SUBMITTED BY DataServ

Artificial Intelligence and RPA — what's the difference and how do they apply to Accounts Payable? There are so many different technologies that can be used to improve processes. No matter what your AP processing looks like now, automation, the reduction of paper, faster approvals, taking advantage of payment discounts and cost savings...all can help. You just might not be sure where to start.

In this webinar, DataServ experts will cover:

  • What is AI
  • What is the difference between AI & RPA
  • How AI can enhance AP processing

Click the link below to view the webinar, or scroll down to read the transcript.

View the Webinar

Webinar Transcript

Tom: Good afternoon good morning, depending on your particular time zone. I’m Tom Fischer, and I’m Director of Marketing at DataServ and I want to welcome you to our webinar today. Thank you so much for joining us. Artificial Intelligence is talked about a lot. In many ways when it’s talked about, it’s unclear in terms of what it is or more importantly how it applies to accounts payable or purchase-to-pay. Our webinar today is going to help you with all of those things, so we are very glad you’ve joined us. I’m going to turn it over momentarily to our speakers, but before I do, just a quick housekeeping note. All attendees are in listen only mode, but you have a question at any time, you’ll see the ability to submit questions via the go to webinar functionality and you also have the ability to do chat. So at anytime feel free to do those, and I’ll be keeping an eye on it. You’ll be hearing from me from time to time. We have a couple of poll questions, I’m going to jump in. But otherwise, I’m going to turn things over to our 2 speakers today, Drew and Jeff. Please take it away.

1:10 Jeff: Thanks, Tom. It’s Jeff Haller from DataServ here. I’m here virtually with our Director of Technology, Drew Gillow. Hello, Drew.

Drew: Hello.

Jeff: Welcome. And we’re so excited to talk with you today. We originally did this presentation a year ago or so at our annual Confluence event with a bunch of our clients, and it was very well received. So we decided to do a virtual version now that we are all remote workers and keeping up with our social distancing. So I’m not technically in the room with Drew, but we wanted to share some of the research that Drew has done and continues to do into this technology that is somewhat confounding as Tom mentioned. We hear it thrown around a lot, as you do, too. It’s hard to know what is what, and what is not. We thought Drew did a really good job of bringing this down to the everyman's level for those of us who aren’t as technically savvy as Drew is. To share a little about Drew, he’s been at DataServ for four years or more now, and is really in charge of the entire tech stack here including the research that we do and the development that we do on software. We found ourselves doing some artificial intelligence in P2P some 13 or 14 years ago, and this is therefore for us not something new. We found we were taking it for granted a bit that people understood what we were talking about and this presentation was designed in an attempt to simplify and explain that somewhat. Hopefully you’ll enjoy it. Tom is going to ask a bunch of questions, because if you’ve been in a presentation with me before you know that I pick on the audience a lot, and I will try to do that today even though I can’t see your faces as I would in person. With that in mind, Drew, do you want to get us started with some definitions and perhaps we’ll all learn something.

3:10 Drew: Yep. Thank you, Jeff. So we’re trying to help people understand what’s really going on in the world. So for the past few years, experts have been telling us that the top digital transformation trends are analytics, AI, machine learning, and chatbots. So everyone’s talking about RPA and AI. You see it at conferences, the consultants, and probably senior management at your companies.

Tom: Thanks, Drew. I’ll take it away. We mentioned that we were going to have some poll questions. It’s certainly OK if you are unsure of what these things are. That’s why you’re here. But if you know which of the following tools you currently use in your AP process, you can check multiple answers here. I can see that the votes are coming in, and as soon as we have a few more I will publish the results, but this will give us a sense of who is on the line and kind of a before and after approach and will help us guide with our content. So I’ll give you 10 more seconds to get that done if you haven’t voted. I know some of those words are complicated but hopefully they won’t be after a few minutes. And I’m gonna close the polls. I am sharing the results. Drew and Jeff, it looks like a bout half of attendees, the big one is analytics is currently being used. Next up in line would be algorithms and process automation with about one fourth of the folks on the line. Then the numbers dwindle a little with machine learning at 17%, RPA at 17% and then chat bots at 8%. I will go ahead and hide these results and get back to your presentation, Drew.

