As artificial intelligence becomes core infrastructure across the payments ecosystem, the conversation has shifted. We’re no longer debating whether to use AI. We’re figuring out how to keep it reliable, explainable and trustworthy when the pace of innovation is outpacing the guardrails around it.

For payments companies, governance isn’t abstract. AI influences whether a transaction is flagged as fraud, how credit risk is assessed and how fast suppliers get paid. A misstep doesn’t just create noise; it creates risk at scale.

What I’ve seen, both inside Billtrust and across the industry, is that most AI governance challenges don’t come from the models themselves. They come from the data. Payments data is messy by nature. It moves through ERPs, banks, suppliers, FinTech partners and procurement systems with their own formats and levels of quality. If we don’t solve for consistency, ownership and lineage before we build the model, the best algorithms in the world can’t save us. Governance breaks down at the point where trusted data should exist but doesn’t.

And that leads to the hardest trade-off: speed. Everyone wants to deploy AI quickly because the efficiency gains are real. But going fast without disciplined data governance is a false victory. You end up paying that debt later in rework, model drift or customer friction. At Billtrust, we’ve learned to slow down at the beginning so we can move faster and more confidently later. It’s not a glamorous answer, but it’s the truth.

Another challenge unique to payments is that so much of the data and model performance depends on third parties. It’s not enough for your own house to be in order if your partners don’t meet the same standards. Governance becomes a collective effort: aligned definitions, shared expectations around data quality, visibility into upstream changes and contractual clarity around how data can and cannot be used. You can’t govern in the dark.

Data governance is not an engineering issue — it’s a companywide discipline. Legal, compliance, operations, product and security all have a stake in how AI behaves. Formalizing that collaboration earlier gives you the structure you need to keep pace with innovation without cutting corners.

Looking ahead, I think boards and CEOs need to push past the broad question of whether AI is being used responsibly and ask something far more specific: Can we explain every AI‑driven decision that affects our customers? If you can’t answer that today, regulators, partners, and customers eventually will ask. And the expectation will be high.

AI will continue to reshape payments. Whether it reshapes it responsibly depends on whether governance evolves with the same urgency as the technology itself.