hello and welcome to today's webinar today we're going to talk about how you've been lied to about your ap automation solution my name is matt wright i'm responsible for customer success here at appszone and i'm excited to talk to you about how artificial intelligence is changing the way that organizations are applying automation to their accounts payable process so let's start and talk about what is the state of the ap workforce today we've had a couple of interesting years that we've lived through what is that translated into in terms of how our teams are feeling and performing today it's been some recent research conducted at the beginning of this year iofm conducted a survey of accounts payable professionals and what what what the data tells us is that our ap teams are pretty distressed um on the top right here we can see that organizations are not really providing a clear career track for ap professionals so it feels a little bit dead end and we have you know if we look here in the lower left 50 52 of the respondents said they're they're feeling burnt out just flat out say it straight up admit it i'm burnt out and you know what's what's going into that is you know the number of hours that we're asking of these ap professionals to work so in this this this particular survey 67 of the respondents are working at least 41 hours per week so almost uh you know uh two-thirds are reporting that they're working more than a standard hour week and 23 almost a quarter of the respondents are working more than 50 hours a week that's a lot to ask of anyone and and when we come over and ask them you know why why are you feeling burnt out no surprise 46 percent of them say they're working too many hours and you know the other reasons you know it's it's difficult balancing work and and family when you work that much and and it's just understaffed when you don't have enough people and you always feel that you're behind it really just it's stressful and it takes a toll on people and and it's hard to go to work every day so what's that translating into you know for for our teams as they you know wrestle with what coed has done to their business life their work lives so understaffed high hours no career advancement no work life balance that leads to burnout right which is what we've seen and when when when individuals are burnt out they they retire they move on they resign and in accounts payable that's really really expensive because all of the institutional knowledge that sits in the heads of those ap professionals that do the work every day it walks out the door with them and that's really hard to replace and it takes a lot of time and has a big impact on ap processes so let's dig in a little bit and see how artificial intelligence can help organizations have a better ap operation so this is what commonly organizations would uh illustrate as their ap process and they would you know most organizations i think would say that they have some level of ap automation in place yes we we do ap automation but is this really ap automation every one of these blue boxes usually represents some sort of interaction from a human to help with these these automated solutions make this decision who who is who is this which business unit does this invoice belong to which supplier am i am i assigning this invoice to which vendor id from my vendor master do i assign how do i code this invoice which cost center i got to help the the automation apply the the two in the three-way matching and then you know do we even have visibility into duplicate payments are we getting an invoice that was already paid for by some other channel of spend by a p-card or an expense report so applying human effort to the ap process even in its you know automated state relies heavily on human involvement that's really because the biggest enemy to accounts payable automation is how organizations rely upon the institutional knowledge that that the individuals the humans have in their heads that they use every single day to process invoices so let's let's talk about what i what i mean by institutional knowledge and what that what that manifests to uh for accounts payable processing so here we have a snippet of a vendor master file and i bet a lot of you can relate to you know what this looks like in your vendor master file right where we have um this this this supplier called east repair incorporated and in this one vendor master record we have seven different instances of this one supplier east repair and they all have different names some of them don't even have different names here's east repair here's east repair but they have different vendor ids right so when an invoice comes in from east repair it's up to a human to figure out which one of those vendor ids do i assign to that particular instance of an east repair invoice so here we have a stack of invoices from east repair and here's that little snippet of the vendor master um that that the human is presented with and you know in in traditional ap automation ap automation vendors will say to you yeah we we help you resolve the vendor id well they they what they do to resolve vendor id is they apply rules-based logic but it says if the name is east repair inc then the vendor id is 987-223 but what happens when that rule comes up against an ambiguous result where there's more than one possible answer the rule breaks down and it can't perform and that's what a lot of ap automation and i use air quotes there when i talk about automation that's what a lot of automation historically has delivered for organizations is rule-based processes that work fine until they don't work when there's ambiguous results which is really common in accounts payable so in this particular case we have the the human who sees this invoice they look at this vendor master id and we've got all of these invoices that came in and they've got to figure out which vendor id gets assigned to it what these humans do they look at the context of that invoice they look at all of the details on that invoice they look at the ship to location they look at the products on the invoice they look at a myriad of different elements on that invoice to figure out what is the