Bridging the GAAP Podcast Series: Behind the green curtain with capture

---

Speaker 1

Hello, and welcome to Medius' new podcast series, Bridging the GAAP. Together with Spend Matters, we've put together a podcast series exploring a wide range of topics from the evolution of technology and finance to tax compliance to sourcing strategies to fraud risk and mitigation. In today's episode behind the green curtain with Capture, Pierre Mitchell, Chief Research Officer at Spend Matters, talks to Katarina Andersson, VP of Capture and AI at Medius, about automated invoice capture, how it works, the history of Capture, and where Katarina sees Capture going in the future. Take it away, Pierre.

Speaker 2

Well, hi. This is Pierre Mitchell. I am the chief research officer at Spend Matters, and I lead our strategic research for our coverage of all things source-to-pay technology related. And I am very happy to be here to interview Katarina Andersson to get her insights.

She's a real expert within P2P and particularly on the invoice automation side. So I have just a ton of questions for her about getting under the hood or behind the green curtain, if you're familiar with The Wizard of Oz. So Katarina, welcome. Maybe just do a quick introduction of yourself and then we can get into the discussion.

Speaker 3

Thank you for having me. I'm so happy. And wow, what an introduction.

Yeah, my name is Katarina. My nickname is Katta. Actually, my mom is the only one in the world that calls me Katarina. But it doesn't matter.

I respond to all the different names that that's in the book. But that's so interesting, because I've been working with Capture since the beginning of 2000. And been in different positions doing it. So my husband actually tells me that the only thing I can do is talk about capture.

My daughter, she rolls her eyes at the dinner table, say the only thing you can talk about capture mom, so please leave the subject and learn something new.

Yeah, so that's me. And right now I'm working at Medius as the Vice President of our Capture Solution at Medius and responsible of that.

Speaker 2

You're the queen of capture at Medius.

Speaker 3

Yep. Or what some colleagues call me, Miss Capture, but that could actually be interpreted in the wrong way.

Speaker 2

Well, that's great. I mean I, you know, and I think the listeners will be really interested in learning a little bit about Capture. And I mean, let's just kind of not to spend too much time going through the time capsule of history, but maybe just talk a little bit about where we've been in terms of, we started with OCR and then we kind of moved towards extracting more intelligence in an automated way out of the documents to help the process. But maybe just tell us just a little bit about, you know, kind of how we got from OCR to kind of where we are today in two minutes or less.

Speaker 3

Yeah. Yeah. It has indeed been a really interesting trip since I've been working with this for more than 23 years. In the beginning, I was, like you mentioned, the queen of capture.

I was out there removing the tedious manual work of manually capturing the data from the invoices with your keyboard. So it's no OCR at all. But at that time, we were introducing templates, you know, what everybody hates today, you need to maintain them, you need to work with them to really get a good extraction rate, meaning capture from the invoices. And then we moved into some more like when you looked at the titles and the position based data capture.

But today it's totally different. We're doing AI machine learning. Actually, we're trying to understand the content of the invoice. So to remove the tedious work of maintaining templates, to configure the system and so on.

So it has really been a really interesting trip, a technology trip, so to say, don't you think, Pierre?

Speaker 2

Well, I do like the Grateful Dead, and I will say what a long strange trip it's been. But in terms of that trip of moving from kind of OCR, this evolution, in terms of OCR and then being able to, as you said, use AI to understand the data that's in there to train the AI to understand supplier invoices and to really ultimately help us get to what we're hoping for which is that 100% touchless match rate, etcetera. Let's kind of pull out a little bit and kind of just take that bigger view in terms of when you're actually trying to do, let's say a P2P implementation and you're actually trying to go all the way through to actually not just get it to P2, ready to pay and approved and ready to pay, but actually taking it all the way through.

That ability to do that onboarding is really challenging because suppliers might not just like the onesie twosie portal and they might not necessarily be EDI-enabled, There's a lot of stuff in the middle. So maybe just, you know, tell us a little bit about how you guys solve that problem of get the suppliers digitized and be able to get that 100% automated no touch match rate, know, eventually. Right? How do you do that just from a project standpoint in terms of onboarding the suppliers?

How does that process generally work for you guys with your clients?

