In B2B Invoicing, Late Payments Vanquished Through Machine Learning

The paper chase bedevils efficient buyer/supplier relationships, particularly with large corporates buyers. Previse’s CEO Paul Christensen says the enemy is process when it comes to late payments — an enemy that might be vanquished with machine learning and artificial intelligence, which can automatically ascertain which invoices to pay…in a literal instant.

Imagine walking into Starbucks, getting the coffee, and telling the barista, “Thanks for the coffee … send me an invoice, and I’ll pay you in three months.”

So offered Paul Christensen, CEO of Previse, in an interview with PYMNTS, as an illustration of the traditional ways in which bills get paid.

But late payments can have dire consequences up and down the buyer/supplier relationship, especially the one between large enterprises and a plethora of small suppliers. For large buyers, deciding who to pay and when can be an inefficient paper chase. For small suppliers, waiting for money means a tougher operating environment.

Late payments across the B2B spectrum have worsened over the last 10 or 15 years, as balance sheets were strained during and after the financial crisis and companies sought to optimize working capital.

But there’s a second reason for late payments, said Christensen: “Large companies aren’t evil,” he said. “They are huge, generally bureaucratic organizations that have processes that they find very hard to speed up.”

“The enemy is process,” he told PYMNTS.

That is the focus of Previse, which bears down on process and seeks to speed up the payments decision for buyers, using machine learning and artificial intelligence to get invoices approved for payment – and to get payments disbursed instantly.

Anyone who thinks that late payments represent a deliberate strategy, said Christensen, might be well to re-think that pigeonholing concept.

“It is not economically rational for a large organization to pay late,” he said. These large firms look for – and, in the present environment, enjoy – a low cost of capital, easy access to money and low interest rates.

But against that backdrop, finance departments must grapple with know your customer and AML regulations. Entrenched ways of doing things means that each invoice may be seen by eyes and touched by hand, and signed off on, slowing things down to a crawl at times.

Regardless of the reason, late payments, in effect, represent a practice of borrowing money from a supplier, inefficiently so for all parties involved. The cost of capital that accrues for the smaller party can be in double digits, as they in turn borrow money to keep operations afloat. This is no free loan to the large corporate, either, as it has a very real impact in the B2B arena among smaller firms that can bubble up to their larger corporate brethren.

In the latter case, consider the fact the supplier has to build late payments into their pricing model – which means they may have to charge higher prices, which in turn has a ripple effect through the business landscape, and margins in particular.

Thus: Previse deployment of AI in pursuit of better B2B supplier relationships, invoicing and payments.

“There is a vast amount of very rich data that is sitting there, not being touched,” among large buyers, Christensen told PYMNTS, likening it to “low-hanging fruit, sitting in the back office, that no one has touched.”

Previse, he said, looks at “every invoice” the moment it comes into the system. The firm measures millions of data points about the supplier, multiple features of the invoice, across metrics such as category of spend (think stationery, or travel) or location of the supplier, the frequency of interaction and payments. Previse’s solution has already looked at this information historically and, as Christensen put it, knows with “a very high degree of accuracy” the possible predictors of an invoice being approved to pay or being rejected. In other words, it can discern the likelihood that its large corporate client will pay the suppliers’ invoices.

What comes out is a score/percentage reflecting that likelihood. Call it looking at the past to look at the future.

If the score is above a certain threshold, the payment can be automatic; if it’s below that threshold, the invoice is held back for further examination. Efficiency improves against a backdrop where invoice processing has been traditionally viewed as a tedious, messy task at best, as noted in this space not long ago. Christensen echoed that sentiment, as phone calls tied to invoices can number in the range of eight to a dozen, hardly an efficient use of time.

The number of invoices that ultimately prove problematic, said the executive, is rather small – and yet, the traditional efforts that go into screening all the invoices, from phone calls to paper chases, are costly to the buyer because they run the gamut across all invoices. Now the sifting is automatic, and only those invoices that should be re-examined are pulled for further scrutiny.

Where payments had once been 60 or 90 or 120 days, the number is now zero. If the score is above a pre-set threshold, payment to the supplier is assured. “The supplier can send their invoice in the morning and have money in their account that evening.” Even as cash flow improves to the supplier, it does not change for the buyer. If the latter wants to pay in 60 days as they might normally do, for example, they can do so (there’s a bank in the middle of the buyer and supplier, and the Previse relationship is funding the payments).

Asked by PYMNTS for a near-term roadmap, Christensen noted that the funds raised this past summer have been allocated to building its team. They currently have 13 people on staff, and expect to reach 30 by the end of this year. Previse is poised to go live with its AI technology, with an eye on supporting roughly half a dozen very large corporate clients this year through incremental rollouts, initially in the U.K. and the U.S.