Robotic Process Automation

Robert Amberg:

This is Strategy That Works, where we discuss practical solutions to companies’ most complex challenges. I’m your host Robert Amberg, chief marketing officer at Jabian Consulting. Let’s dive in.

Hey everyone, welcome to the podcast. Today we’re talking Robotic Process Automation, or RPA, what it is, how it’s useful, and some best practices for companies looking to implement it.

We’re joined today by Will Funderburg, a director in Jabian Consulting’s Charlotte office, and Jed Manfull, a manager with Jabian, also in Charlotte. Will, Jed, welcome to the podcast. Thanks for being here.

Will Funderburg:

Thanks for having us.

Jed Manfull:

Yeah, thank you.

Robert Amberg:

All right, so let’s start with just the basics. Robotic Process Automation, known as RPA, can probably sound to some people like Skynet going live and taking over the world. What is it really and how is it used?

Jed Manfull:

Yeah. Well really it’s an application that sits on a desktop or on a virtual machine, and it emulates integrations. It can emulate traditional integrations between applications or it can emulate what a human would do with applications in the same kind of way.

I gave that definition to a couple clients once, and I got a real blank look back from them. So I came up with a little bit more of a robust definition that it seemed to kind of click with them. So how I wrote it down was, “RPA tools are designed to mimic the same manual process paths humans do, by accessing multiple applications and performing rule-based tasks. An RPA tool operates by mapping, think like a process flow, a process for the software robot to follow via computer pathways and various data repositories.”

So RPA can interface with multiple applications in the same way that a human can. RPA can be triggered manually, so you can launch the bot to go do the tasks that it’s supposed to do or you can have it running 24/7, 365. So anytime it gets any kind of a trigger, like an email in an inbox or something that it’s monitoring, it knows to go do some kind of a task.

It can move or populate data between prescribed locations, it can document audit trials, it can conduct calculations, perform actions, and it can even trigger downstream activities for a person to pick up and go do.

The biggest ‘gotcha’ I guess is it has to be something that’s very logical, something that is driven by business rules, where it knows that if this action occurs, it must do this, and if this action occurs, it must do this. It can’t really weigh them out.

Robert Amberg:

So it’s more if/then and less AI?

Jed Manfull:

Yeah, exactly, that’s a good way to put it. AI would come in and kind of help you be cognitive and think through, “If not this and if not this, then strategize the next thing.” And RPA, for the most part, is not there yet.

Will Funderburg:

So maybe an easy way to think about it is if you ever had Excel and you record a macro that says, “Highlight this and turn them green.” Whatever, very simple. It’s recording those steps for your desktop. So it can open any application, it can recognize shapes and click on those shapes, it can enter passwords at login screens. So when you see a recorded version of sort of a pilot RPA, a bot … And a lot of technology vendors will show you recordings, it’s fascinating because as soon as a web page loads, you’ve got logins already entered and ‘okays’ clicked and what’s making it slower is the internet actually loading the page, not the bot being able to recognize and enter in information, so it’s really cool.

And you can download community versions of these things and play with them yourself. I tried to create one to log my time sheets for me. But to give an example, Jed talks about how kind of fragile they are, I had it open in Excel, and then look for the time I wanted it to enter for my week but because the Excel opened in a smaller window, the page size in the top right corner was different, so it didn’t recognize that shape, and it stopped and failed.

So just as an example, they follow very, very specific directions, and if anything changes, it won’t work. So you got to have a really good handle on the very, very small tasks and things it needs to recognize.

Jed Manfull:

And that’s kind of like part of the development process. Like when you have a use case that you decide you want to go and you want to automate this, then you create your process flow, and then you spend about 1000 different times figuring out the little quirk that you have to fix, and you go and fix it, and then you come back, and so that’s kind of the development process. But it happens in weeks, not months or years, like other interpretations can happen.

Robert Amberg:

Sure. And I would assume that the ROI for doing this, I mean it’s like you said, an incredibly detailed task, but it’s RPA, you know, very popular right now and it’s obviously not unpopular if it wasn’t successful. So why do you think it’s been so popular lately?

