Today we talk with Jim Walker from UIPath about the growing use of RPA, or Robotic Process Automation, in the Government, as well as Simon Wise of SunCorp Group, who is on the front lines of RPA, changing the way his financial services company does business.
Subscribe to Data Dives on any of the following platforms:
Erik Oehler: Hey, mom. How are you?
Mom: Good, how are you?
Erik Oehler: That was my mom, recently retired; she worked for a health management service.
Do you know what RPA is?
Erik Oehler: I'll give you a hint; it has to do with software.
Mom: Software? Okay, so it's some type of rapid pulse automation.
Erik Oehler: Wow, that's a ... That's a really good guess.
The health care industry is one where RPA or robotic process automation is coming on the strongest.
I also asked my dad who turned 72 this year the same question. He had never heard of RPA, and he's a tech junkie. He helped me to learn to program in Basic when I was six; he's always sending me tech news.
Deloitte's global RPA survey in 2018 reported that only 53% of respondents had started an RPA journey, and only 3% had been able to scale their digital workforce. That first number is expected to rise to 72% by 2020 and achieve universal adoption within five years.
And I didn't ask my mom and dad as an indictment of their knowledge of types of automation, I asked because they're not alone. Nearly everyone I talked to about RPA these days, which is a lot, doesn't seem to have a grasp of what it is or what it isn't.
RPA or robotic process automation is an industry blanket term for software robots completing repetitive rules-based tasks ranging anywhere from scraping data from the web to moving files and folders to extracting structured data from documents and executing processes with it.
And maybe RPA has a branding problem or maybe this is planned for it to sneak its way into the mainstream without anyone ever knowing it by name.
Today on Data Dives we're going to cover RPA from two different angles. First ...
Jim Walker: I can bank online, I can call Uber, I can get Uber Eats, I can choose between Uber Eats and GrubHub. But if I want to get a document from the government, I'm getting in my car, driving down to a courthouse, filling it out, handing it to a person who then types it in, and then the necessary work begins.
Erik Oehler: We're featuring an interview with Jim Walker from the UIPath, a software that makes building RPA robots easy, transforming the government to more efficiently serve citizens.
And second ...
Simon Wise: And so they'll see these robots working alongside them, but from a solution perspective it's still magic.
Erik Oehler: All the way from Australia, Simon Wise, a robotic process analyst at Suncorp Group who is on the front lines finding creative ways to build RPA into his day to day job.
I really enjoyed the interviews with both of them, and I hope you do too. From Alaris, a Kodak Alaris business, my name is Erik Oehler, welcome to Data Dives.
If you ever need convincing that RPA is the future you only need to spend a few minutes with Jim Walker. He's the federal CTO and director of public sector marketing for UIPath whose software is among the leading tools for building your own RPA robots. I've used it and it is incredibly easy to understand. Jim has also worked for Deloitte, NASA, the missile defense agency, and he's an army veteran.
And if authenticity is the key to marketing Jim is the gold standard, his passion for this technology and specifically translating it to real-world value absolutely blew me away. He sat down with Alaris public sector director of sales, Joe Yankle, myself, and my colleague Sarah Jones, you'll occasionally hear our voices off mic. Here's Joe.
Joe Yankle: So RPA starts with usable data, and usable data starts with high-quality images. And today we're going to talk about the relationship and technology between Alaris and UIPath, so I'm going to turn it over to Jim and we're going to talk about how our two technologies complement one another.
So Jim, I'll kick it off to you.
Jim Walker: Yeah, thanks, Joe. So exciting to be here, exciting to be having an opportunity to talk to you guys about RPA and how we partner. I joined the army in 1983 and was a field artillery officer in a fire direction center. And they were first learning how to automate, I walked out to the vehicle where the fire controls were done for their eighth artillery guns, and the truck next to it was the computer.
And over the last 20 or 30 years instead of one computer we now have four or five computers that we walk around with every day. And we have all of this data, we have all of these pictures, we have all of these forms despite the paperless society. And workers all over the world, whether they're government or non-government, were inundated working for their computer.