5:00 Drew: Thank you. We may get some different answers as we explain some of these definitions. So are you confused? Certainly, the people I see writing the articles are. There’s a lot of terms out there. I’ve read marketing materials from big consulting firms that misuse the terms. I’m not sure if that was on purpose or they really didn’t know themselves. So my goal is to define the terms, provide examples that hopefully are real world that you recognize, as well as examples that utilize DataServ’s cloud based accounts payable automation solution. So what do you think of when I say artificial intelligence? Well, actually AI is a catchall term that means software that does what humans can do. So this can be as simple as a calculator to as complicated as a self-driving car. AI just means your computers doing anything.

Jeff: Computers can do a lot. So I think that’s a good one to use here, but it’s funny that people think this is something more specific.

Tom: We just tapped into the very beginning of what AI but we have another poll question. We’re asking which of these best describes the majority of the AP approval process. There’s 4 options and they are progressive in nature, so let us know which more accurately describes your scenario. This one’s a little easier to answer than the last one. So I’m going to close the polls here in about 3, 2, 1. I’m going to share the results. About half the folks on the call today have partially automated workflow system, and then after that we’re looking at email is used and mostly paper based at 17%. Congratulations to the 8% of the fully automated workflow systems on the call today. Drew, I’m going to turn it back over to you.

7:20 Drew: Alright. So to give an example, DataServ AutoVouch is a good example of software doing what humans can do. Take an invoice from the top of the stack, go find the relevant purchase orders and receipts, automatically retry for a number of days since invoices often arrive before the materials are received, and then match the lines by product/quantity/unit/price. Compare the prices and quantities. Is it matched within tolerance? If so, automatically approve it. If not determine who should resolve the exception and automatically route the invoice. So really a bunch of automation processes, some machine learning and some not, but really looking at the total AI perspective.

Tom: Thanks, Drew. I promise you’re not going to hear from me for a while after this, but we do have one more poll question. Based on what you just saw with some of the matching process. We’re curious how many of you have an automated matching and exception handling workflow process, or are you doing this manually? If it’s automated great, if it’s manually, there's no shame in that, but we’re just trying to get a sense. And if its some sort of combination, that’s good for us to know as well. And if you don’t know, you don’t know. So we’ve got options there for you. And I will publish these in 3, 2, 1. We can see the results now, everyone on the call. Almost half are using a manual process, in terms of matching and exception handling workflow. That’s the most popular answer. Then another third are doing some sort of combination. It’s a pretty small number. Only 8% are currently automated in this time. Jeff and Drew, are there any surprises there?

9:10 Jeff: I do think it’s always hard to understand. I like the “don’t know,” Tom, because I think there are a lot of people out there who aren’t thinking about how people do this today. There’s just people who do this, and it happens every week, every month. That’s what we run into a lot. Some people don’t realize that you can automate these things. I think it’s getting better and more able to do that, and Drew’s going to talk about what those technologies are that are allowing us to automate some of those human-centric, complex tasks.

Drew: Thank you. So getting into the terms that we saw on the intro slide. Algorithm or process automation? This is software that performs a repeatable task. This is what software has been doing since its creation. Some easy examples that you are familiar with: text search in excel, or pricing in your online shopping cart. In this scenario, a developer programmed a specific set of steps. For example, order total equals item total, plus shipping, plus tax. Within accounts payable within the DataServ automated solution, we have a number of examples of this. An easy one would be searching. Searching for documents that meet a specific criteria. Moving into the next term: machine learning. This is software that learns. This is what people really think about when people are talking about AI. But there is a broad range of what machine learning means. Most commonly it is software that is taught or trained by developers and then used. It doesn’t continue to learn. A developer used some technology to create the original software, and the developer doesn’t actually really know exactly how the software does what it does but it’s happy with what the results are. More complex that continually teaches itself after the initial training. So within DataServ, the Autovouch invoice matching process that we talked about earlier is an example of software that is periodically improved with new knowledge. The most complicated version of machine learning is software that teaches itself and automatically learns on its own. This is probably what you think about when you think about AI. One of the most prominent examples of this is AlphaGo, which was built by Google. AlphaGo is software that taught itself how to win at the game of Go. Back in 2016, it defeated the winner of 18 world titles. And the next year it defeated the reigning world champion. For those of us not familiar with go, it is NOT Pokemon Go, it is a 2500 year old strategy game. Go has very simple rules, but is more complicated to play than chess. I bring up chess, because in 1997, IBM created software that defeated a grand master chess player. It won by playing out as many moves as possible and choosing the best. This was not machine learning, it was just a very powerful computer running through an algorithm extremely fast. It’s not possible to do that with Go. There are more possible board configurations in Go than there are atoms in the known universe. So a different approach had to be taken. In the match pictured here, the reigning champ said the gap between humans and computers was getting too great. He would play against humans, and treat software as a teacher. So a few years ago, AlphaGo taught itself to be the best in the world in just 40 days. Why isn’t AlphaGo running everything? Well, it took 40 days and $25 million in hardware. Not to mention all the software development costs that went into it ahead of time. Much more than the $25 million. The only thing it could do was play Go extremely well. And even more important as I said earlier is Go has 10 simple rules. How simple is your job?