Will Funderburg:

I mean, I think one piece is you got to look for areas where the volumes are really, really high. So this task is repeated 1000 times a month. Nobody wakes up in the morning and wants to go to work and do something 1000 times a month. It’s not invigorating, they’re not passionate about it, etc. or places where, for example, call centers have a lot of use cases.

If someone calls in and because we don’t have all our systems integrated perfectly yet, I need to look up, for instance Jed Manfull, customer number 1-2-3, I need to look that up in nine different systems. I can have a bot do that while I’m greeting him for 30 seconds versus a human having to do that.

So wherever there’s volumes or wherever there is high customer touchpoints or value of that person’s time, that’s where you’re going to kind of get your biggest bang for the buck.

Robert Amberg:

Okay.

Jed Manfull:

I’d also… to add on to that, completely agree. I think the barriers compared to a lot of traditional implementations are pretty low. The cost for the software is pretty low, the licenses for the bots are pretty, pretty inexpensive and the time to implement… If you… this is just kind of round numbers, but if you had a process that you wanted to automate, and if the volume was there but the processes only needed to be automated for six months, it might still make sense because it might only take you a couple of weeks to stand up that automation. Whereas if you were doing an ERP implementation, well you would never even consider doing an ERP implementation for six months, right? You’d want that thing to last for years and years because it’s going to take you years and years to implement it.

So I think that’s one reason it’s so advantageous is that kind of low barrier. And then another way that something can be really valuable is depending on the timing of when the tasks happen… So for instance finance. Finance loves RPA. It’s helping people close books a lot quicker. So even if it’s not the hardest task, giving finance back an hour at the end of the month is very different than giving finance back an hour in the middle of the month. So if you can help them close those books, things like that that are pretty time bound, that’s a huge win there.

Robert Amberg:

All right, so obviously benefits across the organization, who actually… In an organization, who controls RPA? Who manages it? Is it technology? Is it finances, operations, like what have you seen?

Will Funderburg:

Yeah, we had a client experience where a large consultancy came in and did sort of an SG and A assessment across the business and said, “Here’s some big numbers you can go save, and part of those numbers can be saved using RPA.” That was delivered to a finance organization, and maybe it’s just the nature of finance organizations, but they didn’t necessarily buy into the soft numbers right away. But technology saw that same study and said, “Okay, I know RPA is coming, I know it’s a big trend, and soon enough, I’m going to have different business units standing up these things on their own desktop because it’s easy, because they don’t need IT.”

And so our client I think was looking forward in a really intelligent way to say, “I know it’s coming, I better get my arms around how I want these things to be built in my organization, how I want to support them in the future.”

So The short answer to your question is, it depends. And If folks are pretty savvy, any department could take a manual repetitive task, load a bot on their desktop, and automate it. But I think at the end of the day, IT needs to have their arms around and have some policies around how much of that they’re going to support because the worst case scenario is someone goes and builds a bot to replace two FTE, all of a sudden the person supporting that bot leaves the company and now they go, “IT help. I’ve got a big issue and this thing’s crashed, and now it’s affecting business.” So they would have inherited an issue they didn’t even know existed to start with.

So I think IT has to have a handle on it, but who owns those bots day-to-day is a really interesting point to jump into of you need to sort of treat them as they are employees. So if I’m IT and I build you a bot and you’re in operations, I need you to manage that thing. So it can’t just be “Oh, I have an emergency something changed.” You should know what’s going to change in the future and plan for that and then make a request so that we update that for you.

Robert Amberg:

That’s interesting. Treating it like an employee. That’s a good way to think about it.

Will Funderburg:

And there’s also some very interesting points around… that I think organizations are just starting to realize, internal audit and compliance. What if a bot goes and does something? How do you reprimand a bot? Do you dock their pay? No, you know.

Robert Amberg:

Does it take PTO?