And what process robotics is going to allow us to do is to take good work like you guys do, right? Good quality imaging, the ability to provide that imaging in a hurry, and to be able to pick it up your multifunction devices, from your scanners, and do something with it.
Today in an office, any office, it comes in something, gets scanned in, and then someone has to type that information in or they have to grab that and figure out what they want to do with it. And that is just necessary work, it has to be done, but it's not something that has to be done by a person, so instead of us having our people work for our computers, let's let our computers work with other computer systems and pass information around.
Joe Yankle: There's been a lot of talk about the president's management agenda, and some of the three key planks - modern information technology, data accountability and transparency, and workforce modernization. And there's also multiple OMB memos out there that all tie back to digital transformation in both of our technologies. So won't you share with the audience how UIPath helps enable agencies to comply with the OMB memos and with the president's management agenda?
Jim Walker: Right, so one I got to give this administration credit. For technology for the last 15 years, 20 years, all you ever heard somebody say was, "Congress and the administration aren't keeping up with the technology, their policies are all outdated." And you've got an administration now that steps up and says, "We want to have automation." And their memorandum 1823 had very specifically said agencies should use RPA and they should look at artificial intelligence.
About three weeks ago the president signed a new executive order, the American artificial intelligence initiative, telling agencies you need to do research and development around artificial intelligence. And he's not saying he wants to do it because it's a neat thing to do, it's not we want to go to the moon just because, but it's more in all of the things we're putting out, we want to serve citizens better. I can bank online, I can call Uber, I can get Uber Eats, I can choose between Uber Eats and GrubHub, but if I want to get a document from the government - I'm getting in my car, driving down to a courthouse, filling it out, handing it to a person who then types it in, and then the necessary work begins, right?
And so in our case at UIPath, started a year ago in the federal sector, 32 agencies are using that software and they are doing everything from bringing in an email, interrogating that email to make sure it has what it needs, and then put it in the ticketing system. No employee in the government says, "I like putting tickets in a ticketing system." So nobody's arguing with that. And you have the IRS who is using this in their procurement area.
The GSA's procurement office is using it when a small mom-and-pop company make a bid on a government contract, they submit their packet. And that packet might have a requirement that the font be 10 pitch only, and in government terms 12 pitch means you're disqualified. 11 pitch means you're ... Because it's 10 pitch and this type of font. The bot grabs that, checks to make sure it meets the standards.
Maybe the standard is it can only be 8 pages, 9 pages disqualified. It could be the best offer the government was ever going to get but it's 9 pages. How about have a bot standing at the front door, grabs it, interrogated for about 15, 20 seconds, sends it back to them and says specifically, "You have 9 pages, you should only have 8. You use this font, you should use that. And because I know you submitted it on time you're now authorized a 24-hour extension." There is no citizen who is going to be unhappy with that.
But if they wait three weeks for the procurement office at GSA to finally get around to it, and not because they're taking their time, but that's their workload, and are told, "You didn't qualify for something three weeks ago." That's not citizen services.
The Navy is using with an F-35 aircraft, they're using robotics and AI, it is such a new thing, and the best way I can describe this is if I were to buy SAP, a word processor, or a calendar program, any other kind of thing - I buy it for that thing, for that function. And what you're going to be able to do with robotics is you're going to be able to say, "I want you to be an HR application supporter for the next five hours. And then for six hours I want you to support your HR finance and procurement in a process that they do once a quarter. And then at night I want you to take all the documents that get scanned in from the downstairs folks and I want you to electronically file them, I want you to build me a report that says what you filed, and I want you to put that up on our website."
That set of arms and legs for all of the work that it's doing, doesn't care who it's working for, when, or where - That's what makes this technology so different. It's not a finance tool. It's not a word processing tool. It's a tool to work with anything that your current employees work with.
Joe Yankle: And speaking of employees, RPA, the intent is to not take away jobs, correct?
Jim Walker: Right.
Joe Yankle: It's to complement the workforce that's already in place, right? So tell me a little bit about George Washington.
