in about 18 months ago in finance, accounts payable and receivable pretty much like every other person starting an RPA. And was also in charge I really at about the same time, or finalizing our, our transition to the cloud for all our communication and collaboration tools, which got me very quickly to become very familiar with the Office 365 stack. So I also work on Office 365 adoption, including the power platform.
Mark Percival 05:41
Got it. So is that typically the platform you're working on right now the power platform?
Tom Kelshaw 05:45
Well, it's definitely the most common automation framework that we have available. We also have, we also run a UiPath stack with orchestrator. Yeah, bots around the world running on Azure. So got a real combination of UI path where it's required. Yep. For You know, there's legacy systems that don't talk to each other. There's complex things that power automate, cannot do with a little bit of data transform, in nymo, alteryx, alteryx, we have a big data automation unit, which I, I don't run, no skills to do so. And they're very sophisticated power users of alteryx, and nine, and I, I get to play around with that, where you know, you've got data, it's in a tabulated form that you need it to do something else, and maybe have some smarts attached. So use a bit about Azure ML studio nine alteryx, for that stuff as well.
Mark Percival 06:48
Make sense and coming from the marketing world, I mean, obviously, a lot of this is I'm assuming, getting that data from these different providers. How much of it is I mean, a lot of these providers kind of sort of allowing you easier and easier access to this data is one of the things where you really are kind of doing a lot of this RPA just for the sake of getting this data in some format that you can consume.
Tom Kelshaw 07:11
Yeah, we were, we're lucky to have many opportunities to use RPA. Because both our structure as the world's largest media and advertising holding company, we were less of an enterprise, I think everyone will be happy to say that we're a federation of, of small and medium sized agencies that have been grouped together through acquisition and development. We're really you know, 50 to 100 agencies under a group and we deal with, yeah, thousands of clients, sometimes as if we're a federation of thousands of client teams. So standardization has been a big challenge. And then you add on to that the realities of our industry, which is experiencing significant amount of fragmentation, and very fast moving from, you're putting billboards and newspaper ads up and measuring those. And that's a fairly well trodden mechanism. Yeah, we've been doing it for almost 100 years through to AI driven programmatic audio media, which there's no rules, there's no standards. So it's kind of the whole gamut of media. Investment Management, means that we are stitching together systems that were built in the 60s and still ran on mainframes, built in the 70s. With AI driven systems that, you know, from Digital first, you know, media platforms. Yeah, yeah, Google, Facebook, Amazon, etc. So we work on a lot of platforms that are not ours. Industry, ERP systems and mainframes. It's not ours to update. So handling that are stitching data between things that don't stick to don't normally talk to each other, has been the work that I'm doing quite a lot of, in addition to a bit of the standard. You know, SI SIP. ERP finance systems that are pretty universal in RPA.
Mark Percival 09:20
And so when you have that many, I mean, you have this huge set of client teams, essentially. How do you standardize on something like bot development?
Tom Kelshaw 09:30
Yeah yeah. Well, I'll caveat by saying was, it was still pretty early in our journey we really kind of committed to, to we had done some little point solutions and experimentations pilot to prcs. And actually across all the the main RPA technologies, a blue prism automation, anywhere and UiPath. We did sell on UiPath for basically ease of use our most important Ross's his kind of audit, trail monitoring, understanding, really, if we're gonna let 1000 flowers bloom and we'll talk about citizen development, I'm sure is at least having a hitch to orchestrator. Let us know who's using what, and for what, but it's probably without having a fully fledged CIA. It's really no way for someone that's trying to guide even globally with, yeah. 12 13,000 people around the world. Yeah, who's using what? Right UiPath seemed to give us the best chance with the, with the way that orchestrator worked to having some type of formalized platform. So we ended up settling on it, it was easiest to adopt. I run our we call it training, it's really kind of evangelism sessions where we train business users, technical generally, you know, people that write VBA script macros, people that use other data transformation automation tools, but not certainly not in software engineers. Well, I'll train them on UI path, we'll do a two hour sprint, we'll get through something that basically just says, Hey, if you've ever wanted to, I don't know, grab something off a website, put it into Excel, do something with it, and put it back into another website. And I'll be honest, that's most of our job, in a nutshell, is grabbing stuff from one data source, doing something in Excel and putting it somewhere else. Then UI path was by far the easiest way for me to get people to understand how that works. And start to think with like an RPA mindset.
