By Ray Birch
TUCSON, Ariz.—When the next wave of payments innovation hits full throttle, credit unions may find themselves both in the driver’s seat and in the danger zone — especially on the topic of chargebacks.
The advent of agentic artificial intelligence promises a major overhaul of payments and transaction workflows. But for the back-office world of chargebacks, the implications are complex: operational relief, new revenue streams — and yet also new risk, cost, and labor burdens.
For credit unions, the message is clear: this is coming, and they need to prepare, says David Martinez, founder of consultancy Core Credit Advisors.
What Is Agentic AI — And Why It Matters For Payments
Agentic AI refers to autonomous systems that not only assist but act on behalf of users — selecting, scheduling and executing transactions without continuous human prompting. In the payments environment, that might mean subscriptions, recurring purchases or automated replenishment of goods and services that happen without direct user intervention each time.
“The way I think about agentic AI is more like supercharged subscriptions. You have your monthly Spotify bill, your monthly cell phone bill … With agentic AI, everything else that you may go to Costco for, you may go to Walmart for — potentially could be automated in terms of when it’s purchased and that shift massively expands the volume of automated transactions. That’s where the real opportunity comes in,” Martinez explained.
For payments, migrating more of the transaction flow into automated, AI-driven processes means deeper engagement, higher volume, greater stickiness of a credit union’s card or payments product. For credit unions, that suggests a positive: more transactions, more interchange, more product usage, Martinez said.
“If you don’t catch up or you don’t maintain … the credit unions will fall back. They’ll lose revenue. They’ll lose usage,” he said.
So, the race is on. But with this opportunity comes new risk — and in the chargeback world, that risk may show up early, Martinez cautioned.
Agentic AI’s Impact On Chargebacks — The Positive Side
From a favorable perspective, Martinez explained how agentic AI could help credit-union issuers and processors reduce chargeback exposure and cost in several ways:
-
Stronger authentication and intent capture: When AI orchestrates recurring or automated purchases, the transaction logic can incorporate known member preferences, permissions, and “agent” consent flows. That means fewer situations where the cardholder says “I didn’t buy this” when in fact the AI agent did it with consent. In time, that could reduce friendly fraud and thus chargebacks.
- Improved data and context: Agentic payments can carry richer metadata (e.g., “automated monthly replenishment of X”, “AI-scheduled re-order of detergent”), which gives card issuers better signals and dispute context, potentially reducing representment losses (when issuers or credit unions challenge chargebacks).
- Higher transaction volume, more revenue cushion: More activity means greater opportunity to spread fixed chargeback-cost burdens across larger volume, making chargeback cost per transaction lower. Martinez frames it as growing transactions via AI = more revenue = stronger CU position.
- Operational efficiencies: AI may help in the back-office (not just front-end) by automating dispute-triage workflows, smoothing resolution, lowering labor burden (which is a major cost for issuers).
In short, if embraced, agentic AI could help credit unions both grow and protect their margins from the chargeback drag, Martinez said.
Agentic AI’s Impact On Chargebacks — The Risk Side
But this isn’t all upside. Martinez flags several risk vectors credit unions must be aware of:
- Friendly fraud spike: When payments become automated, it becomes easier for cardholders to remove visibility (“I forgot I set it to buy detergent) and then dispute the charge. “Anytime technology takes off there tends to be less rails than more rails. Some bad actors may take advantage. Additionally it’s going to cause a lot more friendly-fraud—those chargebacks where I forgot I set it to buy detergent … I mistakenly either reach out to the merchant or, more likely, reach out to my issuer and say, ‘Hey I didn’t buy this.’” Martinez estimates credit unions may see a 5-10% increase in chargebacks as this starts to roll out.
- Operational burden and cost: The back-office of chargeback management is labor-intensive. Recent data show that for issuers each dispute costs about $9.08-$10.32 to process in the U.S. on average. That means more transactions + more automation + more disputes = more cost if not managed, and additional labor, compliance, technology ramp up.
- Regulatory uncertainty and consent/agent risk: When an AI agent acts on a member’s behalf, the question arises: was proper consent captured? Are the terms clear? If not, then disputes will rise and the issuer will bear the cost.
- New fraud vectors and “agent mistakes”: Automated purchase logic may lead to unintended or unexpected charges — an AI agent might order something the member didn’t anticipate, and the member might dispute it. That undermines the transaction experience and increases chargebacks.
- Technology and implementation gap: Martinez cautions that until industry protocols fully align (issuers, networks, processors), the early phase will see more risk: “Initially unless everyone rolls out the protocols in the same fashion, friendly fraud will go up,” he said.
What Credit Unions Need To Do Now
Given the dual nature of this shift — opportunity and risk — credit unions should act proactively, Martinez said, sharing steps to take:
- Understand the CU’s payments and transaction flows
- Map out where agentic AI or automated purchasing touches the credit-union card/loan members today (subscriptions, automated replenishment, in-device payments).
- Review how the CU’s issuer-processor handles recurring/automated purchases, agent-based consent flows, member notifications, and dispute-data capture.
- Engage with the processor, network and card-strategy partner
- Martinez’s main message: “Make sure you stay up to date. … Know what your processors are doing … Stay in contact with your Visa/Mastercard rep … ensure that you’re front of the line for those protocols when they go live.”
- Upgrade back-office and dispute workflows
- Prepare for higher volume and perhaps higher complexity of disputes.
- Invest (or plan) in enhanced dispute triage, AI-powered analytics, intent-validation tools, and member-notification systems that pre-empt friendly fraud.
- Design member-friendly controls and transparency
- For transactions executed by an AI agent, ensure members receive clear notifications, have easy visibility of what the agent purchased, and can opt-out or dispute early.
Agentic AI in payments isn’t an abstract future — it’s in motion now. Credit unions should treat the next 12-24 months as a critical window, Martinez said.
“It’s coming,” he stressed. “For credit unions, agentic AI becomes real the moment automated purchases begin flowing through your channels. You may see an increase in small, recurring transactions and a corresponding rise in ‘I don’t recognize this’ disputes. That shifts operational focus from traditional fraud reviews to updating dispute processes, monitoring AI-driven purchase patterns, and working with processors and networks on new consent and metadata standards. In short, agentic AI changes transaction volume, dispute behavior, and the operational demands placed on the entire credit union.”
This report is part 1 of a two-part CUToday.info series.