Jim Walker: Yeah, so that was fun. So when I've recently retired about 17 months ago, the NASA Shared Services Center which is a quasi for-profit type of thing when you think about how it runs, it needs to break even, but it's a government private company partnership that services NASA, right? So it's not private company or anything, but it acts and behaves that way. And it was exciting to go down to that because normally you have a budget, Congress tells you, "NASA, you have 19 billion dollars," and somebody would say to the Shared Services Center, "You have $500,000 to survive on for the year."
At a Shared Services Center you charge internally for chargeback. So we came across robotic process automation, did all of our due diligence with market research, narrowed it down and said, "Okay, we have a bot." But we realized that the workforce was going to be afraid of the phrase automation. And so when we were trying to figure out what to do, we said, "Let's name our bot," right? And it's going to be the first bot in the federal government. We were literally racing the IRS because they had a program going also. And so we declared it was going to be George Washington. And George Washington has its own email address; it has its own credentials, its own security and everything, all the things that a person has.
By the time we had our program up and running, people would come to the office and say, "Can George do this," right? Not, "Can your bot do this?" Not, "Can the software do this?" And they were asking can it do things that we don't want to do, right? They have to be done, right?
If you are a clerk in a finance department and you have to do a reconciliation between 1,500 lines of a purchase card and the invoice that you have, that work has to be done. But again, it's necessary, but is it necessary for a person to do? The person needs to know when there's a conflict, something didn't reconcile, that's where you bring the person in.
And most of the time now, 1500 lines, I'm tired, I'm ready to go to lunch, there's one that's kind of borderline, I'm going to let it go. I'm going to get to it later, and I never get to it.
Now when you only have to look at the two that are wrong, you're fresh. "Oh, I know why that's a problem. Let me go fix that and run it through the bot again, and then we'll have a zero balance on everything."
Joe Yankle: So let's play this scenario out - I'm an agency, I'm trying to meet the NARA 2022 mandate for implementing automated workflow solution so I can send digital-only files to the National Archives. So I'm an agency drowning in paper, okay? How does RPA help me if I'm trying to scan in boatloads of paper, because I've been housing paper in a file cabinet, I've been paying for storage? How does RPA help me with that scenario if I'm scanning?
Jim Walker: Right. So the good news especially for you guys is, is we don't scan, right? We work with the scanner. So today an agency not only has to have a person that scans for them, but they have to have a person on the other side that's going to do the work for them. And those two aren't doing the same thing, so scanning takes longer, backlog builds up, that type of thing. What if I take the person behind the scanner and move him to the front? And say, "You two, scan for us. Welcome, summer intern, you're going to catch us up," right?
And in behind that every time something scans in George Washington would say, "Hey, there's a new document in my directory. I'm going to grab it. I'm going to figure out what to do with it. I'm going to file it. And I'm going to do ever the person who moved out from behind used to do," right? "But if I can't figure out what to do with it, if for whatever reason the scan comes back and the bot says I'm not really sure about that, it sets it over to the side. It doesn't say, "Hey, that scan wasn't good so I'm just going to file it over here." It says, "You, human, need to be in the loop."
And so you're able to really make a good human-in-the-loop partnership but you're able to reallocate where your staff are. Eventually you'll get all caught up, all your back log will be there, and one of those two people now gets an opportunity to go do something different.
Joe Yankle: So you could set a threshold for OCR read rates at 90%, if doesn't hit the 90% the bot will send it to a human to maybe review it.
Jim Walker: Exactly.
Joe Yankle: To get an acceptable image.
Jim Walker: Exactly.
Joe Yankle: Excellent.
Jim Walker: And that's what we want, right? We don't want the bot guessing, right? If you're an intel officer at one of three-letter intel agencies you don't want to look and have a bot determine if there was a troop moving in Iran last night if the image isn't great. If the image is perfect and it's clear that it is, you want the bot to immediately do something. But if it's not you want it to immediately send it to the desk officer to say, "There could be troop moving, you're going to have to check this out." So it's not ... there is this narrative that you were talking about earlier about people losing their jobs, right?
I bet people said the same thing when the cell phone came in. But there are many more jobs in the telecommunication industry today, but they're not all the same, they're not all laying wires all over the country. They're putting up satellites, they're putting up dishes all over, they're making covers with bling on them for somebody. In my view there's nothing that suggests that technology has always decimated the workforce. It just said to a lifelong learner, "You get an opportunity to learn again."
