Disputes are a vulnerable, human moment. When an unrecognized payment appears, the consequences could be as mundane as an accidental repeated charge, or as serious as a full account takeover attack. In this age of digital banking, it’s also one of the only person-to-person touchpoints account holders will have with their financial institution, raising the pressure and opportunity of the moment. During disputes, people expect their financial institution to be transparent about the decision, handle it quickly, and most importantly, to take care of them.
At the same time, disputes are one of the best use cases for automation and artificial intelligence in financial services. The processes are ripe for automation, and the heavy information gathering presents an opportunity for AI systems to take care of the manual work.
It would be easy to see this as a paradox: using cold technology to support an emotional human moment in banking. But if we use the right automation and artificial intelligence, while human judgment leads the strategy and decision making, banks and credit unions can improve their dispute processes, build more trust with their account holders, and meet the rising expectations.
So where does AI belong in the dispute process? And where do the human experts need to stay in charge?
Simple cases are an automation problem
If a merchant is going to refund the charge, nobody needs to waste their time deciding whether it’s fraud. In this case, an account holder is looking for a quick fix, and a person in the middle only adds delay. That’s low-hanging fruit for automation and, in most cases, doesn’t need AI at all.
However, AI can be used to make that automation stronger and more robust. It can study an institution’s history to find patterns within the data. It might detect that a bank approves a certain kind of case every time and makes a recommendation to automate it.
AI can also be used to catch account holders who are testing the rules. Frequent disputers trying to game the system might figure out the dollar amount where a bank auto-resolves a claim, then file just under it, again and again. They’re turning a dispute system created for convenience into a ‘free money button.’ While some institutions might just shut the system down completely, or absorb the losses, AI can spot those abusing the system and pull those cases in front of an investigator to improve automation rules, without sacrificing efficiency.
Complex cases need a person, and a prepared one
But not every case is a simple one. In that moment, the best thing a bank can do is show up as a human, not a black box.
Still, AI can be used with purpose to improve the process and meet the account holder expectations we outlined earlier: transparency, speed, and care. AI can upgrade the intake process, surfacing the right questions, so an intake agent captures a complete dispute the first time. For analysts, AI can surface key signals in gathering the facts of the case, such as atypical spending, unusual location, and dispute history. This data gathering allows analysts to focus on what matters in the decision. AI acts like a junior analyst, cutting out the legwork while preserving the power of verdict for the human.
Think of AI as adaptive cruise control and lane assist, not a self-driving car. The system can hold your speed, keep a safe distance and nudge you back into your lane, but the investigator’s hands stay on the wheel. That framing changes the design question from what the model is capable of to what the tool is allowed to do.
What AI should not do is tell the investigator exactly what to decide. In talking to banks and credit unions, investigators and analysts ask to see the levers, not the end result. They want to combat any bias from an automated decision, leaving the final call to the experts. Most AI models are limited to the data they have access to, whereas the investigator might have additional context. AI’s job is to bring the research file, not the verdict.
How financial institutions find the right balance
Every institution has a different strategy around fraud and disputes to meet their goals. Picture the options as a sliding scale rather than an either-or choice. At one end is earning the account holder’s loyalty. At the other is reducing losses and the cost of working disputes. Every institution is chasing both, but some lean toward protecting the account holder relationship, and some lean toward stopping more fraud and protecting the institution. There isn’t one right answer for every institution, which is why the technology needs to be configurable and adaptable.
Ultimately, disputes are more than just an operations play. They’re about handling a high-stakes moment thoughtfully, building account holder trust, and keeping the investigator’s hands on the wheel.
Iris Friedman is a Director of Product Management at Q2, where she leads product for Centrix DTS, Q2’s dispute and fraud claims platform. Over 500 financial institutions use Centrix DTS to take in and resolve disputed transactions across card, ACH, Zelle, check, ATM, and other payment types. Iris has been developing AI for financial institutions since 2016. She works closely with banks and credit unions to understand their needs, modernizing fraud solutions to be intelligent, human-centered, and useful to the teams who protect customers from fraud.