13:20 Jeff: Seems like a lot more if/then statements in our job. Especially in AP, especially in matching. I also like to bring up the fact that invoices have a lot of variability, but there is eventually a limit. You mentioned earlier that Go has more possible outcomes than atoms. That’s just a massive number, and that’s really where the idea of machine learning is really required. Anything less than that doesn’t really need to incorporate such expensive technology for learning more as it goes. It should incorporate updates and experience, but it doesn’t have to be self-teaching in order to take care of the universe of known possibilities of invoice capture. Because there’s only so many characters in the world on invoices, and there’s only so many invoice formats in the world. It is limited, even though it seems like a very large number. It’s proven to be limited. I think we’ve got a good combination of artificial intelligence without needing to go as deep into machine learning as some of these extreme programs have.

Drew: And it’s teasing the puzzle apart. AP is many steps. So one of the processes that we have machine learning involved at DataServ is identifying and removing logos and non-invoice attachments from email. All these vendors send an email, someone sends it in and there’s a logo at the bottom of the email. That’s not necessary or helpful in the process. We’ve trained a machine learning algorithm to go and find 99.8% of those logos and strip them out.

Jeff: And this was a perfect example of how it does help a lot, because they do vary tremendously. And we really do not want to get to know the set of logos. In this case that data is too variable to want to know it. We identified some simple criteria, and that simple criteria applied to 99% of the logos so we said that’s something that we should have the machine learn from and just build up its own database over time. Plus we had a lot of volume going through this, so some 3.7 million or so every month. There’s a lot of volume that’s been able to teach this, so that’s another good example of how to use those techniques.

Drew: And then diminishing returns. Does it need to get smarter than that? How much more time should we invest in getting smarter?

Jeff: There’s only so much we need to know about these logos.

16:00 Drew: The next term I’ll look at is analytics. I think 50% of the organizations represented said that analytics was something they were already using. Analytics is software that turns data into insights. For example, profitability analysis. Which products, customers, or markets (or a combination) are profitable and which are not? The Society of Information Management surveyed 800 CIO’s. Analytics was the number 3 IT priority after security and alignment with the business. So after getting a job done, the number one project.

Jeff: I was fascinated the 50% of our attendees today have some form analytics, but Drew wouldn’t you say that there’s always more to be analyzed.

Drew: Yes, there is. It takes us into our next poll question.

Tom: Thanks, Drew. We heard earlier that about half of you are using analytics. Sometimes it’s another story to know how easy or how hard those are. We like to get a sense from those of you on the line, how hard is it for you to access analytics around AP. We’ve got a couple options for you. We ask that you pick the best answer for your situation. I will go ahead and close the poll and share the results. So it looks like we have about 42% say they have access but it’s not great data. Another 25% says they have access and it’s great data. Loving all the people who said I have no earthly idea. You are in good company, I hope that makes you feel better because a quarter of the people on the webinar today feel that way. 8% of you are not feeling the love from the IT team. We are not going to beat up IT on this call since Drew is part of IT and there are probably others as well. So, Drew, back to you.

18:15 Drew: Thank you folks. So within DataServ, we provide a robust analytics solution with standard analytics that our clients use to continually improve their process. Looking at how they’re working the process, how the exceptions are being handled, and what are the causes of those exceptions. As well as the ability for our users to create their own dashboards and find their own insights by visually exploring, drilling, and filtering the data. Moving on into specific cases of analytics, next one you’re going to see is predictive analytics. So that’s software that predicts. An example that is familiar to many of us is Netflix. More than 80% of the TV shows and movies that people watch on Netflix are discovered through the platform’s recommendation system. So Netflix is predicting what we want to watch and we watch it.