Erik Oehler: To hear the full interview with Jim, go to our website alarisworld.com under insights or subscribe to this podcast. Coming up next, Simon Wise of Suncorp Group, implementing RPA on the front lines of financial services.
Simon, if you could please introduce yourself and explain what it is you do.
Simon Wise: So I'm Simon Wise, I'm an RPA process analyst, and I work for Suncorp, one of the largest insurers in Australia.
Erik Oehler: So Suncorp's a financial services group, what exactly do they do and what has your journey been like?
Simon Wise: Yeah, so we were originally a bank, and back in the ... it would have been the 90s or something, they bought an insurance company and then we kept acquiring new brands and we steadily grew. We're mostly insurance with a bank attached these days.
Erik Oehler: And what is your day-to-day work entail?
Simon Wise: So day-to-day I go and look at opportunities within the insurance section of Suncorp. And we look at different opportunities for RPA and determine whether or not they're going to be worth our time now to perform. And we also determine what sort of tool sets we would use to deliver that process automation.
Erik Oehler: So was this always your role or did you see the wave of RPA coming and changed course?
Simon Wise: So I've been doing this for almost two years now, but before that I actually was in a business improvement type role, so kind of like a business analyst delivering process improvement projects within the business. But we've been doing up here I think for about three or four years.
Erik Oehler: What were some of the challenges you faced as an organization during that shift to RPA?
Simon Wise: The difficulty was that our traditional strategy was around building embedded automation within our applications, and completely excluding this ability to build more simple automations external to those applications, because the cost to modify them is quite expensive.
Erik Oehler: What was the tool or tools of choice?
Simon Wise: So we use a couple actually, so we use automation anywhere, we use UIPath and we also use Python for very simple automations.
Erik Oehler: I haven't used automation anywhere yet, but UIPath's awesome they just make it so ... it's just so accessible and easy to use.
Simon Wise: Yeah, it's brilliant for getting a user in and actually getting it to be able to play around and build your own automations even at home. When I first used that I built little tasks just for my own workflow at home when doing personal things. And seeing what it offers, people who don't have a huge programming background it's a really powerful software.
Erik Oehler: So what was the impetus for Suncorp's transition to RPA, was it like a reactive measure to a mandate or something they sought out as a solution?
Simon Wise: We were very aware that there was a new code coming in at the time it was a year or so away, where we'd be required to make sure we're paying small businesses within seven business days. And being a large scale insurance company we were finding it difficult to be able to pay in a timely manner, especially when there was complex rules to apply to these invoices.
Erik Oehler: What did that payment cycle look like before?
Simon Wise: It could have been between 14 business days to over 30. And it was not ideal for us, and it wasn't ideal for our business partners, so something had to change. And I was involved in process improvement at the time to try and build some rigor around how we handle this and do it just from a process perspective. But we found that realistically if we use technology we would not need people to do such an administrative role. And most of invoice payment is just data entry and then somebody does a quality check at the end to make sure it's within delegations of authority and things like that, and then releasing it for the processing.
Erik Oehler: Was there any resistance internally?
Simon Wise: Absolutely. So we did have a lot of culture issues that we had to overcome within our business to be able to deliver this particular piece. And it depends on the leadership of those areas as to whether or not they're very supportive or resistant to such change. But we did find especially with the processes that are overseas, so when you outsource administrative sort of functions to another company we found that the SME's from those areas would quite often when you're exploring how this could be automated, they would find new ways to add value to the process to make it harder to automate.
Erik Oehler: How much of this did you build in-house and how much did you have to bring in help for? And like what was your journey into this role like?
Simon Wise: So we had an internal capability, but we didn't have access to any sort of natural language processor at the time. And during delivery of our first invoice process I got coached internally over into the RPA space and became one of the analysts for delivery of the project. But I guess initially I was more part of the business, explaining the problem, giving them requirements, explaining the process, trying to help smooth out any problems with that process to help facilitate the RPA.