Will Funderburg:

But the responsibility falls to that manager, right? That manager decided this task was something a bot could do and it’s low risk enough, etc. Those accountabilities need to go to that manager. So that’s something that on the very front end of setting up an RPA capability, you need to get right and get understood between the parties, people who are going to utilize it and then IT who will support.

Jed Manfull:

Yup. We had a client contact that has kind of an interesting structure as well. So similar structure, IT is owning and managing the capability, and they’re in charge of the governance, but they have almost a studio type of experience where someone from a business can come in, and they can play around and they can even develop their own automations. And then the technology would kind of manage that and launch it. But they allow the business to come in and drive and play around and see what makes sense for their organization.

Ultimately business owning those processes that are being done by those robots, but the bots themselves, the application, the technology being owned by IT. And that helps to avoid kind of a spaghetti architecture, where you have a bunch of different people that have potentially different instances of bots, if you really didn’t manage it centrally, to make sure that you’re avoiding that and making sure that you’re pushing updates appropriately, and you’re taking advantage of licensing benefits.

Will Funderburg:

Yeah, his analogy was like the music studio, right? If you think you have a good use case, come on in. Lets.. I’ll sit you down with one of my developers and let’s see how feasible it is to actually do that. And then once we get something that we think is robust enough to actually put out in production or live, by the fact that you came into the studio and helped build it, now you’re inherently bought into supporting it in the future, versus you make a request, I go build it and ship it to you or say, “Hey, it’s live, Rob.” You weren’t a part of making sure it was correct, making sure it was robust. And you putting time into it means you’re bought in and will own it better in the future. At least that’s the hope.

Jed Manfull:

It helps you turn it around faster, too because you have that dedicated business person there that can truly make sure that the… Even if they’re not driving the design and creation of the process themselves, they’re there to make sure that it is adhering to what they’re wanting it to do appropriately. And then from a change managers perspective, it’s huge as well, like you were saying. Now you have someone on the business side who’s developed a process and identified a process and they kind of own it and they were the champion of it. And they can go tell their other business friends in other areas. That is just going to expand and enhance your capability across your organization, which is ultimately what you want and that’s the desired effect. The more places you can be in an organization, the more mature and more required your capability is and the better chance that you can have it for a long time.

Robert Amberg:

So, are you seeing companies hire RPA experts, specialists, like is there a role that is, “I’m the company’s RPA person, and I’m going to manage and oversee and….” Is that an emerging role that five years ago maybe didn’t exist?

Jed Manfull:

Absolutely. I think the way… Whether you’re centralized or federated, there’s usually some kind of RPA lead, and he’s driving the capability. And depending on if you’re bringing capabilities in house or whether you’re outsourcing your capabilities in just outsourcing your design and development, he might have a little bit of a team with him that helps stand up the bots. And it depends on how they want to scale and what size they are and how many automations they want to have, how many organizations with… or how many places within the organization they want to impact, depends on the size of that team but yeah, it’s a cool role.

Robert Amberg:

So, if you’re the RPA lead, what does he or she have to think about from a security standpoint? Like is this something with all these bots out there running, if you can build your own with licensing, with updates, with someone leaving and it kind of going dormant, what kind of security does that pose the organization and is that something that you’re seeing people really take into consideration?

Will Funderburg:

Yeah, I think anytime we ran into discussions around security or compliance, the easiest way to think about it was treat the bot like an employee. So they have to use logins that are provisioned specifically for them. So they can’t just use Rob Amberg’s login, right? I can’t have 10 bots using your login to go do something. So they each have their own login provision to them. And if each bot is assigned to a manager that’s responsible for that, generally those two rules keep you within the realm of you’re not going to get too far outside where you’ve created a huge risk.

But I would say that is on the front of a lot of folks mind, especially the internal audit, IT compliance side as well. And I think they’re just, at least in the example, the clients we’ve interacted with, they’re catching up with it. They’re seeing the benefit that it’s affording and they’re catching up with, “So how do I make sure there isn’t a bigger risk for the organization while we make sure we support the benefits we’re getting out of it.”