Mark Percival 11:37
And it sounds like the users have any sounds like a two hour sprint, and they seem to be pretty, pretty happy with it. And pretty, pretty gung ho about jumping on it.
Tom Kelshaw 11:46
Yeah, I mean, the killer app for UiPath. And I know that the other platforms have this now is the Community Edition is the, you know, zero pay to play. It's just, you know, if I tell you we'll let you install it, then you can be developing bots, and that is just transformational? Um, yeah, it's less about the process just and the power of free there has meant that I can get people using this for zero dollars down. Yeah. And that gets them very excited that this is now a tool in their toolbox. That said, I'll admit we've had not as much success with the kind of, you know, engaging champions or citizen developers as No, maybe articles or thought leaders might promise about citizen development. Yeah, there are some standouts, I'd say most of our automation teams around the world are actually glorified citizen developers were a loosely regulated militia. But we all have other jobs. We have a very small team of full time, automation engineers that do this as a job. The rest of us are building bots maintaining, supporting, you know, from the full automation development model, we are probably, I'd say advanced citizen developers with a very small, committed automation engineering team.
Mark Percival 13:15
That makes sense. Yeah. Now, if you when you introduce this to someone who's more of that citizen developer role, are they looking at this and your company as a, hey, I want to run this on my own system. And it's like an attended bot. And that says situation, are they looking at it, like, Oh, this would be great if I came in, in the morning and said, run overnight, and everything is where it needs to be.
Tom Kelshaw 13:34
They want the latter. And that's really what I'm trying to develop. Because that's really the magic. You know, there is definitely some magic in the attended bot development. Just as I said, in terms of evangelism, if someone can build a bot that does a thing, that that is pretty similar to a job that they have that they don't like, we call that dumb work. You know, our mantra for our automation program is killed on work. I want everyone out there with with a bead with a target, on dumb work that they see in dumb work that they do, and we just have to kill that. And we have, we hire very smart people. And like any industry, some part of the work they do is pretty, pretty dumb, pretty mind numbing, because we just have systems that don't talk to each other and honest monitors are humans. So with their mom, they will law you know, they will get very excited by seeing work that they do and then attended by just happening really fast. And even like when the letters get typed out really fast, you just see this kind of flashing excitement, I think people understand then you know the the reality kicks in that that part breaks that the day after you build it and you know, once the guidance is off, then you really do need to start very quickly bringing in a more centralized, you know, development force with a proper engineering framework. Whether that be yeah agile or, or others in developing bots supporting and maintaining them through a co a, but we, we try to avoid worrying too much about the problems we were creating early on, we did want to do that, you know, let's just start it and say let's create those problems, let's cross those bridges when we come to them. And I think that is still the best way to do it knowing that's wrong. But I would rather get, you know, 20 meetings with 20 people making 20 bots that are going to break and 20 meetings about PowerPoint about a CEO a best way. Yeah. Because that's, you know, we call that yeah, that's pending the nursery before you've even started making the baby. You know.
Mark Percival 15:42
Right. Yeah, I think you see that not in a lot of tech, though, it's the desire to make it perfect the first time. When in reality, I think you mentioned it, which is knowing you're doing it wrong is okay.
Tom Kelshaw 15:53
And you just have to be very candid about that. And you'll have consultants that have come in to talk to you about automation, saying, hey, you're not meant to do it this way. And I'm like, we know, and we know it's wrong. But for us, it's really about getting people's hands dirty. Yeah, getting them to see, you know, those proofs of concept, those proofs of value. So that, you know, we can start to build up the excitement. And the energy and of course, thus, the budgets and the strategic commitment to automation. And for us that was only ever going to happen in a more grassroots or ground up way, just due to the nature of our business being so let's say federated with, yeah, lots of different stakeholders, lots of different decision makers, doing work in lots of different ways.
Mark Percival 16:41
So you've got all these bots, and they're all you know, some are an orchestrator. How do you manage support?