Jeff: Or dictating. Whatever it is.

Drew: Yes, they are providing suggestions, and we are choosing to act on those.

Tom: Sometimes awkwardly, right Drew?

Drew: Yes, that can be the case. But they actually market in ways to generate people to look at what they want them to see. They pick different images to display on the screen based upon demographics and other things. I was talking with a CIO of a manufacturing company. They have about $115 million in revenue. And he was looking into predictive analytics, and he was quoted $45,000 a month for a tool with no models built, not trained, and was going to be given the privilege of being an early adopter of this tool, which was why he was getting a discounted rate. A significant percentage of revenue would be going into that predictive tool. An example at DataServ, we use historical information to predict which invoices are potentially duplicate. The next topic we want to speak to is big data. You hear a lot about that on NPR and other places. People want you to think big data is different from analytics. And there are definitely differences but those differences are in the technology and the approach to solving the problem. So that’s something IT has to worry about. From a business perspective, it’s just analytics on a different set of data. So an example of where big data comes into play would be R&D of plant breeding. They use DNA along with lots of data collected from greenhouses and field trials to make decisions on which varieties and hybrids of corn, lettuce, broccoli, what have you to produce in our R&D pipeline. For example, corn has 32,000 genes. Humans have 20-25,000. It’s interesting to me that we haven’t figured that out. But we haven’t. But corn and many plants have more genes than people do, so it becomes a very big data problem for them to figure out all these trials and which products to produce. But honestly, invoice processing in AP is not big data. Even if you have millions and millions of invoices, and you track all of those documents, it doesn’t result in a big data problem. So I don’t have an example of big data analytics in AP. The next item would be Internet of Things (IOT). IOT is software that captures the status and history of real objects. Complex objects with lots of data capture can result in both a big data problem as well as being IOT. So for example, the Boing 787 collects data on all of the systems, from the landing gear to the flaps to the engines. Over a half a terabyte of data is captured on a single flight. So this is both IOT and big data. Obviously at DataServ we don’t make things. We don’t have any physical airplanes flying around, but you would be surprised to learn that DataServ has been involved in IOT all along. We provide a digital representation of an invoice, the metadata of an invoice document. The workflow path, the audit information about the changes that have occurred within the life of the document. You may not think of that as IOT, but that paper invoice was a physical document and you’ve tracked all of its history of its life from when it was received through its whole life cycle. So DataServ has been doing SaaS since before there was SaaS and IOT for longer than there was IOT. And why do I say that? It’s because consultants create new terms to repackage old ideas. And we’re gonna hear that again later in this presentation

23:20 Jeff: You know when RPA first came out, I remember getting that question a lot from people. What is this? Our research looked an awful lot like scripting language from the green screen days, the 80’s. Turns out that there’s a lot of that that goes on out there.

Drew: Yes, you are correct. So augmented reality, another big phrase out there, you see a lot in VR and video games. But you also have people trying to sell that in the business space. So augmented reality is software enhanced reality. You’re asking yourself what does Pokemon Go have to do with AP processing? Pokemon Go is an example we are all familiar with. But it does have a place in AP processing. So DataServ’s digital mailroom quality assurance review is a real example of augmented reality that provides business value. Our QA operators see a digital representation of the invoice document, augmented to identify where data was captured by the invoice processing machine. So this expedites the verification and correction of data in case there is a QA issue. Next is chatbots or conversational platforms. Chatbots are software that understands language that is written or spoken. Some examples are Siri and Alexa. You may also run across chatbots on websites that collect information from you and prompt you with default answers to questions that you have before getting you to a real person. Today DataServ does not employ chatbots in our platform, but we have had conversations about is there a potential to do instant messaging or slack based invoice approvals, or even using smart speakers to do invoice approvals. Getting the work to where people are working.