And then when I moved across to be part of the team that was delivering that piece, I guess I spent a lot of time working in ABBYY Flex Capture, which was our NLP of choice to help train the models that would glean the data from invoices. And I found that very enlightening for how a lot of these sort of applications work.
Erik Oehler: What were some of the hiccups or stumbling blocks that came up as you implement it?
Simon Wise: So we did find some hiccups in our own internal architecture. So, because we didn't own any of the systems, we had to work with an internal technology team that ironed all the architecture and really work with them to build that relationship to allow us access to the different things that we'd need, like being able to talk to databases which they'd fortified, so that people could not overload them, that sort of thing. As well as getting the information we needed from those systems that we could go and build the training models for ABBYY, as well as making sure we've got what we need so that the robots that go into the claim system can then go and poke the right widgets to make the payment.
Erik Oehler: And what was the volume like that these robots were handling?
Simon Wise: Yeah, so our first invoice payment process managed a hundred documents a day for a particular segment of our business, and then we went and we built the second one which was much more complicated for another area, which does the 2,000 documents a day sort of volume. And I guess when you can turn around an invoice within a day, you're ensuring that for one your suppliers or the service providers aren't chasing you for payment because you're getting the invoice and typically paying it within a day. So by the time their accounts receivable people are having a look at their bank accounts and checking those remittance advices, you've already paid. And it makes their job a lot easier. I think as well it means for them they don't have to employ as many people to chase you down for money.
Erik Oehler: So one of the stigmas is that RPA will replace jobs, and what was your experience like with Suncorp in this regard?
Simon Wise: For some of them, so when we looked at our first process we thought, "Oh, we're going to find so much value here." And we ended up removing the need for any offshore support for that first process. But for the onshore people that were also paying these documents it did just release more capacity for that customer service perspective.
Erik Oehler: Wow, it's like the ideal case, isn't it?
Simon Wise: That's right. And for the second process which was much larger, we did actually reduce offshore capacity. I don't think we impacted our onshore at all, because they didn't really do too much except for supporting the offshore. But we did find that when you're using these complex natural language processing applications you do quite often need to have a ... and we found it was about 10% of the staff that we had handling the verification of those documents. So if you had say 10 people paying these documents, you'd still need one to do the verification in ABBYY, to just help ABBYY with funky fonts, and sometimes the scan might not be great, so being able to just correct any misread data and things like that to make sure it flows through nicely to the process.
Erik Oehler: So though I didn't intend to we've stumbled into my company's territory with document scanners, can you speak to the accuracy of the data through in these paper-based processes, like what kind of error rates you experienced?
Simon Wise: Yeah, so I guess error rates when it comes to documents can be subjective to who you're getting your documents from. We find with the smaller process because we don't have any faxes whatsoever, the scanners are quite good at getting the data with minimal support from people. It really comes down to how you build your project within say ABBYY or whatever to only get what you need. So if there's anything additional that you're adding on top because you might want to down the track ... it actually adds that opportunity for something that needs to be then checked by a person, which takes away that benefit.
But when it comes to faxes, for example, faxes add a new element of pain because they get grainy, they sometimes are sideways and things like that, and that's generally where you'll get more problems. But yeah, I don't ... I wouldn't have exact numbers on scanning issues.
Erik Oehler: So in your cases, are documents scanned into a hot folder and then RPA takes over from there?
Simon Wise: Basically, yeah. So the documents come in and they get ingested by our own internal services that then triage the document to our claims file storage. So each claim kind of has its own document library, that sort of thing. And then what happens is we have a small application that pulls these new documents, and it looks for the applicable document type that we're after, so invoices, so our ingestion can classify.
And when the service detects that sort of document with the right criteria, it then triggers a task in which we have a small script that downloads those documents, as you say to a hot folder in which ABBYY then ingests and applies its rigorous rules against. And from there that then gets output to a database, and then we'll run another series of rules over that data just to decide, "Is it payable? Is it not payable? What's the reason it's not payable?"
And then it then creates a task for the robot that goes into the claim system to pay it. So it'll either log in to the claim system and do the non-payable action, which is quite often leaving a file note and assigning that task to a human, or then going in through the payment wizard and processing the payment, which then just goes to an automated system that generates the check or the EFT payment.