Tom Kelshaw 16:45
Yeah, um, we're, we're, we're just starting to get to the point where that's now a really, really good question is, is we're implementing a global RSA, in order to have it. So it's not the we have a support engineer locally, we have a couple of support engineers, often offshore and globally. And it's now at the point where we're running enough bots that we need kind of a full time in a robotics operation, Command Center, and they're going to be looking after bots so that we can scale those out, you know, it, the way we were doing it for the first year was the devs did the support. And you know, two hours in the morning, they respond to friendly, yeah, Trello tickets or teams or meetings, and then they get back to development. And now, you know, there comes a tipping point where your dev team can't interrupt their sprint to do support on the last part they did. Because scale adaptation, configuration, etc, start to take up. So read about the level of the maturity model where we're starting implemented an RSA.
Mark Percival 17:59
Now that makes total sense. That's a I think that's an interesting metric. It's figuring out when, when to make that change, right? It's sort of looking a lot of industries. Anytime you're adding somebody it's adding that next level of, you know, an a del devel Delve, sorry, developer ops engineer, or somebody who's development support engineer, is when you make it.
Tom Kelshaw 18:20
Yeah for us it was the, yeah, it's when the light went from green to yellow. You know, that's where all we are, where we're not being as efficient as we could be. But we've done enough with it with a pretty lean, initial team and initial investment that we can show value to the people that make the decision. You know, we started from pretty grassroots ground up level, we just looked at low hanging fruit there was going to save money. So we have a, I like to think a pretty well established assessment criteria for the jobs we decided to take on. And one of the ways one of our raises is, is first we stopped paying third party costs. So if we can automate something that we're paying a third party to do, that could be in your back office support or otherwise great that the moment that bot goes into play, it's lessening or negating the third party cost, easy. it's pretty easy to have a conversation that on ROI there, you know, and then we look at, you know, reducing risk, we deal in quite a lot of money. As you can imagine buying one third of the world's media, stuff can go wrong, when you're dealing with that much money and as an agent, we can sometimes be on the hook for that. So reducing risk is kind of number two. And then number three will be you know, upgrading our humans giving them promotions and automating from the bottom. And that's that's a little harder to, to, to manage the ROI. Just given the nature of our business. We don't have gigantic back offices of people that just do the same thing every day. There are their highly skilled kind of consultants. And once they do You know, regular repetitive rules driven work, sometimes they do 35 to 50 of those types of tasks every week. So our P Yeah, our next challenge is really about kind of a broader BPO approach of how if at all, can we reorg parts of the business to be a little more are standardized, and thus, benefit more from automation? Right now we like to say, our first year is about buying people slightly longer lunch breaks, through automation. And you know, pretty soon the ROI, factor rondelle on that invoice start delivering change at higher scale.
Mark Percival 20:45
Yeah, that's an excellent point.
Brent Sanders 20:47
How did you guys when it comes to introducing support engineers into your current development staff? I mean, how did you go about setting that up? I mean, at what point, you know, there are standard terms and ways of doing this. Historically, and like, generally IT industry, this L one L two support, I mean, yeah, how did you guys set that up? And what was the sort of methodology behind it?
Tom Kelshaw 21:13
Yeah, so I think I think this has come up on your podcast before but and we've done the exact same thing is where, for whatever reason, RPA seems to be the it that it don't get to, to do with standards, or they just give it to the enthusiastic hand raisers and the the kind of the maniacs that think they can run a, an IT division without consulting it. And, and we are, you know, we would, I'll put my hand up and say, Yep, absolutely, we did that as well. So we're only now just going back and looking at how support might work, given the very established standards for whether it be, you know, support engineering, onshore, offshore, or shared support services for all of our other IT projects, and digital systems and things like that. So we are just like, we haven't yet solved for that. I guess the short answer is the question. We're just looking at how we might structure it, but it will be closer to how we support other. Yeah, it projects and programs that we have, unless, like, teams of developers on teams and Trello, who answer questions if and when, with me and about five other people orchestrating or managing priorities via tickets. You know, that's not the way to do it. But we were small enough that we could get away with it. We're now you know, serious enough that we need to probably look back at a more established framework for, for managing support. And yeah, luckily, there's plenty of that. It is just there is a kind of a prodigal son returned to the IT department at some stage. I think we're, we're about to turn around.