Jeff: It’s coming. I think if you look around your house, everybody’s got on of these smart speakers today, or 4, 5, or 6, like we do. And you get used to asking the questions verbally, and being handsfree while you are doing something else. We have felt like there might be a place for this in P2P, especially when you talk about higher level execs who are doing approvals mostly from mobile devices. It might be easiest if they got a text from the system. More secure, perhaps, than email even. For those reasons, I think this might be coming, Drew. How about you?

Drew: Yes, I see this on the horizon, for sure.

Jeff: Wouldn’t you say there are lots of tools out there to make this happen, too. That’s another reason that we might want to look for it. It feels like the Amazon’s of the world are giving these tools away. In a lot of cases it’s open source. It’s probably easier to build than in the past.

Drew: The real question is, are the users there to use it.

Tom: If I could interject, gentlemen, for those of you who are on the line if you are feeling a little overwhelmed yet, one of the things we are going to send you later on after the call is over will be kind of a quick recap or a grid. What are the categories of AI that we’re talking about? What is a real world example of it? What’s an example of how DataServ sees it being used in the AP space? So pay attention, but I just want you to know that’s coming.

27:15 Drew: There is a quiz at the end. The next one Jeff jumped into a little early, which is Robotic Process Automation or RPA. Sometimes called bots. Don’t confuse an RPA bot with a chatbot. That’s part of the reason for discussing chatbots earlier, because they’re both called bots sometimes. Someone says, we need a bot, you should ask them, are you talking about a chatbot or an RPA bot? Because they do very different things. RPA is software running software. So imagine something is remotely controlling your computer. Your mouse pointer is moving and words are being typed on the screen without you typing. Assuming that it is software doing this, not the help desk or a hacker, that’s RPA. The promise of RPA is when you combine machine learning with RPA. However, Gartner says that current RPA technologies are just scripting tools with little to no intelligence. So no machine learning, no intelligence on what they’re doing. These are really just the same tools that IT departments have used and struggled to maintain to do automated user interface testing. Again, vendors are repackaging old ideas with new terms, trying to sell them. We have seen people trying to replace automation and integration that they already had with RPA because they had initiatives from management saying spend a lot of money on RPA, it’s the next great thing. So get rid of the tools you have that already work and replace them with RPA. So they spent lots of money, millions of dollars, and had little to nothing to show for it. RPA has issues whenever the user interface changes or an exception arises. We know of a company that uses RPA to login to vendor websites and download invoice images. They have over 50 RPA developers maintaining those scripts because the websites are changing. They have popups appearing. Every year that somebody says, are you sure you don’t want to get this thing mailed to you? All of those popups interfere with their scripts. RPA can have value in certain cases. It has value in repeating high volume repeating tasks that are manual, where the user interface is unlikely to change and exceptions are infrequent. For example, open every email in a mailbox and save all those attachments to a folder. Another good use case is entering data into legacy applications that don’t have API’s. For example, an old school mainframe green screen application that you have that people aren’t making changes to but you just can’t get rid of.

Jeff: I think they’ve been used in a lot of ways, Drew, and that’s really what the current environment looks like to us. We’ve got probably 5 or 6 current projects from companies that are clients of ours that have tried this in various different capacities inside AP, including matching. They’ve come to us because they’ve tried that and it did not work. We’ve got a ton of stories around that, if anybody’s interested in hearing more, please reach out to us and we’d be happy to share. I think to some extent it doesn’t mean it’s a bad technology. I think a big part of what Drew’s talking about today is you’ve got to understand what applies to what problems. What solution you apply to which problem. That’s really the challenge. So much more technology is a great thing. We have lots more technology, but the application of the technology has always been the trick to getting the value. We’re hoping we can give you a sense of where these tools apply best.

31:10 Drew: Exactly. RPA is good at eliminating high volume repeatable tasks. DataServ does not use RPA. We eliminate high volume repeatable tasks using algorithms, machine learning, and other forms of AI. If your organization has goals to implement some of those technologies, we are already delivering many of them to our clients.

Jeff: I love this diagram being gears of different sizes because that’s really the nice thing about SaaS. Because if you are working with SaaS like DataServ, you end up getting on these technologies without even realizing it. Some of our clients don’t even think about the fact that we have this machine learning, the feature you showed earlier that weeds out logos from invoice images for example, that’s a tool that we added without any cost or testing or big learning curve for the clients or big IT project. So we keep evolving and our hope is that we can bring you as clients along with us, and prevent you from having some pain of applying the technology to perhaps the wrong problem or the wrong technology to the right problem. Those are real world issues that I think we help solve.

