Good Morning!
Two things dropped this week that every credit union leader needs to know about. Your core banking vendor announced production-ready AI agents. And NCUA confirmed it will use AI to scope your exams before examiners walk in the door.
Neither of these is next year's problem. Both have a clock on them. Let's get into it.
TRENDING AI NEWS FOR CU
Fiserv launches AI agents built natively on your core
Fiserv launched agentOS. An agentic AI operating system built natively across its core, payments, and servicing platforms. Four agents are live at launch: Commercial Loan Onboarding, Daily Operational Analysis and Reporting, Agentic Deposit Intelligence, and Agentic AML Triage Analysis. Two financial institutions are already running them in beta. Wide availability arrives August 2026.
Why it matters for your CU: This is not a roadmap slide. It is a ship date. If your credit union runs on a Fiserv platform, production AI agents are 90 days away. The institutions that arrive in August with governance policies written, data cleaned, and a use case prioritized will go live. Everyone else will spend Q4 getting ready.
NCUA will use AI to scope your next exam before examiners arrive
NCUA's 2026-2030 Strategic Plan commits to deploying AI-assisted examination scoping tools by December 31, 2026. The 5300 Call Report data that your credit union files every quarter becomes the first read. An algorithm will flag where examiners spend their time before anyone walks in the door.
Why it matters for your CU: You now know what the exam looks like before it happens. The model will flag delinquency trends, capital ratios, and concentration risk relative to the peer group. If something looks stressed, you want to find it first. Pull your most recent 5300 and look at it the way a risk algorithm would. The answer to a bad exam is not a better conversation. It is better data.
NCUA confirms federal preemption over Illinois interchange law with 6 weeks to spare
NCUA submitted an interim final rule on May 19 confirming federal law preempts Illinois' Interchange Fee Prohibition Act, which would have blocked federal credit unions from collecting interchange fees on taxes and tips. July 1 is still the effective date. Federal Register publication is pending.
Why it matters for your CU: Illinois is one state. Colorado, Massachusetts, Delaware, and New Jersey have all looked at similar laws. The payments layer of your credit union is becoming a regulatory exposure, and that is a technology and operations problem, not just a policy one.
DEEP DIVE
Waiting for your core to deliver AI was a strategy. It just expired.
Fiserv's agentOS launches in August. AI agents built natively on the core, not bolted on top. Commercial loan onboarding, deposit intelligence, and AML triage are available out of the box.
For credit unions waiting on their core provider to move on to AI, this is the moment they have been waiting for. The problem is that most are not ready to use it.
Three questions that you need to ask:
1. Do you have an AI governance policy that your board has actually approved?
agentOS ships with auditability and policy controls built in. That is good. It does not replace the governance that your NCUA examiner will expect to see. Who approves new AI use cases at your CU? How do you monitor a model after it goes live? What happens when an agent produces something unexpected? If those questions do not have written answers, August is not a deadline you are ready for. Start with the policy. Everything else follows.
2. What has waiting already cost you?
Credit unions running AI on loan onboarding today are closing funding packages in hours, not days. That gap is real, and it compounds. By the time agentOS ships widely, institutions already in production will have a year of tuning, documented outcomes, and examiner-ready audit trails behind them. Version one of anything from a core vendor is rarely the version you want to go live on anyway. The question worth asking your leadership team is not whether agentOS looks good. It does. The question is whether waiting for it is the right call.
3. Is your data actually in shape?
Every AI agent runs on your data. Loan files, member records, transaction histories. If that data is inconsistent or scattered across legacy systems, no platform fixes it. The credit unions seeing the best results from AI spent 60 to 90 days on data cleanup before they deployed anything. That work needs to start before you pick a vendor, not after.
Three things to do this week:
Put an AI governance policy on your next board agenda as a decision item.
Pull your most recent 5300 and look at delinquency trends and concentration risk relative to peers before your examiner does.
Talk to your CIO about data quality before any vendor evaluation starts.
FROM MULTIMODAL
Main Street AI Podcast: How Wescom's Hashim Forrester Lifted First Contact Resolution by 34%
Hashim Forrester: SVP Head of Remote Service Delivery Operations
Hashim has 30 years in operations and service delivery. His team at Wescom Financial is not piloting AI. They are running it, and the results are measurable: 34% better first-contact resolution, the lowest new-hire turnover in four years, and 100% QA coverage, replacing random call sampling.
On this week's episode, he walks through what actually made adoption work, where they drew a hard line on AI touching member data, and the framing every CU operations leader needs to hear.
Want to see what this looks like at your credit union?
Data point this week
4,287
Federally insured credit unions remain as of Q4 2025, down 168 from the year before.
Source: NCUA Q4 2025 Quarterly Data Summary