Jed Manfull:

I think where you’re going to see audit compliance go is, once they get a little more comfortable, start to think about how audit and compliance applies to bots potentially a little bit differently. For instance, segregation of duties.

You could think about that a little bit differently with bots than you do with people. You might not need as many handoff instances with a bot because, I mean, if you had two bots that were handing off back and forth, how’s that helpful?

Will Funderburg:

How is that different than one with two logins? It does kind of change the rules a little bit.

Robert Amberg:

Interesting. So what are some of those… I mean you might have just answered my question, but what are some of the biggest mistakes organizations are making when it comes to RPA implementation and what can they learn?

Jed Manfull:

So, kind of back to the benefit that I’d mentioned about it being really enticing because it’s low barrier to entry, you can turn around and you can launch things live, you can launch processes live pretty quickly, low cost.

What that sometimes means is that one area of an organization hears about RPA, gets really excited about it and goes to IT and says, “Hey there’s this thing called RPA and it’s pretty cheap and it can automate these processes and I’ve got these five processes. Can we go do this? Can you buy a license? Can we do it right now?” All focused around this one, one-use case, handset of use cases, says one area of the organization that’s really jazzed up about it. And so they stand up this whole kind of practice and capability around this one area that was really excited about it, kind of their champion, and doesn’t think through all of the other factors that you need to think through to stand it up for long-term success.

So what’s our value proposition? What does this look like in five years? How many use cases do we think we can automate? How are we going to scale up? What’s our governance structure going to look like? How are we going to integrate or interact with different business areas? How are we going to do change management? How are we going to market ourselves across the organization because we’re essentially a service that’s being provided as an option to the organization.

An organization could not see or not put as much emphasis on this capability, meaning RPA. And so, if you get into a budgetary cutting situation, you might think of RPA as one of the things… as a place to reduce cost, because it wasn’t adopted by the entire organization or the organization… Maybe the entire organization isn’t really aware that this is a capability that they could take advantage of.

So not doing the groundwork upfront because it’s so easy to get into, I think is a challenge that some clients face.

Robert Amberg:

Sure.

Will Funderburg:

Now, what I’d add to that is not thinking ahead far enough to say, “What do we do when this RPA…” or “How do we not build a bunch of bots that is supporting an entire business function? Are we avoiding the bigger issue of actually implementing a system, an enterprise system, etc.”

We had one client say he considers RPA the least worst way to integrate. So you can get the benefits of integration by having a bot pull two datasets, compare them, etc., load it back up, or you can actually integrate that system. But looking at it as a stepping stone versus a end-all be-all.

Another one I’d say is, we worked with clients and then worked with RPA vendors they had, it seemed like the RPA vendors were a little reluctant to talk about selling a single bot. And when you start thinking about the business case you say, “Well if I automate five or 10 use cases at X dollars a potential benefit, I don’t know that I can justify 10 bots at once.” I realize you might want to sell them at 10 at once, but I’ve got to sort of invest and get the benefit in lockstep and grow as you go.

So just being willing to ask those questions and bring that up before you, like Jed said, buy five bots but you only have one use case in mind, you’re going to quickly outpace your ability to deliver on that business case.

So a different client has been super disciplined about… He will bring in another FTE within his RPA organization, as he automates enough to justify the time saved for the organization. So he sort of stays lockstep. And you could argue he’s being a little conservative but you know they’re three years into it, I’d say that’s probably a pioneer of RPA honestly.

And so he’s doing a really great job there, at least in that client example. So just making sure you don’t over-promise or over-invest in something without a long runway of use cases, and then having a plan for what you’re going to do when you actually replace these sort of least worst way to accomplish a task.

Robert Amberg:

So if you want to go buy a bot, is this like going to the App Store? I mean do you see it going that way where pretty soon there’s going to be… If anybody can come in and make a bot, is that something that you see happening in the future, from literally like App Store-type acquisition?

Jed Manfull:

That’s an interesting way to put it.

