Ep. 75: Shifra Kolsky - The Effective Roll Out of RPA Implementation

June 29, 2020 | 20 Minutes

Shifra Kolsky, Vice President and Assistant Controller in Finance has been at Discover since 2009. Shifra launched the finance robotic process automation (RPA) team in 2018, the first RPA team at Discover, in an effort to eliminate manual processes and gain efficiencies by redeploying human resources to tasks that required higher levels of thinking. They partnered with teams across Discover to help launch an overall Discover RPA program which included establishing governance and key messages for promoting internal use of RPA. In this episode of Count Me In, Shifra talks about what to expect (time, investment, challenges, etc.) and the overall benefits of RPA. Download and listen now to get a great foundational understanding of what is needed to effectively roll out an RPA implementation.

Contact Shifra: https://www.linkedin.com/in/shifrakolsky/

Adam: (00:05)
Hi, everyone. Welcome back for episode 75 of Count Me In, IMA's podcast about all things affecting the accounting and finance world. I'm your host, Adam Larson, and I'm pleased to introduce you to our featured expert speaker, Shifra Kolsky. Shifra is the Vice President and Assistant Controller and Finance at Discover she is responsible for external reporting, the SOX compliance program, accounting policy, corporate accounting and financial systems. In this episode, Shifra talks about the value of an effective rollout and what all aspects of an RPA implementation look like. Shifra  launched the finance RPA team in 2018. The first RPA team at discover. So to hear firsthand experiences and actual applications, keep listening as we head over to the conversation now. 

Mitch: (00:56)
So Shifra, you know, we've had a lot of episodes here talking about artificial intelligence, RPA, and from your experience and you just, how you answer questions regarding these topics. Can you first start off with telling us how is RPA different from AI? 

Shifra: (01:13)
So RPA is robotic process automation, AI artificial intelligence, and the main difference is that the way I think about it as the bots are a little bit stupid. So AI tools can and learn from, the different data that they're exposed to and they can develop more sophisticated responses over time. Bots can strictly do whatever it is you tell them to do so they just follow instructions, nothing more. 

Mitch: (01:45)
So as far as following instructions, you know, I know you are in finance and accounting, right? VVce President Assistant Controller here at Discover,  and again, your perspective, what are the best type of tasks for these bots to perform? 

Shifra: (02:00)
Bots are great at doing simple, repetitive tasks where you can give exact step by step instructions. Some of the examples in controllership might include things like pulling reports or setting up journal entries based on specific data fields in that report, preparing reconciliations where the bot would  compare data from one source to another source and create a list of exceptions. So again, all simple, basic repetitive tasks, but we have a lot of those in finance and accounting, and so they're very helpful to us. 

Mitch: (02:38)
And being in finance and accounting, you know, a lot of people probably outside the function would look at this as maybe a cost cutting measure, right? It's, it's a way to kind of eliminate some of the human tasks that are out there, but from within the function and the organization as a whole, really, how do you get in to get people to understand the benefits of these bots and RPA? 

Shifra: (03:04)
Yeah. So when we first launched our RPA program, we were not looking at cost cutting, and we were looking at, ways to become more efficient and free up people's time to be able to do more high value work, to kind of critical thinking things that you need a human to do and so when we took on this program, we started by, we asked people to tell us about the things they hated doing, the things that they found, kind of mind numbingly boring. And thought let's take that list and see if we can get a bot to do those things instead. We also made it very clear to people that it was about shifting people to doing the higher value work, the critical thinking, the analytics, so that people weren't focused on the, Oh my goodness, the bot is  going to take my job. Building a foundation of trust that it really was centered around helping people, was really important to get buy in and to get people engaged. We also enlisted one specific team at the start, to be guinea pigs for everyone. So they test it out. They were the first ones to have a process automated, and they were specifically selected because they had two clear qualifications. One, they had a whole bunch of tasks that were repetitive and easy for us to automate the box. But the other thing they had was a sort of general sense of excitement about the program and the possibilities, and so they were able to really carry the message and they were able to help the bot developers understand things quickly. And then they were also able to convey their enthusiasm to other people as they started seeing the results. And so having those natural cheerleaders or business champions was a really effective way for us to build some momentum around the program. 

