#86 Let's talk about RPA vs. AI w/ The Automation Guys

Welcome back to another episode of the Process and Automation Podcast with the Automation Guys. Yes, you all know we have a very busy community with plenty of discussions on all things process and automation. And recently was a couple very specific questions on two acronyms you probably heard of in the past. But yeah, we thought it would be really good to talk a bit more about PA and AI, so robotic process automation and AI. So that's, that's what we like to cover today, looking at this, this very innovative technologies and definitely look at the points of where these technologies can work together and or if they can work together and how they can work together and where it makes sense. So yeah, as you know, robotic process automation and artificial intelligence, both offer quite revolutionary prospects for businesses. And each of these technologies comes with a unique feature set and very specific applications. So yeah, you probably have thought, like many others, these two things are interchangeable— robotic process automation, AI— but there are definitely distinct potentials in these technologies and reshaping shape operational dynamics of sort of any business. So, um, so their individual merits, uh, and combined, uh, have a power to revolutionize the sort of complete industries. Um, just looking at insurance, financial services, yeah. And, um, however, to harness that full potential, it's crucial to, to really drill into these two things and demystify them nearly and understand the, the core differences and and understand where these two technologies stand and how they maybe work together. So this is what we want to do today, unravel these very specific technologies. And yeah, let's drill into the first one. So what is RPA? Arno, do you like to share that to our audience?

Yes, thank you, Sascha. Good intro there. And I think that there might be a bit of confusion sometimes in what's the difference between robotic process automation and artificial intelligence. And quite simply, robotic process automation merely refers to software bots or software robots that can emulate things like repetitive rule-based tasks that was previously carried out by humans. So think of it as a digital worker that never sleeps, it makes zero mistakes, obviously provided the robot is running and it's actually programmed correctly, and it performs tasks at really impressive speed. So it is sort of like a digital worker. And the key characteristics of robotic process automation sequences are really that they are developed to be rule-based. So that means that RPA bots function based on predefined rules that we teach them to execute. It's akin to actually training another person to do specific tasks. And this is no different. There's sort of this training element. The second characteristic of RPA is the fact that it's what we call non-intrusive. So RPA bots operate at a user interface level. So there's, there's no need to alter any of your underlying systems. And this could be your legacy systems, your modern systems, your SaaS systems. It can actually touch the user interface and interact with it similar to what a human would do. And then the third characteristic, which is really important one, especially when you look at high transactional automation, is the fact that RPA is very scalable. So RPA can manage increase in task volume by actually taking advantage of more bots doing the processing. So there's that kind of elasticity built into it where if there's more demand, then what you can do is you can just deploy more bots. If there's less demand, you can take those bots and could deploy them elsewhere. So I suppose that again, it is a bit like having a workforce and actually deploying a workforce where it's needed. The most. So perhaps you've got kind of a month-end finance process that's running. So if there's a lot of transactions in that process, what you would typically do is you will throw more bots at it. And in the middle of the month, you might use less bots and deploy them elsewhere. And, you know, I'll quickly touch on sort of other applications of RPA. Obviously, just mentioned that the financial sort of month-end report, which a lot of that happens inside businesses. And, you know, to produce something like that, I think that report generation is a good example where RPA bots can be deployed. So that's a really good example of the application of RPA where you have to generate either month-end compliance, quarterly reports, and the robotic process automation actually emulates what humans will do to produce that report. So it will go and get the data, it will aggregate the data, it will actually create a presentation, and it can even email it to members that need to have access to these reports. So people in finance, for example. Another common application of RPA is, of course, data entry and data migration. So if you've got a legacy system and you've procured a brand new workflow system, for example, that you want to migrate all of your cases and your data across, you could use RPA to do that. RPA is very good at actually logging on to a system, actually reading the screen and taking that information and keying that into your new system or various other systems. Now, this is not just reserved for data migration and data entry type scenarios, you know, in, in all elements scenarios, rather, including the report generation. RPA is very good at actually going to different systems, reading the screen, and going to other places and actually entering the data in there. And I think that's why data migration is actually quite a, a, a key application that we see RPA is used for. The next one is forms processing. So forms processing is merely where we actually receive forms with inside our business. It could be either a form that somebody fills out, an email across to you. It could be an online form, for example, that somebody submits to you. It could be a paper-based form, the form that's scanned and OCRed. So robots are very good at actually doing repetitive type tasks on these forms. Now, later in sort of the episode, we'll go into a bit more of the intelligence side, how to be root forms. But if you have, let's say, a census and there is a lot of forms, you know, hundreds of thousands of forms that needs to actually be read, the data needs to be saved in a structured format with inside a database. Rather than actually employing a sort of an army of people to do that, what you'll do is you'll just teach it how to actually read and process the form data. The next example, and again, this is a finance example, is one we see a lot, is automated invoice processing. Of course, with the accounts payable, accounts receivable scenarios within AR or inside finance. There are an awful lot of sort of mundane activities, high transactional kind of activities that's quite expensive. And we can actually train bots to look at invoice that is received, for example, do some of the initial background checking on the invoice and then process that invoice and perhaps you know, like a human counterpart. If it, if it founds or finds an invoice that's perhaps over a specific threshold, it can just route that to another person to sign that off. So you can kind of like have that human in the loop element. And then kind of the last 4, I'll just sort of quickly touch on those, is things like payroll processing, end of month. Again, you know, a lot of things to get through in terms of kind of doing all of the elements around Payroll processing, automated credit checking. So we can go to third-party websites to actually, as part, let's say, from a loan approval scenario where we can actually look up from a third-party credit rating agency to kind of get a credit rating. We've got things like order processing. Orders arrive, we need to actually push those orders through their normal processing lifecycle. So there's elements there which is robotic, and we want to actually just automate that using a bot. And again, things like feedback aggregation, you know, where you've got survey forms, feedback arriving into inboxes, and you actually want to just aggregate all of that together in a very structured way and get that into the database. So very, very good example of, you know, application of RPA.