Brent Sanders 23:10
talk to you about you know, as you've built up this practice, and you built up bots for a period of time, like how do you manage reporting on ROI as a or an ongoing? Is that something you revisit, like on a regular basis? Or is that a constant use of dashboards? Like, what's what's worked for you?
Tom Kelshaw 23:27
Yeah, we have a couple of metrics that for me, I in ROI, and you know, everything from dollars, save to time, save to your F do full time equivalent FTE employees reassigned and promoted. Again, that's what I see as the driving, kind of, you know, the driving force between all automation we do is to give people better work to do and to kill dumb work. So ROI is staggered. And we measured that in kind of, we develop a framework, when we assess, we use kind of a hodgepodge of what we've seen consultants come in and show us the UiPath automation hub, kind of metrics around assessing your your automation project pipeline also helps a lot in terms of hours saved, in terms of dollars saved, etc. and report that back to various stakeholders. We're not at that dashboarding phase yet, given most of the work done, we aren't measuring in kind of in a volumetric way, you know, we're not, we're not doing cases per hour, because our business isn't related to that. I can measure the improvement in turnaround speed on some things like Ah, yeah, you're quite a cash, invoices processor day and, and we've been able to deliver some real benefit that in time taken, you know, to go from Four hours a day to 15 minutes, when can measure in kind of some things like that we're not running a full ROI dashboard yet that's definitely going to have to come. It's just that there's such a diversity of value currently, right and metrics that it's, it's sometimes hard to, to even show that on a two dimensional plane, you're like, are we talking, FTS, reassigned, reallocated and promoted? Are we talking hours saved? And then do you have a stakeholder that even agrees that an hour saved is actually worth anything? Right? If you're paying an employee to turn up, and you're still paying them, but you've saved half their time, you've only done half the work like automation is only half the job of, of good operations, the entire job is making sure that those four hours a day you've saved are being filled with higher value work. So I'm, you know, I'm in charge of half the job and the rest of the businesses is definitely looking at how we deliver true our AI, which is, you know, moving someone into higher value work, and not rehiring for that position. Because it was Yeah, it was deemed of low value. But are we given just the nature of our work? We have trouble a site, I wouldn't be comfortable putting a dashboard in with hours saved, because someone's like, great, how many? How many FTS to remove from this department to that, like what none. But again, they are really long lunch breaks now. And it's doesn't lend itself just the nature of our business.
Brent Sanders 26:35
Yeah, we were, you know, we're constantly looking at, like, even our own RPA projects of like, what's the right way to quantify and there's always like, Hey, what's a record in a database worth? You would like this super granular level? And we bring that back to people that people like, Yeah, that's great. And it could be worth this, but like, we can't do you know, some of the things that that we've done, it's like, we just can't do this with a person, you know. And then some of the cases, it's not necessarily even replacing an existing process, it's creating a new one that unlocks a whole sort of new Gentry. And it's, it's hard to put a number I mean, it, it's part of our sort of ethos is like, Hey, we want to be able to report on this month over month and understand like, his support and all this other infrastructure is that, is it, does it still make sense to run this bot? would, you know, should we just sunset and go back to, you know, a human doing the whole job, but it's, it's hard to establish those metrics, and I've seen some attempts at it. And it's like a mind numbing excel sheet. I'm just not.
Tom Kelshaw 27:42
And I think all models are wrong, some are useful, but some are just too wrong to be useful. And it's specific. And it's more assumptions than facts. And it's more caveats than actual data. You know, and at that point, you're like, Look, let's take a step back. And is the principle here. Do we agree that saving someone two hours, you know, this team two hours a day, ah, mind numbing, low value work? That has some value? Yes. Okay, great. Do we agree it's worth X dollars a year around? Yeah, probably actually. And I think there is a danger in trying to over quantify some of the returns in the early stages. Particularly when you're, you're still delivering bots at a relatively small scale. It's when you need to, you know, move significant amounts of money and significant amounts of people that you know, you need significant data. A lot of the time you know, it's when I get the excited smiles of associates that don't need to copy and paste out of Excel and they jump up and down and demand more bots that yeah, that's thanks enough, isn't it?