Shifra: (05:11)
And what were some of the recognizable benefits of implementing this program? How did it ultimately impact your team? 

Shifra: (05:18)
So there are a number of different ways that this has helped our team. You know, on the simplest level, it changed the energy. I mean, it got people excited and really thinking in different ways. Our team has long been focused on continuous improvement, but this is really taking things to a different level and helped folks think more creatively about the things that we can do instead of feeling hampered by the things that we can't do. You know, so that's one element of it on the people's side, but frankly, it's also allowed us over the course of the last two years to redeploy about 10% of our headcount in the controllership team to take on new opportunities within the group. So this furious focus on automation has really enabled us to keep up with the growing needs, that are coming at us from all of our business partners and, and keep up with those demands without increasing head count. 

Mitch: (06:25)
One question that I hear a lot when we start talking about RPA is the length of time it takes actually to implement the program. So are you able to share how long this whole process took from the analysis through identifying what people hate until you were able to recognize the benefits and get these cheerleaders for the program? 

Shifra: (06:45)
Sure. I would say for us, the research we did before we jumped into it probably took longer than getting it moving once we started. So we spent a good couple months really talking to a lot of other companies and understanding, you know, some of the things that worked for them, some of the things they wish they'd done differently. We spent a chunk of time looking at the different tools that were available and deciding what the best tool was for us. And then we invested in, recruiting and training some folks, and we did all internal recruiting. We thought that it was smarter to take people who understood the business and understood the business processes, and teach them how to use the tool rather than taking somebody who knew how to use the tool and try to teach them the business. So we spent a couple of months with all of that kind of upfront research and, and training. And once we got into the training it was fairly quick. So depending on the nature of the process that you're trying to automate, things, can we fairly quick, if you know how to use the software and you understand how some of the different connections work, okay. You can get something going in as little as a week when you're first starting out. You're more likely looking at things taking between eight and 12 weeks for a process. And again, probably depending on the complexity and the number of different, systems the process might touch. But we had our first process in place within about 10 weeks  and built on things from there. And one of the things that we've seen is over  time, the more processes we build, and as long as we maintain a focus on kind of building reusable components to the automations, the quicker we're able to develop things. And so we took our average development time from a starting point of about 12 weeks down to, we're now closer to about 7 weeks for average development time for a medium to high complexity process. 

Mitch: (09:11)
You're clearly continuing to implement this. And I think that was going to be my next question also is, you know, how scalable is this, you know, how widespread are you automating things within, you know, your finance and accounting function here? 

Shifra: (09:26)
Yes. So, you know, Discover started the RPA program in the finance and accounting. It's a common place across a lot of companies for things to start because they are very easily identifiable processes that you can start with. It's also frequently a place where, you know, we're looking for that efficiency. And so we did start in finance and accounting, but what we've done in terms of scaling things is really, partnered with others across the organization, to help other people learn about what we've been doing and think about ways that the technology can be used in their areas. And so from a company perspective, it's really developing different spokes. The model uses a hub and spoke. So we have sort of a centralized technology team, that helps with, you know, troubleshooting some of the connections, and ensuring that the bots have you know, the technical space they need to do their work. So we have a centralized technology team, but then we have spokes that are sort of business led teams that are working on automations within their own business units, but it really was a question of getting the word out and educating across the company to build additional momentum and to really get that scale. 

Mitch: (10:58)
And then I suppose maybe the most important question when you're looking into implementing an RPA program, what's the cost? How does this really affect the bottom line of the company? 

Shifra: (11:09)
Yeah I think you would get a different answer well kind of depending who you talk to. The answer is different because the way that companies use it they'll have different things that they measure. You know, some folks are looking at hard dollars, some folks are looking  at hours saved, but also the pricing is different depending which software you're using and how many bots, you know, they'll do tier pricing so that the more box you have, the cheaper they get. But it kind of, it doesn't matter because across the board, no matter how you're thinking about it and how you're measuring it, the cost of having a bot do certain tasks versus a human is a fraction of what the human cost might be and again, because the bot is doing repetitive, mundane type tasks that people don't love doing anyways, there's an added benefit because in the time that you're saving, you're also enabling people to think more creatively and to free up their time for much more important work. 

