Despite heavy investment by banks and fintechs into AI pilots, 95% of them never reach production, says Ben Goldin, Founder and CEO of Plumery, an AI-assisted digital banking development platform.
“The reason AI projects fail is the same reason that other engagement-centric products fail,” says Ben. “The underlying data isn’t in a format that the AI expects it to be in, or it’s not structured properly, or it’s incorrectly exposed.”
Ben has spent nearly 30 years building banking software, and he’s noticed that banks consistently struggle to differentiate their product-centric systems from customer-centric ones.
Core banking is designed to do core banking tasks: processing transactions rapidly, handling loans, savings accounts, current accounts, calculating interest, maintaining balances, and all the vital things to keep a bank going. This underlying banking data is unfortunately not suited for building engaging systems from the user’s perspective. They’re a system of record, not a system of engagement.
One of Plumery’s main tasks is transforming raw core banking data into suitable formats for building systems for end users.
Last chance: Get your tickets for Baltic Fintech Days May 13
Replacing a core banking system is usually infeasible
“Although it’s certainly valuable to add a modern core banking platform, you shouldn’t really start by replacing a 40-year-old legacy system,” Ben says. “Instead, there should be an architecture that supports a seamless coexistence of modern and legacy without negatively impacting the customer experience.”
Plumery provides modular solutions that transform core banking information into engagement-friendly workflows. The solutions allow banks to “plug in” components on top of their core banking to handle engagement-centric needs. For example, Plumery’s web and mobile banking modules can handle inter-account transfers across core banking hubs seamlessly, as well as deal with user onboarding.
“Plumery helps banks and other financial companies build delightful, modern experiences faster, and with a great level of flexibility to differentiate what matters the most for their users,” Ben says.
Demand for AI integration
A report by McKinsey says that any bank or fintech that doesn’t implement AI is expected to shrink by 9%, says Ben.
Shackled by legacy limitations, some banks struggle to implement this cutting-edge technology beyond simple customer support use cases. However, using Plumery’s technology, at least one bank has managed to implement a conversational banking tool.
“Banks tried conversational banking before, but the tools suffered from instructional rigidity and the inability to deduce meaning from context,” says Ben. “They also operated in a siloed manner without access to things such as the user’s financial history.”
Using Plumery’s data transformation and modular capabilities, banks and fintechs can solve the data problem directly, thus providing AI solutions that have better chances of succeeding.
Last chance: Get your tickets for Baltic Fintech Days May 13
Plumery’s architecture is perfectly suited for AI integration
Plumery was already structuring and exposing core banking data through APIs before generative AI came along. Now, instead of only APIs, it can also expose data through MCP (Model Context Protocol), the open-source standard for allowing LLMs to access underlying data sources.
“Using the same modular format we’ve always provided, banks can now also integrate AI into their systems without having to rebuild the entire core,” says Ben.
However, before embarking on any AI project, Ben recommends that banks and fintechs ask themselves three questions:
- Can they measure the outcomes?
- Can they set clear objectives before starting, so that once they’ve implemented it, they can validate if they’re hitting those objectives?
- How future-proof is the architecture that they’re putting together, considering the immense speed at which AI solutions evolve?