Brent Sanders 28:55
Yeah, I'd say so. I mean, that's there is a lot of delight, right? That you can deliver with somebody seeing the thing that they have to do you know, I don't know what most people but I feel like the the first part of the day you have to do with your coffee because it's so it's such a struggle later in the day, you just can't bring yourself to do it and say hey, let's just take that off your debts.
Tom Kelshaw 29:15
Quite a few of the kind of more bespoke, I would say, custom VIP automations that we've been making that are jobs that are done by your management or senior management after their kids go to bed with glasses of alcohol. You know, some of those things like you were saving six hours a day of very expensive people. And you know, those six hours whilst it you know, it's compliance work and stuff that only they're allowed to do but it is not a job for a human being, you know, yeah, it is sending 500 individual emails with a data point in it and they need to legally or for Sox reasons or whatever, get some data back and that is as much as as much fun as it is to do that while the kids are in bed with a glass of scotch not great work?
Brent Sanders 30:01
No, no, I don't care how good the scotch is. Interesting. So, you know, transitioning into some of your experience, I would love to hear about, you know, what has been your experience with quote unquote Intelligent Automation or AI machine learning? Like, how is that blended into your RPA? practice?
Tom Kelshaw 30:20
Yeah, we use it a little bit, we do some of the very standard, you know, the, the very, very common usage in terms of document understanding for for accounts payable, so an invoice comes in, what kind of invoices to be, to be candid, we are we use it as little as possible. Because it is the nature of our business, we buy media, with media platforms, or publishers, you know, TV stations and newspapers and Google and Facebook, and there is a pretty fat head of people we deal with. So if we're looking at invoices that are coming in, we pay media money on behalf of our clients, right? So if Procter and Gamble or L'Oreal or Mondelez want to put an ad on TV, we go buy it for them, we pay the bills, and they pass. So the invoices that come to us come from, you know, a significant volume of vendors, but there is a fat enough head that it is being worth for us, we started in AI ml. And we're like, it's just not as good as as hardcoding, some of these data extractors, another exam, so So we, we use hard coded C sharp, kind of custom or regex, custom data extraction for a lot of things where the volume is there, because it's a lot faster and more reliable, where there is that kind of almost things in our finance department, we've used some AI ml and and in matching, because we're dealing with, hopefully similar data across disparate systems that don't talk to each other. Sometimes that a record that is meant to relate or unite with a record in another system does not have a unifying key. So great job RPA. Because you can get data from one system and put it into another or match it to another. But if you don't have a unifying key, that's really easy for a human, they can tend to look at a bunch of different data. And go I have that obviously, I've been working on in this industry in this business for 35 years. I know that that matches that. Why it's common sense. Well, common sense is pretty hard to code in a kind of if then else logic. That's where we really started to use AI ml, some machine learning to try to build up, you know, a sense of common sense within some of our projects. And that's worked. Okay. I'm definitely not as good as a human. And it's pretty hard to explain to someone here like, oh, I've got this really advanced AI system. And so why doesn't it know that that is that? Well, yeah, no, let's continue training. So we're using it a bit, but really, not as much as I thought we would be using those tools, we are still relying on, you know, established rules based formulas, etc. That said, we're using a little bit of natural language processing, actually, you know, a bit of custom middleware we built for UiPath, to murder teams. So using some chat bots to initiate orchestrator, jobs, and to act as a human in the loop system. So we've actually had to a lot of our work is just consultative enough, that and just not rules based enough that it's not, it's impossible to fully kind of automate a couple of systems together. So we have a lot of human in the loop solutions. Some of those require, you know, free text and we use kind of a chatbot for that. Other times, it's just you need a very small automation kicked off, just open up a team's chat to your bot, and just tell it what to do. And it comes back either with Thumbs up, thumbs down link to a log, our teams, Microsoft Teams has a great new adaptive cards thing, which allows you to have some very basic forms and, and kind of a card data interface there which, which are people like in teams. So we've done some kind of natural language processing are some, you know, document understanding, and then some kind of fuzzy matching 11 gene distancing type for supervised learning.
