Enhance Your Workflow with Basware Invoice Processing for Product Management

Basware invoice processing for Product Management

hi this is Mike Jasper Executive Vice President at velocity procurement velocity is a procurement advisory firm helping our clients to mature their people process and Technology around their source to pay Journey I'm joined today by Ola MZ Vice Pro vice president of product management at basware bassare is regarded as the leading provider for accounts payable automation by firms such as Gartner and Forester today we'll be talking about how we can use artificial intelligence or Ai and to help automate our coding of non-po invoices Olaf it's good to be here with you again today yeah thanks for having me again now Olaf I've seen the non-po invoice process become one of the most cumbersome areas within accounts payable uh between coding invoices getting them routed for approval uh it's a reason why people are pushing so hard on a on a no PO no pay process but as you and I know there's certain invoices that don't bring themselves to that very well uh they're hard to put po back a lot of times it's after the fact POS which are very little value add uh you know and a lot of that comes down to the coding you know we need to get proper coding on an invoice and a PO is a great way to do it but you know it doesn't happen every single time it's not the best uh use case in every single invoice scenario um you know what's bord doing to help autom that process of getting the proper coding on an invoice if there's no PO to go to yeah you're you're spot on right coding is the most complicated part of processing an invoice one way or another and and typically when I have a discussion with a customer and I ask into the room that do you know your own cost center then 90% doesn't know that even and and that is such a basic element so clearly that's a pain point and this is where we started long long time ago introducing different kind of coding templates right can we get some regularity into coding and that works to a certain level fine but there are so many exceptions and there's so much overhead so that's not a long-term solution and this is where we are starting with the AI but you could teach and we we came up with a with a service which we are calling the smart coding who is actually utilizing all of your invoic that you are processing and come up with these kind of predictions and the beauty out of that one is that it's not a one time fits all so it's not utilizing the whole string of an individual invoice it is able to come up with predictions for individual coding Dimensions that are best suitable for that specific invoice and combine those and in that way you would be able to get unique coding combinations that you haven't had before without having to maintain these kind of different templates that you are running there and that sounds like give be a game changer because I've seen so many times where you know we're using history you know how did I code the invoice last time and and that's helpful but it's not always going to be the same and a lot of times that gives a false sense of confidence the user they're going to go in and just pick what they use last time even though it's wrong uh you know and then AP has to go back and clean it up on the back end which becomes its own process so the idea of being able to do that on a code by code basis and you know come up with those unique strings is is something I haven't really seen you know presented in the market before that's pretty exciting yeah and and like you said right is whole accuracy and the confidence is a key element so this this is something that that needs to be a journey between the machine and the human right so you need to guide the user to take an action and that action could be ah this coding looks okay because the confidence is really high or okay I got a proposal but I'm not sure and as long as you can get the user on board with that one you can make sure that they make the required adjustments and actually start training the AI so you get better and better and in the end of the day you will get more accurate coding uh among all of your invoices which is then of course really good for all your spent information that you want to have down street absolutely what's the customer feedback been on this as you have you rolled out to some of the customers what's their what's their reaction then yeah they have been really enthusiastic and that's mainly because we are presenting this in a different way because we have we have a model and we have the data because we are processing these invoice in our system already so we came up with a value based introduction basically to the smart coding where we did a simulation we put your invoices through our model and this is what you can expect so it's very easy to show the customers that hey with your data if you would start today you would get 80% accuracy and from there you can then build up and start the actual Journey because like like I said this is really a data Journey that we are initializing because many customers don't have a full overview of what is really happening on the coding dimensions on the chart of accounts that they are running and there's a wide variety of possibilities there which are one of the root causes for getting incorrect coding and and troubl for AP going going forward and I had kind of a really nice example the other week when I had such a discussion with a customer and we noticed that they had a they had an invoice Dimension where they had over 14,000 different options that the user could pick from and they processed hard l a thousand invoices that require that actual Dimension so no wonder that the users were always struggling with finding the right Dimension because it was just overwhelming that how many do I need but they just didn't know so part of this whole exercise is what can I do to improve should I remove Dimensions I'm no longer utilizing and these kind of things to help everybody in the end of the day uh getting better accuracy that's interesting and you know Ola I know uh smart coding has been around for a while in in a couple different versions uh you know how has it evolved you know what was it a few years ago where are we today where do you see it going in the future yeah we started a few years back to really have it for a specific user so we really thought that okay let's help out for example you if you would get an invoice what would you do what would be your circumstances uh and that worked but then we saw that that your Workforce isn't static so especially with this whole Global recognition wave ongoing people come and go all the time so how do we make sure that that knowledge that we got it from a specific user is moving away is moving with the company at least that the next person doesn't have to start from scratch so it was becoming very clear that this needs to be a global Service this needs to be for the full organization so it cannot be for an individual and this is how we started the next version or the version where we are now at that hey now it doesn't make matter who is actually responsible for the coding because we're just taking all of the invoices and how you will be processing of course the organizations dimensions are very much important in there because you could have country specific dimensions and these kind of things but it's not only the users's own invoices without the whole invoice set where the training is coming from so it's one size really fits all uh for for a company okay and you know one of the things we talked about when we were having our ation on invoice ingestion was AI takes a lot of data to learn from you know the bigger data set you know the better you're going to get on results um you know how does that play into smart coding you know I know you mentioned sort of 80% up front you know based on some of the historical data what do you what do you recommend to customers how do they you know go through this journey from implementation to you know really maturing the use of of smart coding yeah so the beauty of all these different kind of services together is that the more data you capture up front the better it is but the thing is that you don't have to map out all of your Fields any longer towards your dedicated strings because all of these servers are getting so smart that they know what to look for so instead of pinpointing specific Fields you just say check message full message content for a specific value and it's like like with a supplier and if you give your cost center because you ordered coffee uh from from the cafeteria they can place it anywhere on an invoice and it's very hard to map that into your database where the data is going uh towards these services so now the system knows I need a cost center I know how they look like can I find it somewhere for the raw data even though it isn't given as an output from the initial service the data is still available for the next service to capture and like this you can really get the full benefit out of the electronic data set that we are preparing it sounds like it's going faster and faster every day you know going from you know where we were a couple years ago to today to some of the conversations we've had about you know where AI is going with an accounts payable it sounds like things are moving pretty quick and you're you're staying right on top of it it definitely moves quick and this is only the beginning well Ola it's been great chatting with you I appreciate all the conversations we've had all the all the knowledge you've given us around using artificial intelligence and accounts payable you know with how fast things are going we're going to have to get together again pretty soon and go through all the new things that basor is doing with AI and all the ways we're helping to automate accounts payable I'm happy to have the next session great thanks for joining us thank you