right vendor id and they draw the right conclusions so this particular one was this vendor id and this right and so on and so forth and you can see i've i've attempted color code here which ones line up with which vendor id but we rely on humans to resolve these ambiguous results and and rules-based automation simply cannot solve these types of problems and what's really tragic about rules-based technology is we have this this this human who who knows all of the context they they've processed these invoices for many many months sometimes many many years and they know oh for this particular east repair invoice i know it's this vendor id so they have this beautiful nugget of institutional knowledge that sits in their brain and they they pass that that piece of institutional knowledge through the keyboard they digitize it it turns into usable information and it gets assigned to that invoice transaction it makes all the way through the payment and then as soon as that payment is completed and that invoice is processed that beautiful piece of institutional knowledge that we extracted from that human's mind and we digitized we throw it away we don't reuse it in traditional rules-based technology that's kind of the tragedy of historical ap automation solutions is we're never able to learn what the humans know and that's what artificial intelligence can help us with and that's why i believe that ap is the perfect problem for artificial intelligence we have a series of answers and a series of questions and we're marrying those up and we have a mechanism for retaining those correct answers so to compare autonomous finance um with traditional finance approaches there's a you know a couple of things that are really different that i think are worth pointing out here so workflow processes um are really the way that traditional finance has been automated and again we should use air quotes when we talk about automation it's when i encounter this particular situation and i don't know the answer trigger a workflow and route it to this department or route it to that person who can solve that with artificial intelligence that routing process the information that gets applied when it gets routed it gets recorded so while it might need to be routed in its first instance it will be recorded so that it can be automated in the future so that self-learning self-healing aspect of artificial intelligence is every time we we route a document to an individual to ask them to answer a question what we're doing is more than just process the invoice we're making an investment in automation because that answer gets recorded it doesn't get thrown away like i showed on the previous slide rule based decisions right they they they go to if then else and if that else results in an ambiguous result then we got to rely on that human to answer that that particular scenario so if we go back to that vendor master situation we've got the the clear rule that says if it's this name or if it's this id then then it must be that vendor but if it's ambiguous we can't get to that answer but with artificial intelligence we're constantly learning and looking at new patterns and adapting the predictions in in in in the in a way that doesn't require somebody to make a configuration adjustment it doesn't require somebody to write a line of code it doesn't require an investment from your your vendor that that solved that uh that that delivered that technology for you it's doing this all on its own traditional finance automation is really only effective when there's structured data when it's unstructured data it really struggles with autonomous finance solutions we're leveraging the latest in technology for natural language processing computer vision all of these technologies combined with artificial intelligence allow us to extract data very very accurately in both unstructured as well as structured data scenarios when we have rule-based processes if the rule needs to be adjusted we have to ask somebody to come and make an adjustment to that rule come and write a line of code come change a configuration issue a change order through my software vendor with artificial intelligence in autonomous finance these rules are automatically adapting and learning to what the human inputs are as it makes its way through the process and it's always retained artificial intelligence doesn't go on vacation artificial intelligence doesn't resign right the the changes and the adaptations made by artificial intelligence are retained with traditional finance automation a lot of times we ask suppliers to do something different when they send us an invoice with autonomous finance it's supplier friendly we don't need the suppliers to make any changes the technology will adapt to the constantly changing fluid environment that a supply chain represents so what we're kind of left with here is a really stark difference between the old way of doing things and the old way was really just a year or two ago and the new way of doing things is what we're what we're bringing to the market in the past you know 12 or 15 months is really excited exciting artificial intelligence based autonomous finance so let's just kind of talk through the evolution of technology and remind ourselves where we are so in the beginning of of accounts payable processes this was all paper right literally pieces of paper moving around the organization and being written on and then ultimately hand keyed into an environment we we evolved from moving paper around to really cool technology where we actually scanned the pieces of paper and then we keyed them from image and then we were able to archive those images right that was a pretty big advance for us and then ocr came along was like oh well if we've got images we can apply optical character recognition to read those those invoices and we all know the pitfalls of ocr especially with inputs that are as unstructured as ap invoices are they just simply cannot adapt we have to build templates and adjust templates and it's and it just doesn't get the kind of output that that we were led to believe it would when we first looked at the technology and now that artificial intelligence is really