I mean the way that it’s structured right now is you go with a vendor, and you can buy… Well I guess I’m aware of two different structures for how you scale the amount of bots that you have. One is by bot and another is by bandwidth.

So you can buy… I don’t even know what the numbers are or the units of measure are for the latter, but I don’t know like gigabytes or megabytes of bandwidth of bots and however many bots take to do that versus buying bot by bot. there’s a pretty good chance that I would estimate that it stays somewhere like that where you have to be subscribed to a vendor and then through that vendor, you can always scale up the number of bots that you have.

I also think that that is the more popular of the two right now, meaning purchasing bot by bot. My hunch is that it would stay like that for a little while.

Will Funderburg:

Yeah, but I mean if you think about it like buying a car versus renting a car or Uber. So I don’t need five to be sitting not busy waiting on me to come up with use cases to make sure I can keep those five busy. I think you’re going to pretty quickly get the pay for the share of the service. So I think that latter model is going to come, in my opinion, that’s going to come pretty fast because you’re going to say, “Look, at the end of the month, I run X number of transactions, and I’d love for them to happen between midnight and 4 a.m. I need a huge amount of bandwidth for that amount of time. Price that out for me.”

And again I think it’s another example where if you have a long runway of use cases and you can have a business case underneath you, it puts you in a much better position to negotiate with vendors. Now again, we’re not talking about millions of dollars, so it’s not something you spend five months figuring out your strategy for, but I think the more preparation you have, the better suited you’re going to be able to sell it internally, and you’re going to be able to negotiate with vendors you might want to use.

Robert Amberg:

Yep. So the planning has almost less about the dollars and more about the ancillary impacts and the time saved and how you’re going to maximize the efficiency.

Will Funderburg:

I think so, and then back to your question of so do you go the App Store. We saw multiple vendors that have what they call a freeware version or community version. So I’d say go download that and start playing with it and come up with some simple use cases and just understand how it works. I think a couple of days playing with it gets you a lot smarter on how these things work and where you might be able to use them.

And then, like I said, I think the other side is going to be the paying for a share of bandwidth. So it’s not going to be a buy a specific bot. But I will tell you, we talked to one vendor that their view of what the future looks like is every single machine has a bot or at least a share of bot assigned to it, so that a part of everybody’s job could be, what are the things within my role that could be automated and could be done error free and overnight and all these things. It would be a part of everybody’s responsibility to say, “What of what I do can I hand off to a bot that’s available to me or to bandwidth that’s available to me?”

Jed Manfull:

Yep, and then kind of the next level of that being if everyone has a bot on their device, and that bot is running, and it’s attended or unattended, is that bot monitoring your tasks and telling you, “Hey, you do this really simple thing, 15 times a day. I think I could do it. Want me to do it instead?”

Will Funderburg:

Yeah. For a couple of these, you can get to the dystopian future of what happens if the bot’s monitoring you and decides it can replace the task you’re doing?

Jed Manfull:

And then it becomes self-aware and it’s kind of scary.

Will Funderburg:

But I think the funny thing is, as soon as you change a field or you change a period, it goes to the wrong place, or a comma, it fails, right? So somebody has got to have, like you said, responsibility for the actions being performed by all these bots. So if you think about, and maybe I’ll talk about this on our org design podcast, but the whole span of control and how many direct reports can you have, what’s that mean for bots? How many bots or how many routines or use cases can a single person actually oversee responsibly? I don’t know. I don’t know what the number is for direct reports, somewhere around seven. How many is it for bots? 500.

So, these are some of these things that I think businesses, compliance departments, IT, HR are going to have to start exploring as this gets more and more popular.

Jed Manfull:

I see. Basically what you described is I think the future is basically AI monitoring your bots and so you just sit back and watch it happen.

Will Funderburg:

Have a bot monitor a bot, audit that bot. That’s another… Again during lunch, after some of these projects and we’re coming up with strategy, we’re thinking, you know how to sell it to audit? Show them how they can automate some of the audit tasks they have to do until you get into this sort of circular conversation of, so who audits the bot if the bot is auditing?