Nice one. Nice one. And yeah, so the next one, as we said, this is all about RPA and AI. So the next one is AI, artificial intelligence. So So AI refers to the simulation of human intelligence processes by machines. So that's put very, very simply. So it's a wide range of subfields as well, from neural networks to natural language processing, NLP. So that's maybe another buzzword you, you heard. So which enables machines to perceive, reason, and act in ways previously thought reserved for human cognition only. So it really gets into all these areas. Yeah, and the key characteristics of AI, so we obviously know very, very prominently ChatGPT and other sort of AI technologies which popped up massively over the last couple of months. So yeah, but let's look at the key characteristics of AI. The learning ability AI can learn from data and improve over time. So that's definitely what we have seen a lot recently. Decision-making— AI can make decisions based on data analysis, so going through lots of data and come up with the right information or is the right decision. Yeah, it is, it's clear with the amount of data we have, it will definitely be much quicker than what humans can do with speed and accuracy. So it can go through millions of records in, in, in hardly no time, and we wouldn't need like years to do it, and we wouldn't spot everything because at some point we are so bored and tired to identify those things. So yeah, and the adaptability is a key characteristic of AI. AI can adapt to, to changes and can handle a lot of that unstructured data out there. Yeah, and then when we look at the applications of AI, we all deal with it probably on a regular basis with Alexa and with our iPhones and Siri. Voice and face recognition is a very prominent application. Sort of the next big area we all deal with is chatbots and virtual assistants. We see very different versions from very simple to very sophisticated on lots of websites, integrated in, in shopping systems, into the appointment booking on websites for customer service. So, so many applications for chatbots and virtual assistants out there, and we are just at the beginning. So that, that would be very exciting to see where that all develops to. Predictive analytics, a field which which is around for quite some time now, by looking at this massive set of data, come up with the right information and the right decisions and determining trends, those kind of things. Content recommendations, definitely what we have seen was ChatGPT. We just come up with a little bit of content and we get plenty of ideas from the algorithm. Fraud detection, very, very important one for financial services, personalized marketing. So it's very interesting. So instead of receiving sort of bland emails and standard emails from your sort of vendors and shops and wherever you interact with, getting into something personalized will be so much more powerful. Yeah, translation can be done very, very quick with AI. Sentiment analysis. So all very, very prominent uses of AI, and we see that everywhere already. So it's really nothing what is like sort of up there, we don't know how it works. It is already in application and Yeah, that's pretty much it on that part.