Erik Oehler: What types of paper are being handled related to the claims you process?
Simon Wise: So we don't do any paper claims anymore. It's everything, anything paper that comes in, we scan it, and attach it somewhere, and then destroy the original. So it can be anything from an invoice to an assessment report to a quote, it could be a letter just providing information, like a statutory declaration or a witness statement, medical certificates. There's a range of documentation we receive.
Erik Oehler: How transparent has this effort been? Like, we mentioned the stigma around this before, and I know some businesses are going as far as to name their robots to make them more readily accepted by their human co-working counterparts.
Simon Wise: Well, we're really upfront about our robot's presence, so for our robot users their name is always robot and then whatever we've named it, so it could be like Robot John or Robot Whatever, and so they'll see these robots working alongside them. But from a solution perspective it's still magic for most of them. But we are currently doing a sort of like an internal internship where we bring business users into our space, teach them how to use UIPath and get them to build their own process, and then the idea is that once they're done learning they then go back to their business space and look for small RPA opportunities to then continue to make their lives easier.
Erik Oehler: And how far along do you view Suncorp is being on this journey to scale across as many business functions as possible?
Simon Wise: I wouldn't say we're anywhere near halfway, but we're definitely on the right path. And looking back on the sorts of things we were automating this time last year we've radically evolved and become more capable 12 months down the track, especially with the new technology that we now find ourselves with access to, so things like ABBYY, a lot of our team members are learning Python and becoming more and more capable with Python to do document manipulation that doesn't make sense to use, a full heavy RPA application to do like modifying a Word document or Excel. You can have something run super duper quick and just generate what you need, and then the RPA process continues from there, so conjuring Python.
But I guess when it came to looking to automate we look for the highest return on investment for what we've got. So we've got processes that we could automate all over the shop, but looking at the things that give us the best bang for buck with the lowest complexity is generally how we decide what we attack next.
Erik Oehler: How are the financial services landscape keeping up? Like, are your competitors doing similar things and if they are do they tell you?
Simon Wise: So some of us do talk around the industry, and it's not just within financial services, but the service providers outside of the actual insurance companies themselves are also doing automation. So we've noticed that some of our service providers provide our documentation in a way by automation so they don't actually send it to us anymore themselves, and it's manipulated in a way that suits our automation.
But our other competitors are definitely in the playing field as well. You see them at the different meetings and at training sessions and things like that, so I ran into some of our competitors when learning how to use ABBYY. And when you talk everyone kind of tries to keep their cards to themselves, but when you're talking about particular topics you can kind of tell what they're doing and how far along they are. But those that are building automation into their applications as opposed to using RPA in the way on top of the user interface quite often they don't have much to say about it, because it's just software development I guess to them.
Erik Oehler: Are there any broader industry trends that you see coming in the RPA space?
Simon Wise: So I think tying in more and more machine learnings definitely, like a slow build up. Not a lot of people are doing it. And not a lot of people are able to show where they are doing it really well. But I think with these applications being out there, so UIPath being free to assess and automation anywhere, while not as easy to use is now free for small business, you'll find that a lot of smaller businesses may contract in RPA developers and business analysts short term to build solutions to help them out as well. I think we'll find that there'll be less and less administrative functions where it makes sense a more human interaction.
Erik Oehler: Simon, thank you for your time. It's been a pleasure, and I look forward to seeing what kinds of innovations Suncorp puts out into the world.
Simon Wise: My pleasure, Erik.
Erik Oehler: Thanks again to Jim Walker of UIPath, you can check out their incredible software at uipath.com, as well as Simon Wise at Suncorp who you can connect with and follow on LinkedIn.
Data Dives can be found wherever you download your podcasts. If you like what you heard please subscribe, share, get it out there. If you have an interesting data story about you or your company and want to be a part of the podcast, please reach out to us at Alaris World on Twitter or LinkedIn. This and every episode of Data Dives would not be possible without the support of Alaris, a Kodak Alaris business, delivering scanners, software, and service that can lead your company's digital transformation. Find out more at alarisworld.com. My name is Erik Oehler. Thank you for listening and we'll see you next time.