Mitch: (12:27)
Well, that definitely makes sense. And I think, you know, once all this is taken care of upfront and you begin to start the process, as you said earlier, you start to get more buy in when people have this flexibility, but once the program is up and running, I'm sure there were challenges as well, whether it was in your function or across the organization in new functions, trying to adopt RPA. So can you talk to some of the struggles maybe that you came across when looking to set up these programs or things that you would advise others to look out for when they're starting their process? Anything to kind of keep an eye on? 

Shifra: (13:05)
In the research that we did before we started, there were probably three themes that we came across of challenges that people might have. One is around, training, like having the knowledge to be able to actually use the tool effectively. One is  technology challenges you might have in having the bot interact with other systems. And then the third  really, if you want to get the return on investment that you plan for, you really need to have a long list of processes that could be automated and you need to keep at it, over time. So I'll take those each separately, and talk a little bit about how we address those. So with respect to training, I already mentioned, we thought that it made a lot more sense to get internal people who understood the business and understood the processes and teach them how to use the tool. The majority of the software applications available we'll provide, you know, those companies will provide some sort of initial, like self study training to learn how to use the tool. We found it very effective to get a little bit about outside consulting help when we were first getting started. And essentially we had the consultants, you know, with the first process, they did it and we were looking over their shoulder with the second process. We did it and they were looking over our shoulder and what the third process we were doing it and just calling them if we needed help. And that really enabled us to get a deeper understanding. And we picked processes that had enough complexity to them that there would  be certain challenges to work through. And to have somebody holding our hand, as you were troubleshooting on those things and it was a very effective means for us to get folks trained up. And since then, we've developed an internal training plan where, you know, we have folks who are already up to speed and certified developers training any of the new folks who come to any of the RPA business unit teams across the company. The second element was around technology, and what we did there is I mentioned we have this hub and spoke model. So we have some, a handful of dedicated a technology folks whose primary responsibility is to help the different business units spoke teams, when they come across challenges in getting the bot access to certain other applications. And so having people who are technology specialists, but also understand the bot software has been critical to the success of our program. Because again, you know, until you try using the bot with certain other applications, you don't know exactly how it's going to work and there are challenges, you know? You have that little, I am not a robot click box and, and all kinds of things like that, where you have to figure out a way to get around it and allow for the bot to be able to access certain things. And so having technology folks, dedicated to helping with that was also critical for our success. And then the last thing was around having a very long list of processes that could be automated, and what we did there so I mentioned we had, you know, sort of a pilot team that helped us launch and that the team, you know, the cheerleaders and the voice of the program, what we did was we developed with that team, a video that basically showed the two screens next to each other one with a human doing the tasks and the other with the bot doing the tasks. The bots are about 16 times faster than people, and so when you're watching this video in real time with somebody explaining to you, you know, here's how this works. And when the example you're giving is something that other people can relate to. It was a great tool in generating excitement, you know, you could have the chief accounting officer or the controller get up in front of a big group of people and say, this is important program and we want everybody to be involved, and that really only gets you so far, but if you have somebody peers demonstrating to them, like this is how it helped me, and this is what I did, and they're enthusiastic and they're excited and they're sharing it with their friends, you get a totally different level of engagement. And that was a tremendous tool for us. And, and people came away from that with a level of excitement, and started generating more ideas and coming back and ask me, did it work if we tried this work, would it work if we tried that, and so I think that, yet another element that was, that was very helpful, in getting our program launched effectively and continuing, 

Closing: (18:51)
This has been Count Me In IMA's podcast, providing you with the latest perspectives of thought leaders from the accounting and finance profession, if you like, what you heard and you'd like to be counted in for more relevant accounting and finance education, visit IMA's website at www.imanet.org.