So I think it's probably worth just highlighting the kind of the key differences or the core differences between RPA and AI. And the way I usually think about RPA and AI is the AI is the brain and RPA is the hands. And I think that, you know, Both of these technologies drive efficiency. It also brings scalability and, you know, and it is considered to be sometimes sort of the forefront of innovation, you know, across industries that relies heavily on data processing and digital and tech. And, you know, there are some overlapping kind of characteristics between AI and RPA, you know, within the tech ecosystem, because they're not the same thing, you know, and I think it's worth just sort of just looking at those slight differences. So the first difference is the complexity of the tasks that AI and RPA can complete. So, RPA is very good for repetitive mundane tasks, something you could train it to do, and it does it over and over again. Where AI, it shines when it needs to handle complicated tasks requiring sort of thinking and decision-making. So again, you know, one is the brain, one's the hands, sort of train the hands to do specific things, but the thinking element or making decisions, you know, that requires a sort of a deeper level of cognitive ability. And this is where AI takes advantage of machine learning, deep learning, and all of these things to be able to feed the RPA bot what to do next. The second core difference is implementation. What we see generally is that RPA implementation implementations is often quicker as it works really on a surface level and it mimics kind of user actions where AI requires deeper sort of integration and, you know, often more demanding and excessive data training because, you know, AI uses data. We train these neural networks using a lot of data so that takes a bit longer. We need to have a bit more data. It's not procedural in nature. And setting up those models can take a bit longer compared to sort of just creating a robotic sequence, which is just sort of a step or a set of steps. The third one is learning capabilities. So again, as you can expect, RPA does what it's programmed to do, and AI can learn from data and it can evolve. So sort of self-learning, kind of self-correcting, you know, those are all kind of attributes that's strong with inside AI. And then the 4 core differences is really at the scope of deployment. So usually RPA operates in a kind of structured rule-based environment where, you know, we can map things out on a flowchart and we, you know, we know what to do next and it's easily mimicked by a person with not a lot of training, whereas AI really thrives in both those environments where we have got structured, you know, 1 1 2 environments, but also unstructured scenarios where you might have an anomaly, and it can actually handle those ambiguities with ease. So it's not the same thing. It works well together. It can also work in isolation, if that makes sense. And I guess, Sascha, the last point to discuss is sort of just the symbiosis of this.

Absolutely. Yeah. What do we really get when we bring these two technologies together? So that's That's the next thing I will talk about. So while these technologies have very distinct characteristics, so yeah, RPA and AI aren't really mutually exclusive. So when you combine these technologies, it can amplify the automation within the business. So you mentioned that, Arno, earlier, so there's like RPA and AI are sort of the— or maybe RPA is sort of the arms, and then AI is maybe— yeah, it's the brain, isn't it? The brain and the eye, the eyes. So sort of that kind of stuff. You combine it, um, yeah, you have lots of potential, um, to get more stuff done, right? Um, so it doesn't, it doesn't just execute tasks, it understands and learns and makes that decision. So that's, that's, that's what happens when you bring these things together. So let's dive very, very briefly into a real-world example. So a company handling customer support might use RPA bots to scrape incoming emails for specific keywords, and the emails identified as complaints could be directed to an AI system, which is then— the AI system is looking at it, understands the sentiment and the content of it, can then categorize it and prioritize it for, for some human agent. So it will find out someone is very angry, it can find out if it's a cancellation, and those kind of things are really important aspects.

