All right. We’ll, we’ll get started here today. Thank you again to everybody for joining today’s call.
This is the Rethinking Invoice Processing webinar, and this is part one. So today, we’ll discuss some of the manual challenges around invoice processing, what actually goes into an invoice processing process, and what a proven solution might look like to modernize that process.
My name is Matt Hardwick. I’m a solutions engineer at Lydonia. I’ve been with Lydonia for four and a half years, and have some deep knowledge and involvement in some of the implementations and scoping around these invoice processing processes.
If you’re not familiar with Lydonia, Lydonia is a business transformation firm based out of Boston, with offices in all over the country and in India. We drive value for organizations through AI automation, whether that be unlocking value with technologies you’ve already invested in, or that you’d like to continue to invest in outside of your current tech stack.
We focus in AI, agentic orchestration, traditional automation, and intelligent document processing to help folks guide business decisions, design solutions, and deliver outcomes.
And today we’re going to focus in on invoice processing. So, we’ll take a, first, a high-level look at what invoice processing is, what that process entails, and then we’ll move into some of the challenges around that process.
So, what is invoice processing?
This invoice processing process, more often than not, will look a lot like this diagram, with 4 steps, and each step having a variety in how it can be accomplished. So, starting with the invoice intake process, how are we receiving our invoices from our vendors and suppliers?
The most common way is via an email inbox. Oftentimes, we see physical snail mail coming in, and then having those invoices be scanned and processed as scanned images. Or maybe there’s a vendor portal that you have to go out and physically get those invoices from, and the AP team is doing that work now.
Another question we think about with the intake process is, what format are these invoices in?
The typical way we see it, and the most common way, is PDFs, and that’s preferred for many vendors, just the consistency of that document type. But sometimes we see Excel files, we see Word documents, we can even see free text in the body of an email. All that variability adds complexity to this process.
Then that next step, once we’ve intake our invoices, we go through the data capture process. In the manual invoice processing process, this is where humans spend a lot of their time. They go through each invoice, look for any major red flags, whether it be with the vendor information, the payment information, any line item data, and then if they do find any errors, it’s reaching back out to the vendors. It’s having that vendor communication to do the error remediation. And then a lot of times that involves waiting. There’s a, whether it be hours, or it can even be days for vendors to get back to you and resend those invoices properly. And even then, it might not be proper. So, it might be a back and forth of 2 or 3 different communications.
Once that intake and data capture step are complete, invoices are then typically saved into some flavor of document repository, whether it be SharePoint, Laserfish, there’s thousands out there, and they all operate a little bit differently, right? So, we’re capturing that data and saving that data off, whether it be for regulatory reasons or just for later on in the process, we may need to go back and get those.
Once we have all that data, we then need to validate it. Once it’s understood, the AP team has to perform that tedious step of, validating, receiving approvals, and so on, on that invoice. And what we see most often is a scenario for most companies, is invoices will come in, they need to be matched against something, whether that be a purchase order, a shipping document, whatever it may be, we need to ensure that what was purchased is what’s being paid for.
So there’s two-way matching, where we’re matching just the purchase order and the invoice. Some companies will also have to match to a goods receipt, which then it becomes a three-way match. If at any point in that two-way or three-way match process that data is not in alignment, the invoice can’t move forward, and it must be reviewed further. So, whether that be escalating it up, or going back to the vendor, going to the shipping broker, whatever it may be, there needs to be some sort of review process.
And that’s just one way that an invoice can find itself needing further review or approval. Another way we commonly see approval needed is whether a new vendor’s being introduced, and we need to understand if the payment terms are correct. When the invoice exceeds a certain dollar amount is another common place we see approval needed. And then maybe a GL or bank code needs to be validated.
All things that slow down the process, require extra people and extra steps to get the invoice in the right place for it then to be posted, which is that final step.