Speaker 3

That's a very good question. Because just like you say, it's a maturity of the supplier, because not all the suppliers, some suppliers still sends paper invoices.

I meet customers every day that still have a certain percentage of the incoming invoices in paper. I cannot believe that. But the majority receives the PDF documents. But now with the E regulations and COVID, of course, has pushed our invoicing to the future to be the preferred way, because it's not that the suppliers want to send invoices, it's the countries or the tax authorities that wants the buyer and the supplier to send invoices because they want to collect taxes to ensure they get the taxes so they can start build up the countries again after COVID.

So what we at Medius really think it's important just because of this that you can see all these different formats is that we accept any type of format, because our customers or you as a buyer, you don't want to say no to a customer or to a supplier that can deliver what you really want because of the they send in paper invoices. So that's what Medius's vision and goal is to accept any type of invoice and our customers and the buyers do not need to worry about the format. We just take care of it.

Speaker 2

Yeah, it reminds me of kind of like the term like omni channel retail. This is kind of like omni channel inbound invoices, whatever the supplier, however they can do it, just to be able to accept that. And let's say the email with the PDF. Right?

And there's that workflow piece of taking that in, getting the PDF, unpacking it, going to the OCR and the extraction and all the AI and all that. And that that you really bring a really good point is there's the plumbing piece to the workflow automation, but it's also this knowledge piece about how do you understand the tax information that's showing up on those line items? How do you understand contract information that you might need to refer to and pricing discounts and things like that? I mean, there's a lot of information that kind of is behind the numbers, you know, and text that shows up, you know, on the invoice.

So maybe just that knowledge, I think about expert systems or whatever, you know, I mean, just tell us a little bit about how does that AI actually work, in terms of, I think people generally kind of understand it, but just maybe in your own words, say, you know, how do you think about making that automation actually autonomous and using the AI for that?

Speaker 3

Yeah, this is a very technical term. When I talk to my friends and my family, they feel like AI is like the Terminator word, know, on a Schwarzenegger.

But we are not in that world yet. And in that sense, it makes it scary for some people. But I actually believe it can be one of our best friends. What it really does, it collects the data or when you as a user extract the data from the invoices, or correct the data because the technology extracts the data and you verify the data, of course, it learns from your changes.

So you can, in that sense, that AI machine learning is our best friend in the background, that actually takes that knowledge that you're sharing with our AI friend and applies the same thing next time an invoice from that supplier comes in. So that means that you do not need to work on templates and maintain those and configure. It's just this fantastic person in the background, a digital person, of course, that actually learns from your actions without you knowing it and just applies that again and again and make intelligent decisions and adapts over time. So that's simply how I usually try to simply explain it.

But how would you explain it, Pierre?

Speaker 2

Well, I mean, I think you did it well. I mean, I do like the term augmented intelligence where you're just using the technology to basically extend yourself and get yourself out of the repetitive low level tasks. And if you can let the digital bot, which is a software robot and not a scary Terminator robot to actually help do that, that then frees you up to actually do higher level tasks because let's face it, mean, what extent do folks in AP really just love, you know, typing in invoice data? You know, there's a short time that we're on this earth and I think they probably would like to spend their time maybe doing a little bit more and it's just not good job security to do that if that's your only task.

And so maybe let's just talk a little bit about that higher value add in terms of when you have better this automated capture, it then kind of frees you up to now start doing analytics and you can start really attacking the data that's in the invoice as part of the spend itself, And the risk and all these other things that have to do with the business versus just pushing documents and information around. So I will turn it back to you and say, all right, if you think about kind of the next value add, I mean, certainly there's a piece about spend management having good, better data to support what's happening upstream and, you know, the sourcing process and also just in the PO process and making sure you have something to match to and all that kind of stuff.

But, you know, I guess tell me a little bit on the compliance side. Compliance means a lot of things to folks. It's like compliance to what? Compliance to regulation, compliance to internal policy, compliance to our auditors that wanna make sure that we're not there's no fraud that's, you know, happening here.

So just the compliance piece, right? It seems like a big thing where you could get fines and regulations and late fees. And it seems like there's some real money associated with this beyond the labor efficiency of some AP folks. But tell me, I mean, how do you see the invoice capturing helping with compliance or other just pieces of value in the broader process?

