How Mastercard's using gen AI to combat payment disputes

The card networks are making customer service a major part of their generative artificial intelligence strategies, using the technology to manage the pressure that can come from mistakes or fraud in transactions.

Mastercard this week entered a partnership with Salesforce, integrating Salesforce's financial services cloud with Mastercard's dispute-resolution service. Mastercard's new Salesforce collaboration follows a February upgrade at the card network to add more gen AI technology to dispute resolution. Its rival Visa introduced a gen AI dispute-resolution tool of its own, also in February.

In tests, disputes and call center dispute volume has fallen between 20% and 40% when merchants and issuers leverage Consumer Clarity. Salesforce hopes the Mastercard deployment will deflect about a quarter of call-center queries.

"Getting disputes wrong is a problem because people will stop using their card if they don't feel secure," said Guarav Mittal, CEO and executive vice president of Ethoca, a Mastercard-owned fraud detection service. "For merchants one bad experience can cause a consumer to leave entirely."

New AI technology is a big part of Mastercard's quest to monetize non-payment sources. The use of gen AI with a large data set built over time can differentiate card networks from payment fintechs, Mittal argued.

Mastercard's Ethoca Alerts provide real-time notifications to merchants when a financial institution reports a chargeback. Salesforce connects merchant and payment information to back offices at the relevant card issuer. Payments data is fed into Salesforce's cloud, which manages customer relationship, security, and merchant and issuer services.

Salesforce and Mastercard plan to offer the dispute-resolution service to Salesforce's merchant and e-commerce clients, as well as card issuers.

"There is less likelihood of a payment being rejected with this improved data," Mittal said. "For a sample company that does $40 billion in yearly sales, for example, even a 1% improvement in payment approvals would be a big deal."

Mastercard and Salesforce hope to address what Mastercard says is a growing payment dispute problem. The number of chargebacks is expected to reach 337 million annually by 2026, a 42% increase from 235 million in 2023, according to Ethoca research. E-commerce payment fraud stemming from consumers, or friendly fraud, has become the top form of e-commerce fraud, according to Cybersource, a Visa subsidiary, which reports 39% of merchants have been victimized by friendly fraud.

"Banks and consumers can get a more efficient call center with gen AI," said Eran Agrios, senior vice president and general manager for financial services at Salesforce, adding that the technology includes AI copilots that staff can use to aid queries. "We'll be able to open an [AI copilot] and say, 'tell me about this dispute' and get trusted relevant responses in a moment," Agrios said. "The AI copilots can also take action, something like offering provisional credit while a dispute gets resolved."

Elsewhere in the payments industry, Klarna's gen AI initiative reached nearly 90% of its staff, the Swedish payment and buy now/pay later company announced this week.

Klarna, which did not provide comment by deadline, also reported that gen AI is handling more than 66% of customer service queries, with a 25% drop in repeat queries since the technology launched in January and a reduction of the average resolution time from 11 minutes to two minutes.
Visa's dispute-management product uses technology from the software firm ServiceNow to add automation to the process that finds out who is responsible for fraud, whether a payment is an accidental purchase or an incorrect payment. Visa did not provide comments by deadline. "With gen AI you can accelerate feedback and speed the delivery of transaction information to identify which payments may turn into fraud," Mittal said.

Fraud and customer service are a big part of the financial services industry's plans. Fifty-two percent of financial institutions are using or plan to use gen AI for customer service and fraud prevention, according to Arizent, American Banker's publisher, which surveyed 126 financial services industry leaders in January and February.

"This is reflective of the fact that institutions want to harness new technologies to tackle their challenges impeding growth," Michael Moeser, senior content strategist at Arizent, wrote in a payments research report.

Gen AI is in the process of reinventing everything in the contact center, according to Max Ball, a principal analyst at Forrester.

"Gen AI typically does a much better job summarizing calls than agents did, and the agents are free to move on to the next call," Ball said, adding that customer self-service is where the "real value" is.

The average labor cost for a call-center agent interaction is $5.50, according to Verint, which says an automated interaction costs between 25 and 50 cents.

"The [return on investment] on self-service is huge here," Ball said.