RPA Spotting: What is Robotic Process Automation?

 

Lloyd Dugan:
This is Lloyd Dugan from BPM.com. Tonight's guest is Cam Wilkinson, a friend and a colleague from Down Under. This is also concurrent with an effort by Cam and I to package together a presentation at the upcoming IRM UK Business and Analysis Conference Europe 2021. That is happening in September 20th - 23rd. Our session is early morning on the 21st, which is a Tuesday.

Lloyd Dugan:
Robotic Process Automation, RPA. That has been the focus of the conversations that Kam and I have had in general and in our work at Serco in particular.

Cam Wilkinson:
Well, thanks very much, Lloyd, and appreciate you having me on today and looking forward to going through some of the hot topics that we've been discussing over the last couple of years together. The experience and exposure that I've had here in Australia around the automation and the aspects of how large and small organizations kind of digitize and and transform with some form of automation, whether it's the the new world of RPA or new-ish world. And I was just reflecting, as you were talking about, some of the history and the the background that I have in automation. And I guess it kind of stems back to the days when I got involved with computing, as it was in my teenage years. And, you know, I'm in my fifties now. So that was that was a bit of a stretch. And we were connecting computers to telex networks. This was before the advent of fax machines, before the worldwide web was kind of invented. And so we were automating the connectivity of systems back in the in the 80s. It was to kind of manage the Treasury operations for some of the large broking houses across the globe. So that was the kind of start of RPA for me. And ever since then, I've kind of been in and out of that field over the last three decades. So it's been fun. It's been really good. And the aspects of the current kind of challenges that we see with regard to process automation in the world of outsourcing for large government contracts like what we've been discussing with Serco and their clients is quite similar, but it is much more humanistic aspect to things nowadays. So that's that's the really juicy bit that I like and how we can ingest the human centric design concepts with whatever it is that we do as we help organizations transform.

Lloyd Dugan:
The human centric design is a term that you've introduced to me, and I have come to understand it better in our conversations. The topic of the paper or the presentation, rather, is called RPA Spotting the Art of Looking for RPA Opportunities. Business analyst conference is more about business analysts. They're not that technically deep, but they have to be savvy enough in the in the usage of technologies to know when they should be applied and to and get the right people to help make that happen, even though they may they're probably not part of the technical implementation team. So the presentation is really built around that level of comprehension about RPA. Even having said that, though, it's still pretty dense and it's on the advanced track at the conference. The term RPA spotting was potentially chosen by me because one of my more subversive favorite movies is Trainspotting. It introduced us to a bunch of people we would recognize nowadays. And it has some really hilarious moments and some pretty gross out moments. But it's an excellent film. But anyway, it's about a practice that's apparently a thing in England where people gather on hills with binoculars and notebooks and logs and spot which trains are coming down the tracks. Kind of like bird spotting, just different. So the theme was we have to have that same kind of nod and a wink perspective when we look for opportunities. It's not a hard science yet. So we have in the presentations about suggestive moments for people to pursue. And I'm going to say that is the framing device for what we're about to get in to as far as the podcast is concerned, is how do we see what RPA does and how do we describe it in ways that can be more easily accessed by the larger business community? Because if it just becomes another bit of technology to deploy, it will suffer the short lifespan and burnout focus of every other technology that that's come out to try to answer a need the business doesn't really understand.

Lloyd Dugan:
So to start that off, we'll just say what is what is robotic process automation? I'll take a first crack at that and then I want to hear Cam's perspective on it. To me, RPA is a software robot, meaning that it is a programmed entity. It's not a machine, a physical machine. But it is a software machine. And it is programmed to execute certain instructions. So how is that different from any other software? What it's programmed to do is things that the human being would otherwise do. So if if there is a job, a system maintenance job, for example, that requires a human being to log on and launch something and then mother it while it's working and then close it after it's done. That's a waste of that person's time, a robot can be designed to do everything I just described. That's a very simple case. But that is a particularly easily accessed example of an RPA software robot moment. More complicated ones are when you mix that with other things going on, like maybe I'm a user, I'm working some kind of transaction processing moment through a series of screens. And, you know, at some points I'm going to make a reach out for robotic help. And again, the software robot will answer the call and do some work for me then. Otherwise I would have had to take a lot of time to do.

Lloyd Dugan:
You can extend that further to situations like virtual users where the angle is I'm going to sign on as a virtual user, as a software robot. I have been instructed what to do with this screen, just like a person would be trained to do something with the screen. I as the robot will work that through until its conclusion. You could do that as well. That's a more complicated kind of RPA. But it's it's also what's out there in terms of first gen.