Brent Sanders 34:49
Interesting, I'm going back to a prior point that you were talking about around you know, organization, and this is a trend that we're seeing from consultants that We speak with that are in one of the management side, and they're saying, Hey, you know, we've embraced RPA, we've embraced automation. But we need to create this sort of new structure that is sort of automation first, and creating, basically departments that have cross pollination sort of built in, you have an element of IT and have an odd automation element of finance, you know, one team, how have you guys been looking? I mean, it sounds like it's still pretty early on. But what are your thoughts around that? I mean, have you seen it implemented? Have you gone ahead and made some changes already? Or are you thinking about some of the changes?
Tom Kelshaw 35:37
Yeah, so we've been able to see some kind of, you know, early, kind of green shoots come up for actual, you know, proper transformation of the business based on automation. And that is, as I said, there are parts of the job that are quite consultative, and we deal with clients, we interface with vendors, we come up with creative solutions to for marketing and other business problems. And, and what that does is mean, we were often structured in quite like an artisan or workshop where people come in, they join a client team, they figure out how that client likes to work. And each account team kind of developed its own culture. And that's kind of a times its own way of working. And, and it becomes very hard to standardize other work that they do, where are other agencies that we work with, and we're lucky enough to have a big family of sister agencies that do roughly the same work as a group. So yeah, that other agencies in the group have been able to distinguish and consolidate parts of the work. And we've been able to look at that as a model and say, well, when you do go from, you know, an artisanal workshop to a more, you know, taylorist, industrial revolution 1.0 factory, wow, yeah, you really can justify bringing some tools to bear for those people. Because rather than doing 55 things a week, they're doing five things. So automating two of those is significant. Whereas automating, you know, two tasks, or two processes out of 55, let's say for the month, it's really hard to kind of make a dent. So we've been able to look at some teams that are that given that the diversity of the teams, you have some teams that are far more regulated, standardized, are more compartmentalizing the work everyone does. And see that that is definitely of interest in moving forward to a more automation first automation focused or really efficiency focused business.
Brent Sanders 37:47
Interesting. Interesting. Yeah, it's, uh, I always wonder about for companies that are sort of already you know, many years into it, and mid five may have a lot of success in what they do. And, you know, if you could flip a switch and reorg without there being any sort of issue or, or to say, Hey, I'm gonna start a brand new agency. And if I could, we did the same exact was, this is how I would do it and try to think about that from time to time. And, you know, it's hard to know that, you know, a reorg is actually going to help me obviously, you never want to throw the baby out with the bathwater. There's obviously a structure that makes sense, and it's there for a reason, but was interesting kind of thought experiment.
Tom Kelshaw 38:28
Yeah, definitely. As I said, given the nature of our holding company, we often and have done globally acquire, you know, a hot new startup advertising media or technology company and we can learn from that because we can see where we didn't know start at the same time, media agencies actually spawned out of creative agencies became a specialized different unit. And then they know, split and combined and wavemaker itself as a combination only two years old, a merger of two global agencies. So there's such a flux in the industry. The advent of ad tech and marketing tech on the kind of accelerates that that we actually do get, it's a fairly um, heterogeneous ecosystem, at least in our illicit web. So we do get to see examples come up fairly regularly of different ways of working and even of holy net new born in a digital age agencies, and they do inform kind of broader strategy at the at the group and agency level, we can see how things are done when you are not shackled to, you know, legacy systems, frameworks, ways of working, even business models. There's enough diversity in our group that we can see work that might be interesting to to adapt that might inform either a reorg or restructure or just an adjustment of day to day this strategy. But yeah, it's definitely it's like your say, Well, if you just started again, you would, you know, build from automation at the center, and you do away with a lot of the ways that we do things. And that's, that's a great flight of fancy. But yeah, it's often not practical, any type of significant IT, or operational technology reorg for us would be, you know, what we call a skeleton transplant? Maybe possible, but pretty painful.
Brent Sanders 40:30
Yeah, it's interesting to see, I know, you know, I'm, I had a little bit experience working in sort of creative technology and, you know, had a run in with plenty of agencies and I, it's funny, I do feel like the, the industry has started to shift in and sort of meld with, you know, the, I should say, like the Deloitte and the McKinsey they were they, they even spun up their digital practices. And it's, yeah, that sort of stuff that you would expect a WP company to do is now in that sort of management consulting realm, and vice versa, it's kind of that circle is getting smaller and adjusting, obviously, the agency I'm putting up in front of yours, you can't see. But that industry's changed a lot, even the last five years.
