They’re an intricate choreography of data, regulation and risk management that must work seamlessly and securely.
As artificial intelligence transforms how money moves, the question confronting global institutions like HSBC is not merely how fast payments can travel, but whether they can be trusted every step of the way.
That challenge, according to Tom Halpin, regional head of global payment solutions at HSBC, is precisely where AI and governance intersect. In a conversation with PYMNTS as part of a series-long dive into B2B payments modernization, Halpin said that “trust is at the heart of payments,” and relates “to [a client’s] reputation. It goes to their ability to make their commitment.”
Defining a Trusted Framework
Halpin described HSBC’s “trusted framework” as the foundation on which its AI strategy rests.
“Transparency and the ability to explain the approach” are “non-negotiable,” Halpin said.
Advertisement: Scroll to Continue
That means clearly documenting model inputs, data sources, training methods and expected outcomes. The bank’s emphasis on data integrity, or ensuring that information is “high quality, unbiased and representative,” is central to governing AI systems effectively, he said.
Equally essential is traceability.
“You have to be able to govern that data … because the old adage is garbage in, garbage out,” Halpin said.
Proper audits, rigorous testing for accuracy and scalability, and continuous monitoring form part of what he termed a “feedback loop” in which AI models learn and humans remain accountable.
“This is not just a technology event,” Halpin said. “This is about technology, governance and people.”
Balancing Regulation and Reach
Operating across dozens of regulatory regimes, HSBC must harmonize compliance while maintaining flexibility.
“Building AI requires us to adapt to local regulation and adhere to our own levels of standards and controls,” Halpin said.
The bank’s answer is a global design principle committee and value-stream approach, ensuring “front-to-back” alignment from cross-border wire systems to machine learning monitoring. That structure “limits the fragmentations and ensures that we deliver the expectations of our regulators, as well as to our clients and ourselves,” he said.
He cited ISO 20022, the global messaging standard now reshaping data-rich payments, as an example of how coordination among regulators and networks can create common approaches, common assessments and oversight.
AI as a Force Multiplier
As real-time domestic and cross-border payments scale, HSBC uses AI to heighten speed and security.
“Some people think it’s to trade off one versus the other,” Halpin said. “It’s not. I don’t think we should ever think about trading off velocity for risk.”
Instead, AI allows HSBC to model variables, like sector assessments, region assessments, time of day, volume and client behavior, to detect anomalies and enhance fraud and sanctions screening in real time.
“We don’t believe AI by itself is a standalone tool,” Halpin said. “We actually use AI and think of AI more as a force multiplier to achieve the risk-velocity objectives.”
Assurance at Scale
Halpin called this strategy “assurance at scale,” which “enables trust in the AI models and ensures fairness, accuracy and proactiveness.”
Human oversight remains integral.
“Most of the AI cases will always have a human in the loop,” he said.
HSBC tracks “error rates, incorrect responses, daily peaks and lows” across clients and geographies to reinforce accuracy and accountability, he said.
The Proof of Trust
Ultimately, the measure of trust lies not in dashboards or algorithms but in customer experience.
“The best proof points that anyone could ever get is actually what the client says directly,” Halpin said.
HSBC backs that feedback with transparent audit logs and digital tools showing how payments flow, pause and resolve. These insights feed the next cycle of model refinement.
“Our business is predicated on trust,” Halpin said. “Clients have been trusting us to make payments. Now they’re trusting us to continue to make payments on their behalf … leveraging the new technologies to their advantage, and not breaking that trust by having data go in different directions or making inappropriate decisions that slow them down.”
For all PYMNTS AI and B2B coverage, subscribe to the daily AI and B2B Newsletters.