Lloyd Dugan:
The summary statement to characterize first gen or first generation capabilities of RPA is that they can range from session emulation - the last thing I talked about - or automated step execution - which is the first thing I talked - about or anything in between, which is the second thing I talked about, which is repetitive, routine and time intensive with only deterministic rules in play. In other words, there's not a cognitive evaluation moment by the robot. It's not the it's not deciding good versus evil. It's just it's deciding going left or going right. As long as we have those kind of conditions than first gen RPA tools exist to answer our need in that regard. Now, I'd like to flip it to Cam in particular to talk about what he thinks next gen RPA may bring to the table to make it more interesting in the next go around.

Cam Wilkinson:
When I first heard the term RPA, it kind of intrigued me. And then as people started explaining what it did is like - but that's just really simple. That's just so it's like a macro running on a screen scrape, which is something that has been around for 20 odd years. And I guess that's the conundrum here, is that we've got these capabilities of typical emulation software or automation software routines that just do things that they're programmed to do. Then there's the world of AI and robots, the kind of expectation that, well, if a robot can do this, why can't it think better and act better and do better and faster as well as a human? So that to me is the next gen is why? Why can't a bot do more than just a copy paste function or an automation routine? And that's the kind of questions I like to keep asking when we're talking about deploying RPA is how can we actually improve the process as well? So one of the concepts that we came up with for a deployment was using the RPA to record all the activities and steps and processes and tasks that have occurred and build a track, a running track and a running record of the activities which we found in the insurance field has been fantastic because it provides a level of governance and understanding and for quality assurance purposes. It just gives a much greater level of assurance and understanding of what decisions have been made for, for instance, a claim assessment. So, yeah, I think there's a whole realm of areas that RPA is going into now. People are extending and pushing the boundaries and tying in machine learning routines and scripts to to kick off and and to have a more thoughtful response rather than just the typical copy paste type functions that RPA was initially deployed with.

Lloyd Dugan:
In the earliest days I had a similar reaction, which is, gee, this looks a lot like screen scraping, but I also thought it looked a lot like what is usually thought of automated functional testing, which again, is session emulation. It just prompts a bunch of test data. And so the thing is, with screen scraping, it was a similar thing, but it was much more limited in its focus. It wasn't designed to repeat and extend. It was just designed to scrape and use. That technology had been around for a long time. So essentially the vendors took something that had been there for a while and essentially repurposed it for a new set of use cases. The functional testing angle, though, I don't think I've seen as much recognition of that connection, but I think it's there because the functional testers would essentially create testing. They would script the execution through the user interface workflow. And that's essentially what you're doing with the RPA design. It just just without the testing. So, again, you know, there's the technology, but it was used for testing purposes. And here is being again repurposed for something that can be repeated and extended across a much broader set of situations. So, again, a new kind of use case. So the what's old is new again kind of aspect of this, which I'm pretty sure is easily comes to mind when you hear this stuff. Is it to me one of the more interesting phenomena I've observed in the IT industry in the last couple of decades.

Lloyd Dugan:
The other thing that I want to know ask again in return is. Where do we see? Where do you see the applications of machine learning and artificial artificial intelligence in particular? I'm going to set you up for one which doesn't bind you to pursuing it, but I need to give an example to when I'm trying to get at here. If there is work, for example, and it's a virtual user that's doing it, and it comes across a document that has been imaged in with data extracted through optical character recognition. So now it's in the record somewhere. And we may actually have different versions as we may have some of the data on record already. And now we've listed something of the document. You're trying to match it up. So there is software that's been around for a while, and the identity management space that's able to take the data that's been extracted from the document, compare with what's on file or one record, and then determine within a confidence interval. Is this the same person? Because a lot of the times you sample transposition errors or just that they're human beings have made simple mistakes. That doesn't mean they're not the same person in order to make this work. These these this class of software has a knowledge base of such situations that give rise to these statistical probabilities, as well as some kind of heuristic engine that, again, makes inferences based upon the qualifications of the differences that it's told are there. And then it returns an answer that says that's a big difference since we can't tell you the same person or that's a small difference. We can assure you that's the same person, that kind of thing. So that's a small, very specific example of a kind of limited and artificial intelligence. It's really just the heuristic engine with the noun paired up with the knowledge base applied to do that.