Once that invoice is finally matched, it’s finally approved, it’s ready to be posted. And again, this is where we see a lot of manual, human steps being taken, is that data entry. A lot of AP teams will throw bodies at the data entry piece of this to ensure that it’s being entered properly into our ERPs, our payment systems, and so on. And this is, again, where we see a variety of systems. Since not all ERPs allow for the data to flow directly into the GL or the payment processing system. So again, there’s that disconnect between systems that requires human intervention.
And oftentimes, this piece, again, is a multiple, takes multiple steps. The invoice information’s being entered in multiple different places.
And once it’s posted, then at that point, it’s complete, we can move on to the next. So, this entire process, it can take, the happy path, it may take 10 minutes. But oftentimes we’re seeing it take hours, if not days, to go through this process, especially in that data capture and that validation piece.
Since everyone does invoice processing as a part of their AP workflow, everybody knows that this is complex, and it’s also critical. It’s a critical process in the, in our AP processes, and it comes with a lot of difficulties and bottlenecks that can hold an invoice back from being paid.
First and foremost, inconsistency in the invoice format and data can create challenges for manual processing, since the key data may be difficult to find or interpret across vendors, or may be missing from the invoice altogether. These quote-unquote dirty invoices lead to more communication with the vendors, as we try and get things added, clarified, fixed, and it can elongate that process, as I mentioned before.
Another bottleneck in the process is that matching process, that two-way or three-way match. When invoices don’t match or partially match, it results in an internal escalation that can take the time of management resources, or just emails back and forth, right? That takes time off of people’s days. Maybe we’re going back and forth internally, maybe we have to go back to the vendor. Again, it just causes a lot of delays in that invoice being paid.
There’s also the challenge of the downstream systems not always fitting the needs of the process, whether that be an issue with the auto-matching logic, the internal pricing data not being accurate, or simply the ERP or other systems not being purpose-built to handle invoice processing.
All of these examples result in greater manual effort for the AP team, and risk errors and longer processing times throughout the process.
All these difficulties that result in the invoice processing process taking longer bring about the possibility of losing out on early pay discounts or incentives. They can cause late payment penalties, or worst case, we’ve seen invoices getting lost in the shuffle of approvals and countless handoffs, which results in the invoice not being paid at all.
The downstream effect of this externally is vendor relationships being harmed over time as those issues become more consistent leading to worse relationships, maybe you get cut off, maybe you’re not able to buy the parts that you need, to run your processes and run your business.
Internally, the downstream effect of this is a jilted process with delayed visibility to cash flow, as well as unexpected hits to cash flow due to loss of discounts or incentives, the addition of late fees on invoices changing the cost. And there’s also the people cost, right. Your AP team getting frustrated with the process and the countless back and forth. Maybe you’re a sales organization whose salespeople rely on getting paid based on invoices getting paid. So, there’s a lot of different downstream effects, both externally and internally, that can affect the business as a whole.
And a lot of folks will try and solve this with automation of some sort. And over the past, let’s call it 3 to 5 years, a lot of people have turned to traditional automations. They want to dip their toe in automation in the automation waters with some of the traditional techniques that are touted as quick wins or have a low cost to implement. The path of least resistance to automation and invoice processing is to typically go with an AP-specific solution, whether that be a bolt-on to an existing ERP or CRM system or payment system, that’ll proclaim to streamline your invoice processing. Or maybe it’s an AP point solution that will do the same, but it’s hosted in a third party.
These solutions will typically leverage, some more legacy technologies, like OCR, or a simple machine learning-based solution, to be good enough at reading your invoices, while providing some basic processing logic that aligns with that point solution or that ERP system’s understanding of invoice processing that leaves you little to no room for customization to fit your process.
And these solutions just aren’t built to scale. They typically aren’t purpose-built for your organization. And they have technological limitations, such as the template-based OCR solution that we typically see, or basic machine learning models that are only trained on the top vendors. Those things are brittle and they result in low straight-through processing, or STP, as we call it, for your invoices, which means a human will still have to review and correct about 7 out of every 10 invoices that come through the process.