Speaker 3

It's also a very good question. What's really on everybody's mind today is fraud to help you protect yourself from fraud. If that is compliance, in some sense it is, I would say that. But with the AI machine learning, we can really help with that as well. Can identify anomalies and we can really see changes in behavior from the suppliers and also see that it's been received differently from different IP addresses or from different email addresses and so on. So we can help on that. And of course, we mentioned e invoicing before the regulations that is happening in the world now because COVID, that is definitely compliance.

You need to ensure that you're following the rules and regulations and how the invoices are received and sent and registered and paid, of course. But what is your experience with this, Pierre?

Speaker 2

Well, I mean, I'm just the facilitator in this conversation, but I will share my two cents on it. Yeah, no, I think this is one of the, I think you hit two of the big things certainly is around kind of the regulatory aspect, and certainly there's the fraud piece of it. And this is just a great place where AI is. Think about our consumer spending and you think about the credit card companies and the level of AI deployment they have around on the consumer side.

It's just like many things, just bringing that over to the B2B aspect, right? So that we can have the good bots watching out on our behalf, you know, in terms of identifying these things. But I will just add kind of the broader piece around source-to-pay and the spend management and the piece around working capital too, right? Where efficiency really unlocks the door to kind of effectiveness.

So if you can get that invoice in really quickly, now you have that ability to maybe offer early payment discounts to do supply chain financing, to start to maybe go after these opportunities where suppliers with interest rates are going through the roof right now. And people are crunched for cash. Great time to do a working capital program for a supply chain finance program potentially with some of your suppliers. So if you're gonna unlock that value, you got to get the plumbing really done well.

Got to get have the clean water if you're going to make some nice drinks out of it. Right? And I think the other piece is the spend management piece around the P2P process to make sure that invoices are coming in and really as part of the design of how you want things to happen, three way match, assumed receipts, recurring invoice, whatever that design is around that spend category should ultimately carry through into execution process and being able to tie the contracts and be able to generate all the detail that you need from a spend visibility standpoint so that the savings that you, you know, maybe negotiated upfront are actually gonna get realized, you know, on the back end.

So I do think there is that big spend kind of impact aspect of it. So you asked me a short question. Gave you a long, long answers.

Speaker 3

But I loved your answer.

I fully agree to everything you say. So it's a really interesting area that we can help our customers.

Speaker 2

Yeah, and It's the choke point, right?

I mean, if you're a lot where does it all come together? It comes together at the point in which you actually set up vendors, you know, and in your systems and actually, you know, paying them when the moment of truth is when that invoice comes in. And if you actually think about the definition of what does a perfect order, a perfect supplier order mean, it means on time in full excellent quality and the invoice information that you need to be able to match to and get that thing paid flawlessly. That's a little bit of an Achilles heel, right?

That is an important piece to solve that. So anyway, that's I'll get off my little high horse and let me send it back to you. All right. So tell us just a little bit about Medius around your differentiation as a provider, you know, it's a, you've got some patent pending stuff, but just maybe talk a little bit about, you know, the differentiators and what, you know, what helps you win in the field against more, let's say traditional scanning doc management workflow kind of oriented solutions that are out there or whatever other options are.

Speaker 3

Yeah, I'm particularly proud of two things we're doing at Medius. And one thing I'm really, really proud of is our vision that our solution should be easy to use and require almost no configuration. Would actually say that we're trying to automate automation and that result in autonomous. And autonomous to me is that you do not need to configure your solution.

It's actually learning from your daily work and adapt to it. So you don't need to maintain it. It's just there. And it's improved.

And it helps you improve. And it is improving over time. So it makes you more automated. And usually when I say this at the dinner table, I mentioned this in the beginning, My daughter says she just rolls her eyes and says, Oh, configuration is so 2000, mommy.

And I would actually say so because if you look at our kids today, they are not configuring their iPhone. It's like, right, plug and play and it's done. Right. And it's same thing when you Google something, it actually learns what ads you're looking for.

So I think that is the vision and really good vision and goal that it should just work. Yeah. And the other thing that I'm really proud of is our invoicing solution. Because for once, and I think you can agree Pierre, even though invoices or EDI invoices, it's claimed to be 100% touchless, It never is.