Well, that's a very simple example, but I think there's a lot of inboxes out there where lots of emails arrive and helpdesk@company.com or complaints@company.com. So I think that The, the RPA bot is the thing that actually reads the data, scrapes it, and then that gets passed to the AI system then to interpret that and make the right decision, even if it's just an initial vetting. It could be that somebody has complimented our service, and we might just wanna write them a complimentary message back. Now, you don't need a human to do that. You can train an AI system to do that. So again, it's a deliberate sort of simple real-world example, but that's a good use case for it, especially when you have got high volumes of transactions that you don't necessarily want to deploy a lot of human resources to actually analyze those and deal with those. And I think the last point I want to make is kind of like looking a bit into the future and kind of trying to sort of portray or to convey to our listeners, you know, how do you make the right choice when it comes to RPA and AI? And, you know, it is quite a pivotal decision because it might be that it sounds like you need an AI system, but also perhaps in scenarios you need to, you know, you want to rely on, you know, your requirement is more stronger for just RPA. And it can be quite a daunting decision to make for any business. And, you know, what I can say is that both of these technologies does present transformative opportunities, you know, to businesses. And, you know, a lot of it really utilizes on the kind of like the specific business you're in, your specific scenario, you know, and you want to kind of get an initial place where it's sort of easy to exploit it. and you want to then scale it. So, so, so the point really, when you kind of make a decision on is it RPA only, RPA and AI, AI only, it's really you need to look at your needs, identify your needs first. You know, if your primary initial requirement is, is just automation of repetitive tasks, you know, without sort of altering your, your systems, you know, you've got preexisting systems, you don't want to change it, you just want to train something, a piece of software to be able to interact in, then, you know, RPA is the way to go. But if you need sort of more insights, predictions, adaptive decision making, then, you know, then you need to look more towards AI. The other sort of point to consider as well is return of investment. RPA usually offers quite a quick ROI. Due to the fact that it's implemented very quickly. You know, you can use a sort of RPA studio to develop a bot fairly quickly. AI, on the other hand, can provide sort of long-term strategic advantages. But, you know, it can be sort of take a bit longer to develop. But, you know, that could entail or, you know, in return sort of create new revenue streams for you, enhance customer experiences. And I guess that the other element of this is really sort of just to stay informed, you know, go out there, listen to podcasts like this, go read, you know, this tech landscape is ever evolving. You know, it's, you know, there's a lot of experts out there. You know, again, on this podcast, we've got the Automation Guys, and a lot of people ask us questions about these things. We try and answer all of those. You know, other ways is to attend seminars and really sort of just stay up to date with these two key technologies, because I do think they are truly transformational. And, you know, on the one side with RPA, it's quick to get started. And then the other side, once you mature your RPA story, you could also then plug in your AI into your RPA. So there are various options. It's not necessarily sort of a big initial investment of time and effort to get both of these things deployed at the same time.

Yeah, absolutely. Yeah. As a conclusion here, the debate shouldn't be about RPA versus AI. That definitely should be, should be about the understanding how these two technologies work. And then yeah, in combining them, I think that, and potentially combining them so you get the full power of automation wherever you want to use it. So yeah, the Automation Guys, we talk a lot about these technologies. So we have lots of episodes just about RPA, lots of use cases just about AI, and then also we talk about other other things. So yeah, listen back to our previous episodes. And then yeah, getting, getting hands-on on these things is also very important. As Arno mentioned, staying— stay informed, experiment with things, and what is the possible, what's nearly, and see what's the art of possible. We discuss a lot of these things here in our community, theautomationguys.net. You can find a sandbox you can start using and experiment with these new technologies, um, and find out about different use cases and get your hands dirty. So yeah, whether you're starting at the moment or you just like to get hands-on with these technologies, visit us on theautomationguys.net. And yeah, we will be here, the community will be there for you. And yeah, thanks a lot for listening to today's episode. Hopefully that was useful. Yeah, if you like that episode, leave us a nice review on Spotify, iTunes, wherever you listen to that episode. Yeah, it would be highly appreciated if you send us some feedback, if you have some ideas on other episodes, other questions. So very happy to pick that up and make this topic here in the podcast. Yeah, thanks again for listening to today's episode. We will be back with another episode very soon. And until then, let's automate it.