Lloyd Dugan:
Why is this important for the robots? Because in first gen, it cannot think for itself, right. It cannot sit there and evaluate. John Smith vs. Johnson Smith without you programming into it, the very kind of heuristic engine properties that this other software already has. So there's other software will then return a deterministic result that says this is the same person or it's not the same person, and the robot can react to that. Given today's first gen capabilities of RPA. But one would imagine that at some point, maybe the RPA doesn't have to go through the extraction process in the way of describing it. It might be able to read a document, especially if it's been Full-Text indexed to determine what information is there and then needs to be compared with what's on file. And I could see a moment like that happening in the not too distant future. So along those lines what else would you say, are use cases that bring machine learning and artificial intelligence into better play?

Cam Wilkinson:
The key thing I reckon, that RPA brings is orchestration and management. And thinking back to your comments about functional testing and the automation and this whole macro screen scrape history in the background of what is old is new again, one of the biggest value adds, I think, that the current RPA platforms offer is that ability to do a management of the data pipelines. So the flow of information, the flow of data, so collecting it, whether it'sdoing a screen scrape or monitoring an email inbox or web form submission. So the trigger events that occur are then a consequential flow of activities and a stream of decisions that need to be performed. And and ideally, as many of those automated and creating a capability platform or a capability engine that can do many different things. And I reckon, yeah, OCR of in bound documents in Australia. We've got a Know Your Customer requirement, a regulatory requirement for opening a new account with a financial institution. So there's a heap of crosschecking, like you mentioned, that needs to occur is the same person. Can we validate that? There are three points of reference crossmatch when we are looking at that process, you know, human doing it, they're they're kind of going through a set of prescribed tasks where they make judgments and assessments and then they to document those and and create a point of view.

Cam Wilkinson:
We're relying on so many skills and sensory capabilities as humans to perform that. And I think the RPA platforms are about combining those in a digital sense. So whether it's that the comprehension of this person, the knowledge that we have of them from a lens of geographic logistics. So being able to sequence those together by using standard rules and then applying more probabilistic approaches. So some of the things that we've seen that are quite useful is around the building of a knowledge base in the world of AI and machine learning. Some of the really difficult areas that we are helping clients with at the moment is around looking at unstructured documents, unstructured text and creating categories and hierarchies and linking those documents and the concepts that are contained within them in a way that they're capable of being searched through from a database. So by doing that, you're creating these taxonomies, I guess, which then simplify the whole process of navigating and mapping a decision. And I think that kind of aspect where a robot acting as an orchestration engine can be checking one database against another database, against another rules routine or a rule formula. If those two things match, then it continues and executes the next steps in those areas. Probably from my perspective, the next gen and the the the really useful aspects of automating decisions. That's what RPA is for me.

Lloyd Dugan:
It's an excellent answer. Or just starting with that kind of thinking here at the work in North America, because up to now, the bots have been more limited in scope. But the more that can be asked of them to to do multiple queries, to associate the results of different queries, all as part of this orchestra, you know, all under this kind of orchestration orchestrated sequence, the easier it is, right. Because we take care of a problem that we'd otherwise have to to resolve later.

Cam Wilkinson:
I think one of the surprising aspects of the work that you guys do over there, surprising for me was the scale and this and the scope that you had deployed into. So it's the management of all those activities at large when you've got hundreds of bots and hundreds of people all interacting together, and you've got to somehow keep the flow of data moving along and know at a point in time where things are so that you can kind of pick it up if something falls over, if there's a blackout or if someone wasn't able to submit something or the bot wasn't. So it's that kind of tracing and tracking is quite a task and being able to manage that. That in itself is is quite an effort. I guess there's there's different applications of RPA and it depends on the industry and and the size of the deployments. But they're they're going to be adding value in different ways, some of them just purely because of the the bulk of the work and the time saved and others because of the enhancement with the type of decisions that are being made.

1000 Characters left


Lloyd Dugan

Lloyd Dugan is a widely recognized thought leader in the development and use of leading modeling languages, methodologies, and tools, covering from the level of Enterprise Architecture (EA) and Business Architecture (BA) down through Business Process Management (BPM), Adaptive Case Management (ACM), and Service-Oriented Architecture (SOA). He specializes in the use of standard languages for describing business processes and services, particularly the Business Process Model & Notation (BPMN) from the Object Management Group (OMG). He developed and delivered BPMN training to staff from the Department of Defense (DoD) and many system integrators, presented on it at national and international conferences, and co-authored the seminal BPMN 2.0 Handbook (http://store.futstrat.com/servlet/Detail?no=85, chapter on “Making a BPMN 2.0 Model Executable) sponsored by the Workflow Management Coalition (WfMC, www.wfmc.org).