Aside from the low straight-through processing, those maintenance costs are also critical, too. The brittle nature of these solutions, allow for maintenance costs to continue to rise as you add templates to an OCR-based solution, or fine-tune a model to handle more vendors over time. And then when those vendor formats change, or you’re introducing new vendors without retraining, the templates break, or the model struggles to adjust, which only harms those STP numbers, and the overall, time on task for your humans, it alters that greatly.
It’s also just not great solutions to be able to scale. They’re not great communicating with other systems. That ERP bolt-on for invoice processing, for example, isn’t going to be able to intake your invoices directly from Outlook or that vendor portal. It’s not going to be able to communicate with those systems, and it’s going to struggle to store those documents automatically in that third-party document repository you may be using. They’re limited to the processing of documents within that system, so if everything’s housed within your ERP, great, that may be a good option. Again, a good enough option. But anytime we’re stepping outside that process, or outside that point solution’s capabilities, we’re still using humans.
And so, Lydonia, on the other hand, what we’ve done is we’ve built in a proven invoice processing solution that combines our experiences implementing robust invoice processing solutions for customers with the cutting-edge technologies like Agentic Automation and what we call generative document intelligence, which is sort of the next step beyond the machine learning models and that OCR-templated approach.
These technologies and our proven solution combine to help companies avoid the pitfalls we discussed earlier, and build a scalable solution that is customized to the intricacies of your internal process.
I mentioned Generative Document Intelligence, or GDI, this technology is the latest and by far the greatest in invoice extraction, because it uses a series or chain of large language models to accurately extract and validate information from invoices. This means we can treat the extraction of header data and table data differently, with purpose-built models to achieve higher accuracy.
Whereas a machine learning model is just going to try and attack both of those sections the same, and it may vary in the accuracy. We may do great with header data, and it may struggle with table data.
So that means we can, we can extract accurately, and then we need to validate that data also, right? So, in that chain of large language models, there’s validation models that we use as well to go ahead and validate what the previous models have extracted to, again, ensure that we’re pulling off the right data and that we’re going to be able to cut down on the human intervention time, the exception handling time, and increase that straight-through processing percentage.
So, with the Lydonia proven solution, we typically see a human may only have to review 1 or 2 out of every 10 invoices, down from the maybe 6 or 7 that, the more traditional solutions would have gotten us before.
Our Lydonia proven solution also leverages agents to handle the process exceptions, those low-level approvals, and some of that vendor communication that have typically been manual tasks, but only up until the point where you’re comfortable internally with automation handling those tasks.
We know that, we still want some human touch in that mix. Sometimes the vendor communication needs to be handled lightly because we know that this particular vendor needs to handle things a certain way, or we get to a low-level approval that an agent can handle, but if we need to continue to escalate it up the ladder because it’s the, let’s say, for example, the total of the invoices so far and above our invoice approval, handling capability, that we need to bring a human in.
So we can give agents the knowledge base, the context, and the tools required to handle these lower-level tasks with the confidence that they need, and still give them the ability to reach out to a human when it’s beyond their capabilities. So when humans do get involved in the process, our solution also leverages a simple, user-friendly exception handling interface to do so for that extraction accuracy piece, and then we can also customize the solution to fit your company’s needs, whether you’re leveraging Slack, Teams, Outlook Email, whatever the best way for you to get those exceptions and handle them, we’ll work with you to leverage that technology. And for those trickier areas of the process, like two-way and 3-way matching, as well as that GL or bank coding, we still ensure that the solution is custom-built to fit your process, not just what we think your process is.
