Artificial intelligence-based language systems have sparked a market rally and attracted a lot of media attention. Checkout.com hopes the technology can also improve payment processing in far less sensational ways.
Checkout.com recently debuted Intelligent Acceptance, a product that uses updated AI to add billions of additional data points tied to payments. The London-based merchant acquirer and processor is using AI to turn data into actionable recommendations, hoping to reduce false declines and cut expenses by automatically routing transactions.
It's also betting AI's booming popularity will draw attention to the service.
"AI today has become a meme. It's trending," said Meron Colbeci, chief product officer at Checkout.com. "Generally speaking, that means merchants may be more open to figuring out how it can be leveraged."
Checkout.com is most immediately concerned about fixing false declines. False declines refer to payments that are mistakenly rejected, often resulting in the customer abandoning the cart — or the merchant — out of frustration.
While false declines are most often associated with a payment being erroneously flagged for insufficient funds, there are a variety of other avoidable reasons a payment can be rejected. A payment can be declined if a consumer is in an unfamiliar location, makes mistakes on the checkout page, picks a shipping window that's less than the normal fraud prevention time or requests that an item be mailed to an address not associated with the card.
"There is tons of data around accepting a payment, and there needs to be a lot of micro decisions," Colbeci said, adding that it's more difficult to do proper vetting as processing speeds increase.
The fixes required to reduce false declines include updating policies or changing the rules that are used to spot signs of an unusual or fraudulent transaction. Updating these rules manually risks falling behind the pace of payment technology, according to Colbeci.
"A merchant may not be seeing enough data for all of its users," Colbeci said. "AI can look across a broad network, take inputs and run tests."
Checkout.com feeds data into an AI engine that supports pre-processing tasks such as messaging and routing, and post-processing tasks such as adaptive retries — which refers to attempting a rejected payment a second time after more data is added and analyzed at the point of sale. The tool is designed to improve as it gathers more data over time.
The product makes decisions on whether to use tokenized account numbers, for example, by determining if the token will be less expensive and more likely to process without a decline than other forms of authentication. It also automatically amends or adds "strong" multifactor customer authentication or other vetting protocols based on the requirements of the merchant, issuer or jurisdiction of a particular payment; and routes debit transactions over the network with the lowest fees.
"While it's improving how to produce more contextual answers for customer service it can also answer internal questions such as how to document payment flows or provide insights for developers that are working on merchant tech," Colbeci said. "AI can be a co-pilot."
Checkout.com's Intelligent Acceptance clients at launch include Klarna, restaurant payment firm Sunday, Ant, NordVPN and cross-border payment firm Reach.
About a third of U.S. consumers say they would permanently stop shopping at a merchant following a false decline, according to a
As
"There is an AI wave," said Stewart Watterson, a strategic analyst at Datos Insights.
AI as a general concept has been used for ten years, maybe more, to vet payments on the fly for suspicious activity, Watterson said. What's changed is the same advancements in AI that are creating buzz in the mainstream are also improving how transaction details and consumer data can be crunched to spot possible fraud or other opportunities to make transactions flow more smoothly — a key task as payments become increasingly instant.
"AI has gotten more sophisticated and more effective," Watterson said. "It's now possible to take large amounts of data and provide much more insight about the origin of that data, and how different data points connect to the larger data set."