Jed Manfull:

But that’s a true benefit to audit and compliance is, humans are going to sometimes record the steps that they take that would then be audited, sometimes they’re not going to record the steps that they take. They forgot to, they didn’t want to, or they were really busy and they decided not to.

If you tell a bot to record its steps and if you have an audit log of that bot, it’s going to record everything. And if you train audit on how to review said log, you have really good adherence and really good compliance to whatever policy the bot is supposed to be doing, whether it’s doing it 100 percent of the time or zero percent of the time, you’re going to know.

Robert Amberg:

Yeah.

Will Funderburg:

For just the amount of data that’s generated—

Jed Manfull:

Right

Will Funderburg:

That’s not when humans do tasks. If a bot is doing it, the data is there for you to go audit.

Jed Manfull:

Which then gets into business intelligence, it gets into big data. You have a lot of analytics to play with.

Robert Amberg:

Awesome. So let’s say I want to get into RPA with my organization, right. I’m sold. I like it. I want to jump in.

What’s the one thing you suggest someone should do if I’m going to go pitch RPA to my organization and say, “I think this is what we should do.” How does… What’s that first step? What should I do?

Will Funderburg:

Yeah, I think one thing we’ve learned over a couple examples is we’ve talked about having use cases. We’ve talked about having a large runway of them. But as you do that and you go interview folks across the business in different departments, instead of just focusing on, hey, what’s the worst part of everybody’s job that they’d love to get rid of? And you know, in some cases, people get excited about that, and they’re like, “Oh yeah, I have 15 things that my people hate doing.” But also focus on the other side, what would you rather do with your time? So that you have sort of a push pull of what pain could I remove for you and what gain or what are you excited about doing? So that you have the positive side of change management too. So that’s what I would offer is gather use cases across the organization and have sort of a hit list wish list side so that here’s the things I want to get rid of and I want to automate and I want to be error free, but here’s all the stuff that I want my people to be excited about the opportunity to be able to do.

Jed Manfull:

Yeah, I agree with that. I think starting with engaging different areas of the organization is a great place to begin. Obviously, do your ancillary research ahead of time that way when you go to said places you can answer their questions about “Well what is RPA?” and “Can you show me a little bit about it so that I can then have a conversation with you about what things might be available in my area to be automated, either things that I want to get rid of or things that I’d like to go do but I can’t because I have these things?” And build that consensus and build an idea of what the appetite for RPA in your organization holistically is because that’ll help you make a next level decisions like, how am I going to set up this governance and how am I going to set up this center of excellence and how available, am I going to make it to the organization kind of depends on those conversations that you had. And then that also allows you to kind of build out the range of potential benefit and cost and build out a more traditional business case in that sense.

And then the really cool thing I think about implementers and vendors in the RPA space is how willing they are to come in and do free work for you. If you have… and not just the easiest use case that you can come up with. Come up with a decent use case that would really kind of test the ability of the bot and see how it feels within your organization and how it feels with your infrastructure and reach out to implementers. You can reach out to more than one, and a lot of times they will come in, and they will automate a use case for you just to show you what their capabilities are.

And that’s a great way to kind of get into piloting pretty quickly, and then have some additional proof to show different areas of the organization.

Will Funderburg:

And by proof, the recording of that bot doing that thing that used to take three hours. That may be the most valuable piece. And again it’s a piece… It’s valuable for them because they get to show, “Hey my tool x did this, look.” And they can have their logo in the corner, fine. But you can demonstrate that to a lot of people within your organization so take advantage of what those folks are willing to offer.

Jed Manfull:

It’s a huge wow factor, right, because you’re looking at a desktop, and you’re showing it to a business person who’s done this thing and it’s taken them an hour on a Friday afternoon, and you just showed them it getting done in 30 seconds. They’re going to be… It’s pretty cool.

Robert Amberg:

Awesome. Well, the future of RPA certainly seems bright. Fascinating.

Will, Jed, thank you very much for being here on the podcast, this is Strategy That Works, we’ll talk to you next time.