you know the important um uh lingo in the market uh these these legacy vendors are saying gosh i've gotta i've gotta make sure that i'm bringing artificial intelligence to my products so these ocr vendors take these these tools that they've built over you know the last really series of decades and they bolt artificial intelligence around the edges of it and it certainly makes that ocr perform better than it did before but at the end of the day it really is a traditional ocr tool with some ai bolted around the edges of it but what we're talking about today with absent in autonomous finance processing is an ai first approach where the technology was built from the ground up with artificial intelligence in mind that gives us a huge advantage in our ability to deliver productivity that was never possible before and it really starts with this shift that we're making in in moving away from these human-centric workflow processes which are rules-based and rely on a variety of individuals to provide inputs into the ap process and a lot of those individuals don't work in accounts payable or they don't even work in the finance operation right so we're outsourcing the work of accounts payable and finance to other parts of the business procurement and buyers and vendor master teams have to get involved to resolve exceptions through these cumbersome workflows and heaven forbid one of those transactions gets stuck in a workflow cube that somebody is not attending those invoices age right and we have upset suppliers and the supplier calls the buyer and that's frustrating for the buyer and then it lands in accounts payable and it's late and then it's a fire drill to get that invoice processed right that's kind of a day in the life of an ap professional is always wrangling the the the workflow processes that are not optimized so with ai first autonomous processing we can get to a point where invoices arrive in accounts payable they're processed in accounts payable and they get paid by accounts payable and our reliance on workflows and other members of the organization are are drastically reduced and minimized to just the bare number of exceptions that we need their help with so how do you become part of this shift to an ai first ap operation let's talk through a couple of those examples really starts with the foundation of this this these prediction machines that artificial intelligence represents and if we if we just take a second and break this this kind of complicated picture down we can see on the left is a whole bunch of inputs coming into this prediction machine there's a lot of different data that gets uh used when we're processing accounts payable invoices there's vendor master data invoice documents purchase orders all flowing into this prediction machine and each one of those data elements represents a pattern right so as an invoice gets processed it creates a processing pattern artificial intelligence observes those patterns and it makes a prediction and depending on the level of competence it may or may not uh need to present that particular transaction to a human to confirm whether the prediction that they made was right or wrong if they do need some validation they present it to the human if the human says yes you get it right that goes through a feedback clip when we tell the prediction machine good job you get it right if it's wrong and the human makes an adjustment to that particular transaction we also feed that back into the prediction engine so that the next time the engine sees a transaction like that it will have adjusted its pattern prediction and say ah i learned the last time the last time i asked a human what was the right answer and it told me what the right answer is so this time i'll have a new prediction so it's these feedback loops that are really this is where we're harvesting that institutional knowledge that sits in the mind of this human but instead of throwing it in the bin we feed it back into the prediction machine so it can leverage that information the next time it sees a transaction like that so let's look at an example of of pattern recognition and and how you know the human mind versus the the artificial intelligent mind works so here's an invoice um an image of an invoice that kind of looks like that east repair invoice that we showed earlier and i've purposely made it smaller so we can all take a look at it you know apply our our human recognition talents to this particular image now if we look at a collection of invoice images we as humans we can zero in and we can look at each one of these invoices and we use the pattern recognition that the human mind is really geared for right so the human mind it looks like things like colors shapes right logos so these kind of things jump out at us and we're relatively easily able to pick out that that i think pretty sure that is the image um that we saw on the previous screen now contrast that with the way that an artificial intelligence engine would look at a transaction like this with artificial intelligence the computer mind has so much more processing capacity than the human mind it looks at not just the patterns that we see when we zoom out but it knows every single data element on every single one of these invoices with perfect precision so rather than looking at these blurry images it's comparing in real time at computing speed the entire transaction set the entire element uh excuse me the entire collection of data and it's it's using that type of pattern to figure out which is the pattern that i'm trying to match to so we can dig in a little bit deeper and we can take a closer look at what that really looks like so imagine if you can i'm surely relate to if we the the larger the pattern the bigger the pattern is the more accurate the predictions can be because we're we're taking into account as much data as possible in any discrepancy big or small in those data sets represents a difference and we're able to predict patterns differently so the way that artificial intelligence is applied with absent in the accounts payable process is we monitor an invoice transaction from the moment it arrives in the ap inbox right so the very first step in the