CUSTOMER STORY

Invoice errors are inevitable when you process a high volume of invoices, even with a world-class AP team and tools.

Operating 300 branches and stocking 160,000 parts, the finance team at Euro Car Parts – the UK’s leading distributor of car parts – were painfully aware they needed more control against the risk of invoice errors.

Baz Durrani is ECP’s Continuous Improvement Manager. Having joined in 2018, his first remit was to review and improve the Procure to Pay system. The company had a basic process that created many inaccurate results (false positives). It was immensely time-consuming, and the real risks were hidden in the noise of false positives.

“90% of our invoice processing used OCR scanning technology. It would scan each PDF for key data and send the XML file to our ERP system. Far from perfect technology, it was heavily dependent on the quality of data coming in. For example, scanning errors could occur when an ‘s’ was read as a 5, or an 8 was interpreted as a zero.

We only had three variables to work with when checking for duplicates: invoice number, invoice date, and invoice value. Whenever we looked for duplicates, we ended up with loads of false positives. In fact, it was one person’s full-time job to process the potential problems,” said Baz.

Seeking to reduce false positives

Like 90% of ERP solutions, the system in place at ECP could only check for exact match duplicates, leaving the business at risk of errors and threats.

To better capture indirect invoices, ECP had already added Coupa to its technology stack. Whilst this had given ECP better control, it also was only capable of highlighting exact match duplicates.

Keen to address this problem, Baz first looked at an in-house solution using queries in Microsoft Access. But there was too much reliance on manual checks, and ECP was still limited to exact matches.

Baz explained: “We downloaded some of our incoming documents to Microsoft Access, still only able to work with the same three variables: invoice number, date, and value. But Access could only identify a duplicate where, for example, it discovered 1234 twice. It wouldn’t pick up an entry of 1235 and question it. There was no intelligence to achieve that level of checking.

At one point, we tried looking at invoice value only. But that dramatically increased suspected duplicates. For some suppliers, every invoice can be the same value. So, someone had to physically check every single one, which could create further human errors.”

Discovering the value of AI

When Baz came across Xelix, the difference was immediately evident.

“Straightaway Xelix was visually appealing. But the thing that got us most was how it learned. For example, it learnt invoice patterns, meaning our duplicates would reduce rapidly. It didn’t waste our time.”

“But our Project Management Office initially said we already have two other duplicate checks in place, so why do we need another? That’s when we carried out a proof-of-concept exercise with Xelix.

We gave them a sample of our data and they gave us ten potential duplicates the AI platform had flagged up. Eight out of ten were duplicates - to the value of £50,000. That was enough to win our board over. It was a no- brainer. We knew we’d prevent far more with Xelix – even with Coupa in place,” Baz explained.

Setting up Xelix was relatively easy, said Baz. Flexible in the files required, it could have been up and running in a week. Babbage proved to be a headache for ECP as they worked to extract data from “an ERP that’s not very user-friendly”, but the Xelix team were on hand to help.

Two hours to ten minutes

Having worked with Xelix for some time, Baz is crystal clear on the transformation it’s delivered for ECP. And he explained how that started from day one.

“We first did a retrospective duplicate checking exercise with documents from the last two years. This resulted in us recovering £250,000 on day one. It’s a huge return and covered the cost of Xelix for years to come.”

A second transformation concerned staff efficiency and impact: “Because Xelix all but eliminated false positives, they were only looking at the realistic duplicates. This reduced the workload from two hours to ten minutes a day.”

They now have time to tackle the root causes of any duplicates to stop them from happening again. It’s been a huge saving for us. Everyone quickly bought in to something that would put an end to firefighting and make the job much easier.”

And it’s not just duplicate checking. ECP is successfully using Xelix to identify ledger errors, such as incorrect VAT and invoices posted to the wrong supplier. For example, using self-learning, Xelix flags when a supplier or an account code looks out of place.

Another benefit is how Xelix continually reviews master and transactional data, helping to keep it clean.

“Xelix gives you that master data cleanse tool, which is a big challenge for any business. It’ll flag supplier duplicates and little-used vendors. And it’ll spot potential errors with account coding,” said Baz.

The way to reduce your risks

Whilst some organisations fear the onset of AI, Baz is quite clear it’s “the way forward” because it eliminates the risk of human error and is self- learning. But what about jobs?

“You shouldn’t always see AI as a threat to people losing their jobs. We’ve demonstrated that clearly, utilising the saved resources elsewhere in the business. So, we’ve had more resource to tackle the root causes of duplicates and radically improve our processes.” “Xelix is a simple, reasonably priced solution that reduces your risk using AI and self-learning. It’s honestly a no brainer to me.”