Tom: That takes us to our last poll, Jeff and Drew. After what you heard today, do you think your AP process can be improved with any of these areas? Let us know which one you think is most advantageous for you to focus on. We’ll let the results pile up here. I see the votes slowing down a bit so I will close the poll and share the results with everybody. The leader of the pack would be algorithm and process automation. That seems to be where the most potential might be with attendees on the line. And that would be followed up with RPA, as well. Any surprises there, Jeff or Drew?

Jeff: No. I think this is what we have all grown up to learn. I think it’s nice to see the interest in process automation. That’s always been a very strong thing. On the RPA side, it’s great to see people interested and again find the right application and it makes a lot of sense. We did hear a couple of stories where the approach to how those tools were implemented and utilized was very intelligent. That came down to having the process owner own the bot, the RPA bot, rather that what would typically happen which is that it becomes an IT project, an IT burden, because it’s pretty scientific, pretty technical. But if you can get a tool that is very easier for a user to adapt, that tends to be a better alignment for making sure it’s applied to the right process. Because the people who do the work and own the process, they know the details there and can make a better judgment call, whether this a bot or a person. Some shared services organizations we work with globally embrace this as a tool set for their shared services processes in AP and AR, in particular. And it’s been fascinating to see that. I think this is all continuing to evolve. I’m not surprised as the analytics at zero. It seems to be like an analytical tool, a tool to figure out how you are doing rather than a tool to automate.

Tom: Drew, I’m going to turn it back over to you, I think we’ve got one more slide. As a quick reminder, if you do have a question, I know we are nearing the end of our time, but we do have time for questions. So if you have anything you want to ask Drew or Jeff, feel free to submit that in the question section of the go to webinar functionality that you are using.

35:40 Drew: Thank you, Tom. Really just to recap, trying to take the mix of words that was at the beginning and hopefully put them in some organization and hopefully they make sense to you now after this presentation.

Tom: Drew I do see one question. You mentioned AI and that growing smarter as a new client. How long does it take to learn us?

Drew: Sure. There’s a lot of different steps. The biggest aspect for you would be our invoice processing machine which is the understanding of the invoices that come from your clients. It’s pretty seamless. There’s some onboarding that happens as a new client where we ask you a few questions about how you want the outcome of your data to look. For example, utility bills don’t have an invoice number and everybody has a formula for how they want to generate that, so it’s really about us knowing how you want that to be formatted on the outside. But it’s pretty straightforward and seamless and behind the scenes to learn your invoices.

Jeff: That’s a great question. I’m glad you read that one, Tom. It’s fascinating to watch, actually, but the biggest thing that we brought to the table when DataServ developed the invoice processing machine probably 11 years ago now, we really drove from the idea that we should reduce the pain that clients felt when they were onboarding. Because if you look back on OCR projects from the past and other similar things from the past with AI and machine learning, there’s always this giant learning curve that’s very painful, and it takes longer than you think, it takes more effort than you think, and you have test more than you think. So our goal was to reduce or eliminate that. I think we succeeded at that with this product, this invoice processing machine. The digital mailroom takes on a lot of that cleanup and learning process, that feedback loop that goes back into it. It does tend to ramp within about 3 periods, 3 months or so. You won’t notice the pain, the behind the scenes pain that DataServ’s team goes through is over in about 3 months. The clients don’t actually notice it because we are absorbing that with our team internally doing the verification and cleanup work, and that’s one of the things we try to bring to the table is to get you out of that pain of automating and get you straight to the benefits.

Tom: Good. One says, another question online, can you tell us is there any next tech trend that we should watch for in AP?

38:40 Drew: That’s an interesting question. I think if you read the literature, they’ve been talking about RPA for 3 years now, and it’s really maybe coming to where it’s supposed to be in certain aspects. But I think at this point it’s still about machine learning and RPA and the reality that companies that are not driven by analytics and the data, at some point they won’t be companies.