absent process is we monitor the ap inbox for a new email to arrive when that email arrives artificial intelligence goes in opens up that email extracts the attachments and first makes a decision is this an invoice or is it not an invoice if it's not an invoice it goes down its own track for the invoices that get extracted from the emails it gets submitted into the ap processing flow and it flows through this kind of chain of events right so we we hit the vendor master file the transaction will be adjusted when we apply the vendor master id if it's a non-po invoice it gets coded it gets approved so it's changing right that the transaction set is building and growing if it's a po invoice we match to the purchase order we find the good receipt we make the three-way match if the invoice makes it into the erp maybe there's a journal entry that needs to be made maybe somebody in accounting makes a final adjustment to that invoice before it's ultimately ready to pay and when it ultimately gets to that okay to pay is when it's all done right so what that represents in terms of artificial intelligence in the pattern uh recognition in development is when an invoice arrives all the way on the left of our screen we know absolutely nothing about that invoice but as it makes its progress through this chain the development of the pro the pattern and the progression of the invoice is all being recorded by artificial intelligence until we get to the end of it where we know everything about that transaction we know all of the good information all of the answers right so over here we just have a whole bunch of questions with no answers but when we go through the process all of those questions have been answered and that's the pattern that artificial intelligence records and that's the pattern that artificial intelligence uses the next time an invoice starts on the left and it represents a whole bunch of questions is it goes into the library of answers and it's trying to find the answer that matches the questions that that are on the pattern that showed up as a new as a new invoice so we have to start to think about artificial intelligence differently than the human mind it's truly the biggest brain in the room it has more processing capacity than the human mind and we've all been you know raised to believe the human mind is a very special thing and it is it's very unique right but the computing mind brings a different level of capability to the business world right so the processing power you know we it wasn't that long ago when we had technology and if we needed to add processing power that meant somebody had to order a new server and put it in the rack and and then you know network that that server into our pro our computing environment but with cloud-based technology today we don't even have to ask for more processing power with the if if our engines and our uh computing needs exceed our current availability we're all in these cloud-hosted uh data centers that that just turn the dial up automatically and we don't have to ask for more processing power it just arrives and then we get a bigger bill at the end of the month right the the the the data sources the data services are automatically adjusting to the demand that we're you know putting into the computing uh environment so it's almost limitless how much processing we can get out of these processing platforms and the amount of data that's available to us to apply to these patterns right so we showed earlier that that transaction flowing all the way through and being able to record every single data element and every evolution of the data as it's changed as it goes through that process and record that and create a new pattern for it that we're able to record against is how we're able to use this massive computing processor to look at massive amounts of data and compare it against other big massive amounts of data and come to a very discreet result at the end which is one transaction i think this transaction looks like this transaction so i predict it will behave the same as the one before right and and cloud computing right we talked about that earlier the fact that all of this sits in the cloud it's available anywhere it's very scalable we can process massive amounts of data and have limit the pr limitless processing capacity to perform uh these really large calculations that we need to challenge uh the computing environment with so with artificial intelligence very very different uh computing capabilities than the human mind and and these are the types of problems that we can solve with ai this is what it looks like when we apply artificial intelligence to the accounts payable problem so on the left side we have invoices coming in xml invoices scanned invoices coming directly from the email inbox all mediums are eligible to go through this engine it's not just paper invoices and it's not just pdfs it can be xml it can be edi and all of these flow into this artificial intelligence engine and then they get processed right and as they're being processed we create the invoice on the platform we validate it apply accounting strings do the matching with the purchasers if it's a po invoice and then we have the feedback loop anything that gets adjusted gets fed back in and we're all set and okay to pay so using artificial intelligence for accounts payable is is truly a revolutionary process for ap automation so what are some of the outcomes that are achievable when you use a product like absent for autonomous processing well firstly more invoices can be processed per fte per day so before absent or artificial intelligence is applied each individual fte is only able to process a small stack of invoices and the the productivity rate grows very very quickly with artificial intelligence so out of the gate we can almost double the productivity just because the extraction the accuracy of extracting the data using natural language processing and computer vision is so much better than any other extraction technology available so we automatically make big gains right out of the gate then if we apply the historical data all of all of the data that has been processed prior to this uh introduction of of