Can AI be used to solve the expertise gap in trade finance operations?

Editor's note: This essay first appeared in Trade Treasury Payments. We're republishing it here for our community of trade finance professionals working through the same expertise transition. → Read the original on Trade Treasury Payments

---

I remember my final years while working in the banking sector when the looming expertise crisis in trade finance operations was already being discussed in whispered conversations and behind closed doors.

Fast forward a few years, and this issue has finally moved into the spotlight as one of the sector's most significant challenges. Unfortunately, the fact that we're now discussing it openly simply indicates that the problem has become more urgent and that we've failed to address it effectively.

I won't explore the various reasons that got us in this situation today, but rather focus on one of the many proposed solutions that has gained considerable traction over the past years: AI as a replacement for traditional job roles.

For years, we've heard about technology's potential to replace human workers across various fields. While this threat has been far from realized, it has grown louder and more concerning in recent years with the emergence of the most advanced AI technology to date. Those more technically inclined may have noticed the shift from previous AI technologies, such as machine learning (ML), toward the latest applications of generative AI including large language models (LLMs).

What LLMs can deliver surpasses what earlier ML techniques could achieve by significant margins. For the first time, the concept of 'robots' replacing humans in trade finance has become a genuine discussion point. This explains why we're now witnessing debates about this topic across specialized and mainstream media, among tech leaders, and even within government circles.

However, what may seem like a threat to some, particularly in certain industries, could actually be a blessing in the trade finance sector, where this technology presents a perfect opportunity to address the pressing expertise crisis I mentioned earlier.

Although I consider myself a technology advocate, and having worked with these technologies for several years, and as part of Complidata, a leading company in LLM technology for the trade finance sector, I don't believe this represents the sole solution we can depend on to address the expertise crisis.

To understand this, we need to move past the hype, noise, and apocalyptic predictions, so that we can examine AI's actual capabilities in addressing this issue. While these capabilities are numerous, we must also understand AI's limitations and why Trade Finance still needs to develop and rely on its human workforce for some of the most critical tasks in the field.

AI excels at many of the repetitive and rule-based tasks that consume significant time from document examiners, for instance. AI can flag anomalies in a 50-page document set faster than any human, or detect patterns across multiple transactions that might otherwise go unnoticed: a shipping container reused across unrelated trades, an unusual payment routing, or an invoice date that doesn't align with the bill of lading.

AI also proves effective when operating from clearly defined checklists: identifying red flags based on pre-set criteria or validating structured data against expected norms. These capabilities are particularly valuable as transaction volumes grow and compliance demands increase.

These improvements represent a massive opportunity for efficiency gains: higher output at lower costs, improved business performance, and more effective teams. Essentially, achieving much more with much less. Hasn't this always been the holy grail for businesses worldwide?

However, efficiency is not expertise. Pattern recognition is not the same as insight. This is where the limitations begin to emerge.

The boundary between automation and judgment is where the need for expert human adjudication comes in. While AI can identify potential issues like a mismatch in document data, a suspicious trade route, or a flagged counterparty, it cannot determine yet whether that issue matters in context.

Is the discrepancy material? Is the risk acceptable? These are not yes/no questions with clear answers. They require experience, nuance, and an understanding of the commercial relationship behind the transaction. This kind of decision-making remains firmly in human territory.

AI can raise the flag, but it can't decide whether to waive, accept, or escalate a risk. That's still up to human judgement.

Even then, let's not forget that for today's implementation of these new technologies to succeed, we need not only AI tools empowering humans, but also humans empowering AI with their knowledge and hard-earned expertise of many years in the sector. The same expertise we risk losing with the retiring of experienced professionals and the lack of new-joiners.

Many of the most seasoned trade professionals are approaching retirement, and there aren't enough new recruits coming up behind them. The work is specialised, complex, and rarely taught in schools. And over the past decade, investment in structured training, particularly rotational programs between product and operations, has quietly disappeared.

The answer isn't to replace people with AI, but to rethink how we grow people and how AI can support that goal. If we're serious about solving the expertise gap, we need to treat training as core infrastructure, not a nice-to-have.

That could mean bringing back apprenticeships, pairing junior staff with senior mentors, or working with external bodies like the Walbrook Institute London (formerly the London Institute of Banking and Finance, LIBF) to provide accredited learning. For smaller banks, this is especially critical. They can't afford to wait for experts to appear. They have to create them.

This is also where explainable AI plays a role. When used thoughtfully, tools like TradeSpeed don't just automate document checks; they help junior staff understand why something was flagged and what to do next. That kind of support makes it possible to scale expertise, not just output.

The goal should be to free up human capacity for more value-driven tasks, working with clients, solving complex cases, and growing into senior roles. Efficiency is good. Capability is better. And unless we start investing in both, the expertise crisis will only deepen.

There are a few questions we should all be asking. How do we attract young people into trade finance operations? How do we grow and develop the junior staff we already have, and help them become tomorrow's senior experts? How do we do this fast enough, while the experienced professionals are still here to pass on their knowledge? What can be done at the industry level, and what needs to happen inside each institution?

And perhaps most importantly: how can we make the job itself more exciting and meaningful?

What worries me is the pace. Adoption is still slow. Too slow, too cautious. If we take five to ten years to address this, we may miss our window. Even with the help of AI, we still need experts, and they are leaving.

Continue the conversation

If this resonated, our webinar Solving the Expertise Crisis in Trade Finance Operations covers the practical side: what banks are actually doing, what's working, what isn't.

---

Ben Arber CEO Ben was the Head of Financial Crime Compliance, Commercial Banking and Global Trade and Receivables Finance Head at HSBC in North America and Asia Pacific. He is co-chair of the ITFA financial crime compliance initiative and has a keen involvement in AI & blockchain technology companies. He is a frequent speaker at SIBOS, GTR, ACAMS, BAFT and other industry forums.