Banks Don't Trust Their Own Data: Where AI Delivers Value in Payments Today

Quick Summary

Natasha Lapierre, who heads innovation and product strategy for financial messaging at Bottomline, says the keyword when asking where AI delivers value in payments is today. The potential is huge, but most banks are not ready to let AI make payment decisions on its own, largely because they do not fully trust the data they feed it. The use cases going live now are the ones where AI accelerates a human decision rather than replacing it.

What is AI actually good at in payments?

Lapierre starts with the basics. AI is exceptional at a few things: processing huge amounts of information in record time, identifying patterns that a human would find difficult to spot and would typically miss, and surfacing unknown unknowns. Because of that, she says, the range of use cases where AI can bring genuine value is close to infinite. In payments that includes:

• Anomaly detection
• Operational monitoring
• Intelligent orchestration
• Automated alerting

Why are banks holding back from full automation?

The reality today, according to Lapierre, is that most banks are not ready to let AI be fully automated and make actual payment decisions. The industry is at a certain stage of readiness, and the main reason is not the technology itself. It is that banks don't fully trust the data they are feeding AI with in the first place. Until that trust exists, handing over the final decision on a payment is a step too far for most institutions.

Which AI use cases are going live now?

The winners today are the use cases where AI makes a person faster and more accurate, with a human still making the final call. Lapierre gives two examples:

• Improving the accuracy of sanctions and fraud alerts, so that analysts spend their time on the alerts that matter.
• Suggesting how a payment field should be enriched or adjusted to meet a particular country's market infrastructure and rulebook requirements.

In both cases AI does the heavy lifting on information, and a human being decides. Those, she says, are the use cases Bottomline sees going live today.

FF NEWS TAKE:

This is a useful corrective to the agentic hype. Lapierre is not saying AI cannot make payment decisions; she is saying banks will not let it until they trust what goes in. That puts data quality and accessibility ahead of model sophistication on the roadmap. The near-term money is in making analysts and operations teams faster, and the institutions that fix their data first will be the ones ready to automate later.