artificial intelligence and we feed history into the ai engine so it can start to develop patterns that pattern development allows it to start to make predictions i think this invoice i've seen it in the past so i think it will be coded like this it'll be this gl account this cost center and route to that approver and after we've been running the tool for three months that learning has grown the patterns have developed the patterns are more precise and we continue to continue to make processing gains and the beauty of artificial intelligence is it's always learning and growing and and and even more importantly is this these productivity gains they don't come at the expense of somebody writing lines of code or somebody making adjustments to templates or a software when they're coming in and and issuing a change order to make some optimization this is what artificial intelligence is designed to do this is what you're buying when you buy artificial intelligence is you buy a growth a a progression of of of productivity another big change that happens is the reliance on external parties outside of the finance organization outside of the uh ap organization is much much smaller before artificial intelligence procurement would get presented with a lot of exceptions and some of those exceptions might not be really procurement related exceptions it might have been a capture error right so maybe the ocr wasn't able to extract the purchase order number properly and it got routed through the workflow to procurement to say hey what's the right purchase order for this invoice well that's frustrating for a procurement representative right there their job is procurement and now it's we're asking procurement to come in and fix an accounts payable problem which is extracting the data properly if the po number had been extracted properly the po would have been matched properly and we wouldn't have to ask somebody like a buyer to resolve an extraction error so what we get left with at the end is truly the exceptions that we need procurements help for like i captured this this invoice i matched the purchase order i was matching up lines and i noticed that the price on the purchase order is different than the price on the invoice me and accounts payable i'm not sure what to do with that i need to route that to the buyer who created that purchase harder had to have that person decide is the purchase order right is the price on the purchase order wrong or did the suppliers send the wrong price on the invoice and i need to call that supplier and say hey we got to change the price on this invoice please resubmit right so we're we're getting to the point where only the exceptions that procurement needs to be involved with are being presented to them this creates a much happier environment for our partners in procurement it also creates a happier environment for our suppliers right suppliers are being paid more accurately more promptly they're not calling up their buyers saying where's my invoice where's my payment this is all happening more smoothly so if you look at all of these that might have been exceptions before they're flowing through properly and they're being paid on time without the delays of having to be workflowed and sit in a queue and ultimately get results so everybody's happier in this environment other people that are happy are cfos we have visibility into liabilities very quickly so when you have artificial intelligence monitoring the ap inbox in real time and extracting invoices and performing the autonomous extraction and populating the accrual report that takes minutes in the artificial intelligence world so there's no more concern about uh what what what invoices are sitting on somebody's desk or sitting in the ap inbox unprocessed that i don't have visibility into what are those liabilities that i'm unaware of like that's a nervous situation for for everybody at the end of a reporting period to go gosh we've got 150 invoices in the ap inbox that we haven't opened yet so i'm going to have to perform an accrual to estimate what my liabilities are for this period with artificial intelligence observing that ap inbox you almost always have real-time visibility into an estimate of what's what is uh your outstanding liabilities you can also create choice for suppliers right one of the things that buying organizations have been working hard on over you know the past decade or so is having you know uh different invoicing processes that they're asking their suppliers to comply with so go to this portal send me an edi upload a cxml and those those can be those can be onerous for some suppliers some suppliers aren't able to adhere to those requests so giving them the choice to say hey if you can go to this portal and submit your invoice that's great or if you have edi and you can send it to me that's great but if you don't don't worry about it you can just send your invoice to invoices at customer.com right send them into the ap inbox previously organizations hesitated to open up that channel of invoicing because it represented work and inefficiency and workflow and rules-based processes but with an ai process like the one that we've been talking about up to this point those invoices flow through just as smoothly as the invoices that come through these channels right so we now are able to confidently give our suppliers choice and reduce the friction that we introduce to suppliers when we ask them to invoice our organization a different way than they invoice their other customers this is a pretty unique opportunity with absent is we're able to aggregate all of your spend across all of your channels your expense reports invoices your card transactions any interface files that come through your gl and look for duplicate payments in a and and perform artificial intelligence driven audit solutions that truthfully could never be cost justified in a human-based environment it would be so expensive to have a human look line by line at every single transaction and try to find a duplicate invoice some organizations do that they have really they know they're prone to duplication like that so that's super expensive but with artificial intelligence