Usually, your vendors or your suppliers to put the PO number. This is a typical example that we can see the PO number can actually be put in like, I think it's up to six different places in the invoicing file in the XML file. And how would you as a buyer be able to adapt to that change if you have a thousand or more than thousands of different suppliers send the invoices to? And this is where our invoicing and our patent pending solution comes in.

First, it's included in Omnichannel we'll be talking about previously. We just take any format, we adapt to it and you teach the system as a user. If something has not been captured correctly from the XML, you can actually teach the system where the reference or the PO number is, if it's put in the wrong tag. So in that sense, we can make the invoices 100% touchless.

And we can really ensure it's touchless all through the system, making it more autonomous, all the different invoices. So I'm really, really proud of the invoicing solution as well. I think we are making the invoices autonomous or really automated.

Speaker 2

Yeah, no, it's true. I mean, it's easy to kind of say, hey, like with analytics, we go from descriptive and diagnostic to prescriptive, and then you like make this leap to autonomous, but to actually do that, having the solution be able to learn from you and learn from the suppliers and to really help guide them to just make sure the process is actually working, but also, you know, to just have that the AI have learning and empathy, if you will, actually just help you make that process get to near 100% versus, here's a big dashboard and portal and it's still up to the human in the loop to actually figure out how to keep the things running, the AI should be kind of guiding you as to where you need to focus and also asking you questions of where it can help.

Speaker 3

Yeah, I fully agree. And I came across a really interesting example yesterday when I was visiting a customer. And they were processing a lot of invoices, and they had some issues with the invoices. And they came with a request and said, When can you have this done for me?

And I was like, well, it was delivered six months ago. And they were like, okay, but why haven't you told me? And then and I'm like, yeah, but it's in the help file. And it's, we posted it on the internet and everything in our blog.

And they were like, yeah, but when do I have time to read that?

So it's actually, I really believe this autonomous is the future, because who has time to read documentation to get better or look at this different dashboards?

Speaker 2

Yeah, I mean, you see this in the supply chain very much around just the number of messages of just the supply chain is overwhelming and invoices are kind of the last mile, if you will, when I was talking about the perfect order side, That is also just a very, just think about the millions of invoices that companies process. But yeah, I think it's a good point to have the AI in the system to just let you know that all is well, or maybe there are some things that require attention and here are the things that you might wanna prioritize them in the following way and have the system guide you to kind of feel it's like Jerry Maguire, help me help you, you know, have the system say, help me actually go towards autonomous. So I think that's the exciting part as you're saying around your solution is is, you know, automating that evolution itself.

Well, we've really kind of gone lots of different places in the conversation. Is there anything that we missed?

Speaker 3

No, I don't think so.

Speaker 2

Well, the AI just gave us a score of 8.9. So we're doing well.

Speaker 3

I'm happy to hear that. That makes me happy.

Speaker 2

And I'm just glad I could provide a different audience than your daughter rolling her eyes, you know, and we could kind of geek out in this area because I think not just us, but folks in the business world are going to find this really important because it is just such a pain point at scale, right?

Speaker 3

Definitely, definitely. And that's why we love our job because or work because we can help, right?

Speaker 2

Yeah, it is good to be useful and it is good to help folks and free them up from the drudgery so they can go after the, you know, the bigger problems, right? And we have a lot of bigger problems to solve in the world than this. So, and we'll leave that on an optimistic note, right? So, let's go get at it. And so, Katta, thank you so much for taking your time. I know you're busy. Your clients are probably wanting you to be spending more time with them than me, but I really wanted to.

Speaker 3

I enjoyed having this chat with you as well. I learned a lot from you as well.

Speaker 2

Well, there you go. So we learned from each other. It's a proverbial win win. Okay. Well, you everybody for tuning into this conversation and hopefully we'll have some future discussions on other topics that may not be as exciting as invoice capture, but you know that that will still be able to add value nonetheless. So so thanks again for your time and thanks again, Katta.

Speaker 3

Thank you.

Speaker 1

That was behind the green curtain with Capture. How does automated invoice Capture actually work? Part of our ongoing podcast series, Bridging the GAAP. Special thanks to Pierre and Katta for their time. If you like this podcast, please be on the lookout for our next installment.

And to take all of your business spend management far beyond basic automation, visit medius.com. Thank you for listening.