And so, because of our, the combination of best-in-class technology and a solution that’s tailored to your invoice processing process, Lydonia consistently helps companies achieve that high straight-through processing, and as a result, reduce the time on task for your AP clerks and other subject matter experts involved in the process. It allows you to operate with a clearer, more stable picture of your cash flow through the collection of pricing incentives and elimination of late payments. And improve those vendor relationships, which allow for the opportunity to negotiate more favorable pricing in the future, whether that be through deeper discounts or just a better pricing structure overall. It also allows for your people in the process, whether it be your AP clerks, your approvals team, your, maybe you just have folks dedicated to data entry. It frees them up to do the higher value work within that, within that process to be able to honestly, just have a better work-life balance, right? Be happier with the work that they’re doing, be doing that more critical thinking. Maybe it leads to better employee retention, because you’ve taken someone that’s been bogged down in invoice entry, and you can allow them to do higher value work, right? So, a lot of different soft benefits, as well as some of the harder benefits of the elimination of late payments, the collection of pricing incentives, the improved cash flow visibility, things like that.
So, overall, what we’ve done is really designed and delivered a solution to help modernize that invoice processing process from top to bottom.
And that’s sort of the meat of what we wanted to cover here today. I think what we’ll really get into on our, our Part 2, of the series is kind of how we’ve done this with a customer.
So, I appreciate everyone taking the time to join the webinar today. I would invite everyone to join us again on Wednesday, April 8th, for Part 2 of our Rethinking Invoice Processing series, where we’ll speak to the team at TurnPoint Services, who’ve leveraged the Lydonia proven solution to modernize their invoice processing process.
I will take a look at the chat and see if there were any questions.
I see, does Lydonia have any truly differentiated advantage over others related to invoice processing? Time to value, implementation time, STP, ROI that’s realized. Yeah, that’s a really good question. I would say our differentiated advantage lies in the extraction piece of it itself, utilizing that GDI tool that is only available to Lydonia and Lydonia customers, which gets us that higher straight-through processing percentage. It gets us faster time to value. It allows us to implement faster and be more agile in our implementations. And then lastly, it would help us realize that ROI faster, because we’re able to get that higher STP percentage and turn around those invoices quicker.
Another question I see here, how do we assess whether or not there would be a significant ROI for customers, so we can justify up the chain?
Another really good question, I think, let me see, I think we work with our customers to understand their process where the levers are. How are we going to help them save money, avoid risk, save time, cost avoidance. Those are the types of things that we work with our customers to understand up front and then once we understand what the solution will look like, what the value is, we will then use our value engineering tool, our Lydonia value engineering tool, to showcase what that value may look like, and provide you the tools to go up the chain, to whether it be your CFO or your leadership team, to go justify that solution.
How does Lydonia’s proven invoice processing solution differ from what I have in my Oracle solution?
I would say, again, on the front end, the ability to intake invoices, the back-end solution of maybe we’re pushing information into our payment processing system, and then I would say the overall robustness of the solution that you get with the Lydonia proven solution, as opposed to an Oracle solution is really a huge differentiator. The ability to customize a solution to your process, the ability to escalate externally outside of Oracle when you do need an approval or an escalation, and the ability to handle that complex logic that maybe Oracle’s solution can do, but it’s not necessarily tied to how you do things.
Then I have one other question here. What else do you take into account besides efficiency gains? So the main levers we look at are cost takeout or cost avoidance, revenue generation. It’s going to be risk mitigation as well. Those are kind of the big three things other than just efficiency gains that we see, and all of those can be affected by the invoice processing process.
Alright. And that’s all I see for Q&A. If there’s any other questions in the last couple minutes here, feel free to get them in. And if not, you can feel free to follow up with the Lydonia team afterwards, and we’ll be happy to answer more of your questions, or take a look at your process, and see what kind of value we can provide for you.
So again, looking forward to part two of the webinar, in a couple weeks here, as we talk with TurnPoint Services at how they’ve modernized their invoice processing solution.
But if there are no other questions, thank you everybody again for taking the time out of your day to join the webinar. We greatly appreciate it, and we look forward to talking to you soon. Thank you.