Tom Kelshaw 41:16
Yes, both management consultancies and marketing and media consultancies thrive in uncertainty. You know, a client knows what they need to do, that is they need to sell stuff, you know, they need to delight their customers and continue to grow. Um, given the nature of, you know, the fragmentation of media, the acceleration of digital technology, change, change, change, how they go about solving that problem? Well, with enough change, and enough volatility and enough uncertainty, a lot of people can come and tell you that they know how to do it, and they know what to do. And so that's when you do get, you know, and we work pretty closely and often quite well with some of the management consultancies. And, and, you know, there is, though, more and more overlap, I think just given that, because there are no experts, and there are no right answers yet. So, you know, how do you deal with? Yeah, how do you deal with selling, predominantly in person experiences and products and services in the COVID era? Well, we have a POV on that McKinsey. And BCG and Accenture have a POV on that, you know, what we're all about as knowledgeable as each other. I guess it depends whose PowerPoint deck you want to buy?
Brent Sanders 42:44
That's really well put, I do give that some space, that is an interesting take, because I just do remember, you know, very early in my agency career, seeing this, like, you know, we're doing creative stuff now. But like, you know, let's say it's a website or a web app, and this is like, early 2000s, you kind of start tracing, okay, where's this gonna go? You start, you very quickly start getting an operations. And it's like, what's the difference between, you know, pretty pictures now that we're getting into technology? We have to integrate closely with the Oregon it's been fun to see that evolution. I mean, it's obviously as you say, uncertainty, I think we're gonna see, this is not so much the forecast of the industry of RPA. But it might be an interest, you know, might be related in a way. But I do think you're gonna see the rise of the agency, again, when this sort of, hopefully soon post COVID era. But in the general uncertainty, as you said, I mean, it was a flourishing time after it was a great time to build an agency, when the stock market, you know, basically went bonkers the whole bailout that was happening and people just the hires weren't there, the towels, and as things get a lot better in the economy, the agency started folding up, and yeah, it's anything I'm curious to see where you know, what role RPA takes, I think it's going to be a significant one.
Tom Kelshaw 44:09
Well, I think, you know, what, the industry is not kind of monolithic. But I think some of the truth there is, agents, agents, any business that sits between, you know, an end consumer and another business, we've, we thrive during periods of growth, and we thrive through periods of change. Because, you know, in this case, a marketer. When there is a new digital platform, a new data lead way of selling something coming out every year, you just don't have the time to invest in that. You want to go to the marketplace and pick an agency that specialized and has the diversity of experience of deliver on that. There are these things that I'll butcher the name called k waves, Kondratiev waves, um, and they really talk about that ebb and flow the rise and fall of different technologies. Geez. And you know, back in the day the.com boom that you talked about it was digital. And it's like, we don't know anything about digital, we needed agency to build us a website to sell stuff. And then all the agencies grew on that. And then there was this low, and then came kind of social and mobile, you know, and that the ultra fragmentation of media, and at the same time, the consolidation of, of audiences into these aggregators were really, you know, it's globally. It's Google and Facebook. Yeah, that's how you advertise. There's certainly a very strong place for your broadcast television and print and everything. But Google and Facebook, we've hit Yeah, almost, we're hitting that the down point of the explosion and fragmentation of new media vehicles and new ways to advertise. And it's kind of another, perhaps another Valley, and there's k waves. And that's where, you know, it's time to tighten the belt, where it's time for, say, operational efficiency to come back to the agencies where we're not just growing like crazy. And basically, it's like, higher, higher, higher, you know, like a, like a startup, you know, with VC money underneath it. It's now time to look for efficiencies. And that's why RPA yes, we do some work in automation for our clients, and solving some of their bespoke needs. Really, I think now's the time to look at ourselves and, and look at the best way we can service our clients and help them drive growth is to be lean and mean and really efficient and effective ourselves. And so we're looking at your RPI being one of those tools to drive that. Yeah. Pretty massive improvement.
Brent Sanders 46:42
That's great. That's a great insight. Mark, anything else to add? I feel like this has been really cool. Yeah, for sure. Now, this is wonderful. Thanks.
Tom Kelshaw 46:51
Yeah guys. Thanks so much. I appreciate that I hope some useful stuff in there.