again that big brain that we talked about a few slides ago you throw a problem like this at it and it can aggregate that data and it can very quickly and very confidently identify a duplicate that it would want to present to a human say hey it looks like i paid an expense report for an amount that just came in an invoice a lot of times those are not you know kind of nefarious activities you know an individual might go to a business and purchase something with their p card and put it on an expense report and that that that business that they purchase from has a process in place where it just automatically triggers an invoice so inadvertently that invoice gets sent to accounts payable and accounts payable sees that invoice and they're often unaware that that payment was already made for that product or that good or that service through an expense report with technology like appsend we're able to apply artificial intelligence to find those duplicates and present them to you so that you can resolve them and and reduce your your duplication of spam we also have some really cool architecture with uh the technology that we're deploying we have both our we have our our own processing platform right we call this the the absent mastermind platform where we perform all of the extraction and validation we also have a connection layer where we have our own kind of uh internal and self-hosted middleware layer that allows us to call apis and and and interface with our customers computing environments very very easily so this with this type of approach the reliance uh for our customers to have their it teams get involved is much much smaller we're able to independently reach out and connect to the apis that are available in most of our customers cloud-based environments and make those connections independently of our customers i t teams and move that data both into your environment as well as remote pulling the data out that we need to put on our platform in order to process those invoices so we can build these integrations very very easily and quickly without a lot of reliance on it tunes this is uh something else that's really pretty unique about uh an autonomous ap implementation is previously you know with with rules based technology you know the the workshops and the amount of information that used to be required to to get the information ready for an ap platform we would start with like all right tell me about your vendor master data how is your vendor master data structured you know do you have parent-child relationships right tell me about your purchase order data how do you how do you use blanket orders do you have um service orders that we need to contend with right all of those things needed to be understood and then configured and built and if they were wrong they would need to be adjusted but with artificial intelligence we don't need to know those answers artificial intelligence adapts to the processing of the historical transactions so we use history to guide what a new invoice should be processed like so this drastically reduces the effort associated with implementing the technology so if you've ever lived through an ap technology implementation you'll be really excited about what uh using artificial intelligence can mean for your next implementation so how do you kick off an ai project well let's have a plan right so first define the problem what are we attacking what is the invoice volume that we want to attack is it just the pdfs coming through the ap inbox where do we want to go after xml invoices scanned invoices what what is the problem make sure that the technology is cloud-based right we want to be able to access it anywhere and take advantage of the mass and computing uh powers that are available with cloud-based solutions and this is a really important one and it's very different when you buy an artificial intelligence product which is you really you know when you're when you're buying artificial intelligence it's different than buying rules-based technology because what you want to see is how well does the technology learn right that's really what we're talking about is if i show you a series of transactions and then i show you another series of transactions that are similar but different does the technology actually have a learning component and does it adapt and can i see that you should be able to see that in a proof of concept so when you're contemplating buying artificial intelligence technology it's different than you know the traditional approach of you know selecting a series of vendors having them do some some presentations filling out some rfps there's more to it you really want to you know feel the learning effect of artificial intelligence and then once you've you know understood that you want to communicate with your teams that you know artificial intelligence isn't here to replace jobs it's here to augment jobs it's really here to take out the keeping the simple stuff simple so that we are providing the complex and challenging scenarios for our folks that really understand the business associated with it so getting everybody to understand that artificial intelligence is really here to just smooth over the bumpy spots so i talked to the very beginning and the whole premise of this this this presentation is um that that you've been lied to about how how how automated is your ap process and this is something that we're really excited about here at absent is we have a free trial you can you can follow the url that you see on my screen and you can go and register for that free trial and you will be able to log into our platform and upload invoices um right away this isn't the kind of free trial where you go and you enter your your your contact information and somebody reaches out to you and has a conversation with you you can sign up for our free trial today and upload your invoices today so we encourage you to discover the difference and understand how autonomous ap can apply to your business and what real automated processing looks like for accounts payable so thank you for your time today i encourage you to check out this website follow this url and give our free trial a try thank you
2022-08-04