Jeff: The evolution is the part that’s scary to me, Tom. That’s a great point you are making, Drew, and the fact that it’s just evolving faster and faster. What we’re hearing from the marketplace and from clients is the next big thing for them is to get where they’ve wanted to be for a long time with digital. We’ve all talked about digital, and digitizing and digital transformation for maybe 10 years or more. Probably 15 years that we’ve been talking about it. I think the COVID crisis along with the economic impact of that has caused some speeding up of evolution. If you look at history, that does happen. Every time there is a recession, certain companies are weeded out or eliminated for NOT doing these things, whereas other companies really get on board quickly and really make the changes necessary. I think the answer might be to take advantage of all the technologies that are out there and to get as far as you can on that curve of evolution because if you don’t it’s a threat to your wellbeing as a company, and your profitability, and your staffing levels, and your ability to work remote, and your agility. To me, I think next would be the voice and text kind of approvals are going to happen. I think also this idea of just getting caught up on as much automation as you can to be digital across the spectrum.

Tom: We have one more. Can you cover that example again of Internet of Things? I’m not sure, I don’t think they were clear on what that meant.

Drew: Sure. The IOT is about tracking all of the information about a physical object. So the airplane is the example that has lots of parts and it tracks all of those pieces. GE actually makes the engines and they track their own data on the engines so they can predict when a blade will fail before it happens. That’s the intent of the IOT data, is to get ahead of maintenance or manage issues. What analytics and what insights can you gain about the device. In our example of IT, we track all of the history of the invoice. Who touched it? When did they touch it? How long did they hold onto it? In theory, one of the things we’ve talked about internally in predictive analytics would say can I predict that this invoice that just arrived today will be paid late because it goes to a 3 approval step process, and the 2nd or 3rd approver, Bob, is always slow. We just got it, it needs to be paid in 15 days and based on the history of similar invoices going through the system, unless somebody starts to expedite it now, it will be late in the future.

Jeff: That’s a good one, Drew. I think along with that we’ve seen some push in Europe, in particular, with the European payments council for things like QR codes or sophisticated bar codes on invoices that tie into the payment that allow the bank to get remittance data all the way downstream in the payment cycle. I would say the internet of things has not done everything that it can do. We were very early to adopt this philosophy and apply it to images and documents but I don’t think it’s done yet. I think it’s going to keep evolving, and I think it goes back to the source of the invoice at the vendor. If they were to attach to an invoice a bar code of some kind that would identify key values, that could be consistent throughout that whole process. Really could keep track of where is that invoice? Is it still in customs? Is it on the ship? Where is the product? You could tie back information like that about logistics and supply chain details that are separate from the invoice. We’ll see where that goes. I think that’s going to happen.

Drew: I think what’s driven a lot of internet of things applications is the ubiquity of the internet. So a thing, your car, is driving around the road, and in theory could have a cell phone connection and send data to the manufacturer whether you like it or not. And so it could send the data, and the manufacturer or the person you leased it from could be maintaining your schedule and making sure that you’re doing all the right things. There’s insurance companies that are tracking whether you are a safe driver from your phone, if you allow them to do that. That’s an internet of things on your driving.

Jeff: That’s a great example. I don’t like that idea at all. Someone once said I drive too fast. My insurance company does not need to know that. I hope they don’t know that. They might be listening right now. I think that’s true, Drew, and there’s that connectivity within the supply chain as well that needs to be tied in here. Hey, Tom, any more questions?

Tom: I think we are at our time. I don’t see any questions online. None were submitted. You have our contact information, so feel free to give us a call or send us an email if you have any questions. But as follow ups, we will be sending everyone here a recording in case you wanted to listen to certain parts again, so you will have access to that. And then we’ll send you the recap grid that talks about these different types of automation and AI along with examples. So those things will be forthcoming this week, and again we are all thankful for your time. I will let Jeff and Drew close it out, but I will say that we are planning another webinar on July 1. Look for invites for that. We are going to cover the matching process, the 2-way and 3-way matching process that occurs in the AP function. So with that, Jeff and Drew I will give you the last words.

Drew: Thank you for attending.

Jeff: Thanks, Drew. We really appreciate it and we’re so happy to hear from you all. Please keep the questions coming. We are happy to share more as we know and we’ll probably end up doing this on an annual basis at our Confluence event for clients, but I think we also ought to share this out to the rest of the world because it is good data. Thank you, Drew, for sharing.

View the Webinar

Subscribe to our Blog