Banks Face 18-Month Window for AI Payment Overhaul

Competition has become another consideration as consumers use FinTechs and neobanks for more of their primary banking relationships. Reis said he is wary of “lift and shift” projects that transfer existing applications to the cloud without changing their underlying architecture. Such migrations may lower hosting expenses. They leave much of the original technical and functional debt intact.

Modernization includes shorter software release cycles, lower operating costs and simpler technologies. Reis contrasted release cycles measured in weeks with legacy environments where software may be released only twice a year.

Changing that architecture creates its own hazards. Reis identified integration complexity as one of the major risks, particularly when banks attempt to connect new technology to a dense collection of existing systems.

Banks should also manage migration risk, Reis said. A complete replacement may be necessary in some circumstances. Other institutions can migrate individual portfolios over several years. Another approach separates the customer-facing experience from the back end, allowing banks to modernize specific use cases while older and newer infrastructure coexist.

The front end can change before every component underneath it has been replaced. Back-end functions can then migrate individually, reducing the operational exposure associated with a single large conversion.

AI purchasing, tokenization and passwordless authentication will add new requirements to payment systems. Banks that carry every legacy product variation, integration and decision process into the new environment may preserve many of the constraints they intended to remove. Reis said the 18-month timeline means issuers need to start simplifying now.