AppZen – Using AI to audit 100 percent of spend, before you pay

A fairly new solution provider to us, AppZen, whose core offering is an AI-based pre-payment spend auditing platform which integrates with almost all SaaS platforms, has actually been around since 2012. But it just came to our attention fully last month when we reported that it had raised $50 million from investors. The company now serves over 1,500 enterprise customers, including four of the top five banks, four of the top media companies, and four of the top ten pharmaceutical manufacturers.

As we know, it is more difficult and time-consuming to reclaim money once it’s left your purse, therefore conducting an audit prior to payment seems an ideal solution. So what does AppZen really do and who does it benefit?

We talked to Anant Kale, CEO and co-founder of AppZen to get the story.

Anant’s background is in software development and engineering, he was Vice President of Applications at Fujitsu America, responsible for the management and delivery of the company’s global enterprise applications and infrastructure. Together with Kunal Verma, who led research teams at Accenture Technology Labs that were responsible for developing AI-based tools for Fortune 500 companies, they founded AppZen. Their combined expertise meant they could develop the platform’s core artificial intelligence technology and bring it to the world of business functions starting with the back office. They focused on building an AI-automated audit function for accounts payable and expenses, using natural language processing, computer vision, semantic analysis, and deep learning.

Said Anant, “In any kind of organisation we found that spend controls and verification processes relied on individuals who were already dealing with huge workloads. We wanted to make these processes, like invoice approval, expense approval, and so on, easier, and we knew there was a better way than relying on someone to do their job well. After all, most people don’t have time to look in detail at all transactions. Our idea, at the highest level, was to combine different key technologies to allow those workflows to be carried out automatically, saving much time and resource, but also to bring deeper benefits.”

He laid out one type of workflow they wanted to automate as an example: “Let’s take invoice approval, traditionally this would involve someone receiving the invoice as an expense or a purchase, cross-checking many documents, verifying everything, and, if they think it’s all ok, sending it to Finance to take a second look. It involves a lot of work to verify that the invoices matches not just the PO, (if there even is a PO) but also the contract, sort it into the correct category (by vendor, by transaction type, by department, by spend volume, etc.), confirm the amount is correct, that it has been fulfilled, look for any anomalies, and sign it off for payment. And, depending on the organisation’s spend thresholds, it might have to then go somewhere else too.”

“So it’s a laborious process, one which we knew we could automate fully. However, the unstructured nature of the data posed challenges. Yet, data can come in so many forms, maybe a paper receipt, an invoice from many channels, a credit card statement, an airline ticket, and so on. The cross-checking to ensure everything is in alignment is also a challenge. For example, does it comply with the contract? Does it comply with our service-level agreement? Do the transactions match up? There are so many documents to review, but existing ERP or P2P systems rely only on structured data to automate those workflows. We saw an opportunity to improve this process by using AI to understand structured and unstructured data, enrich those data with intelligence from internal and external sources, and calculate a risk score for each transaction.”

So what AppZen built into their platform was essentially what became its key strength: the ability to extract information, like currency amounts, merchant names, terms, rates, and other key pieces of data from all documents, and understand it semantically. Whether your invoice is in Coupa, Ariba, Oracle or something else, what might be present in one line, mirroring the PO, could be 10 different lines on the PDF invoice, so being able to understand what these items really are, compared to what’s in the  P2P system and how they match up, is essential. In order to verify data and identify problems, AppZen built hundreds of different models that could legitimise all the data, regardless of the source.

“What we came up with was a concept we call Star Match™,” he said. “It allows us to expand our scope of validation – that means looking at online reviews or checking that the transaction is in line with what other customers are paying. It also means looking at different documents that might not be part of the invoice or in the contract repository, including external systems, such as Google Drive, the internet, and e-commerce sites like Amazon. Our models check and analyse all these platforms simultaneously before payment happens, and raises a flag on anomalies, risk, or potential fraud, so that the control person makes an appropriate decision. What can take hours of work is done instantly.”

So the software acts as a compliance layer sitting atop all these applications, with the flexibility to integrate with every system, even your share drive, p-cards, corporate accounts, expense system and so on, to discover anything that might potentially go wrong. “Because our model integrates with and can extract data from all systems straight from the box, we don’t have to ask companies to change a thing, so they can see the value straightaway.”

As well as the legitimisation of spend, AppZen has another string to its bow

“Firms we deal with are processing billions of transactions across all industries, from high-tech firms to banks, and the risk-related insights our engine generates from millions of customers’ feedback, mainly from auditors, or compliance experts, or financial operators, is propagated throughout the cloud, anonymously for the benefit of other users. And as this happens, in real time, our AI model becomes more and more intelligent, accurate and powerful.”

So what might you learn from these insights? You can form conclusions, forecasts, the state of business spend, and change policies or procedures accordingly as intelligence is gathered. See AppZen’s new report The State of AI in Business Spend.

Why does this interest Procurement?

“Charged with optimising spend, Procurement must ensure that their hard work on negotiated contracts pays off. AppZen ensures that spend complies with a company’s policies, that every invoice is correct, that there are not duplicates, that the volume discounts buried in contracts are applied, that all of Procurement’s supplier master data is accurate and up to date to feed the systems. Overall, we are helping the CPO to maximise their influence on spend, have visibility of how it is being spent, by whom and on which categories, and act before something goes wrong.”


Our analysts at Spend Matters will have a deeper insight into the technology in the coming weeks, look out for that in our Pro series.

In the meantime check out this infographic


Share on Procurious

Discuss this:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.