The three largest U.S. core banking providers have been making some important AI decisions. Fiserv is working with OpenAI. FIS has aligned with Anthropic and Jack Henry has announced plans to deploy Google AI across its platform.
Taken together, those three providers serve more than 70 percent of U.S. depository institutions. That is a lot of banks whose future technology options will be influenced by partnerships negotiated somewhere else.
I understand why the vendors are doing it. AI capability is moving quickly and no core provider wants to spend the next few years explaining why its platform fell behind. I also understand why a bank would welcome new capabilities without another costly integration project.
The part that deserves more attention is what comes along for the ride.
A technology choice becomes a bank decision
Think about the conversations you have with your core vendor today. Most begin with functionality, timelines and pricing. AI changes the discussion because a feature may depend on a model, a data service and an infrastructure relationship the bank never selected directly.
A bank can choose whether to activate a particular feature. It may have far less influence over the architecture behind it. Those distinctions matter. If an AI-enabled service draws on customer information, produces a recommendation or drafts a communication, the bank needs to understand the operating boundaries before it goes live.
Jack Henry serves roughly 7,400 community and mid-size institutions. The scale of that installed base makes its Google relationship worth watching. The same principle applies to Fiserv and FIS. Product roadmaps that once centered on core processing and digital channels now carry another set of dependencies.
We have seen this pattern in core transformations for years. The vendor sets much of the available architecture. The bank still owns the decisions about risk appetite, customer outcomes and what it will accept into production.
I would put three topics in writing with the account team.
- DATA: Which information reaches the AI provider? Where is it processed and retained? Can any of it be used for model training? What visibility does the bank receive when the arrangement changes?
- ACCOUNTABILITY: Who investigates an incorrect or biased output? What audit evidence will be available? How are the bank’s responsibilities and the vendor’s responsibilities documented?
- CONTROL: Can the bank disable a capability, restrict a use case or delay an upgrade? What does that choice do to support, pricing and the vendor roadmap?
These are familiar vendor-management questions applied to a technology that can move into sensitive workflows very quickly. They give a board and risk committee something concrete to discuss.
KPMG’s 2026 Banking Technology Survey found that active AI use in financial services rose from 30 percent in 2024 to 75 percent in 2026. PYMNTS also reported that 51 percent of banks were piloting AI agents for productivity. Adoption is broad enough that quarterly discussions alone will struggle to keep pace.
Someone inside the bank needs a current inventory of AI-enabled services and the authority to raise a concern before deployment. Technology, compliance, operations and the business owner need to know where the handoffs sit.
Federal agencies also acknowledged in September that concentration in the core-provider market can limit community banks’ access to information and contract flexibility. That makes clear documentation of vendor requests and responses especially valuable.
The next AI conversation with your core vendor should go beyond the demonstration. Ask about the decisions already being made upstream of your bank and decide which ones your institution needs to own.
That is a practical way to turn a vendor announcement into a governance plan.
Covered in Don’t Blink, my weekly Banking Intelligence Service brief: Issues #24 and Issue #33.
