Banks overlook critical AI infrastructure for scaling

Many financial institutions are overlooking core infrastructure requirements for deploying AI at scale. Despite increased technology spending , 41% of banks lack a centralized or optimized data foundation to support AI initiatives, according to a September report by data and software provider SAS . Data lineage is an often overlooked component when building AI infrastructure, Sathishkumar Palanisamy , expert enterprise architect at Capital One , said at the recent FinAi Banking Summit in Denver. In AI modeling, data lineage refers to the end-to-end tracking of data origins, changes and destinations, according to IBM . “The moment when your model makes a trading decision or flags a transaction, someone is going to ask, ‘Where did the number come from?’” Palanisamy said. “If your answer is going to be vague, then your AI governance is becoming a risk instead of a business advantage.” [caption id="attachment_57383964" align="aligncenter" width="800"] Capital One's Sathishkumar Palanisamy discusses AI architecture at the FinAi Banking Summit in Denver.[/caption] FIs must view data lineage as a core infrastructure component, rather than just a reporting feature of audit trail, he said. “Without that, scaling in banking becomes very tough,” he added. While cloud computing and AI models continue to advance, an insufficient data pipeline could render such advancements useless, Andrew Braun , director of engineering and platforms at the $1.6 billion Grasshopper Bank , said at the event. “If you don't have the gas for your car, it's not going anywhere. The data that's going to make everything interesting and differentiate your content from competitors is the data that resides inside your structured and unstructured datasets and bringing that into AI is really the keys to the kingdom.” — Andrew Braun, Grasshopper Bank Institutional knowledge a must Undocumented institutional knowledge is another hindrance to infrastructure development, Madeline Fredin , vice president of partnership strategy at Alloy Labs , a collaborator with FIs, said at the event. Institutional knowledge is important because it affects how information is stored and the explainability of AI models, she said. For example, an FI may lack a specific policy for how Microsoft Copilot is used, creating inconsistencies and inaccuracies in the system. “That's not an AI problem, that's an infrastructure and procedures problem,” Fredin said. While agentic AI is taking some pressure off institutional knowledge documentation, “the rationale for all of that institutional knowledge that surrounds it is missing in most institutions,” she said. Read coverage from FinAi Banking Summit here .

Banks overlook critical AI infrastructure for scaling

FinAi Banking Summit 2026

Empathy, education, experimentation crucial for AI training

TD executive: AI is raising the bar, not taking jobs

Related Posts

Chief AI officer not needed at nCino right now, CEO says

JPMorgan, AMD, OQC join forces for quantum-AI data center

Compliance, data accessibility will differentiate fintech winners

TD executive: AI is raising the bar, not taking jobs

Stay Informed with Our Newsletters

EMERGING FINTECH DIRECTORY

The Buzz Podcast

SPONSORED

How AI and Product Experts Turn Fuzzy Requirements Into Focused Dev-ready Roadmaps

Is Your Technology Supplier There for You?

Hiding in Plain Sight: How to Use Data to Spot Consumer Accounts Being Used by Small Businesses

Connect

Welcome Back!

Retrieve your password