Where should a treasury team start with AI?
Start with a small number of recurring treasury decisions. Map the data, process and judgement required for each one, then target the parts where AI can reduce preparation work or improve the information available to the decision-maker.
AI will change treasury work faster than it removes the treasury role
The AI discussion often jumps from automation to job replacement. Treasury is a poor fit for that simple story because much of the function combines information processing with judgement and accountability.
AI can take on more of the work involved in gathering, cleaning and consolidating data. It can support forecasting, prepare exposure information and help users retrieve the evidence needed for a decision.
The remaining work becomes more valuable, not less: deciding what the information means, setting risk appetite, responding to events that do not fit the historical pattern and taking responsibility for the outcome.
Treasury needs to solve data and process problems before expecting strong AI results
A treasury function can have sophisticated AI and still get weak results if the underlying data is fragmented or the process itself is inefficient.
A useful way to think about readiness is in layers. First comes the data: where it sits, whether it is consistent and whether the organisation can combine it. Then comes the process: how information moves, where latency appears and which manual steps exist only because they have always been there.
AI becomes most useful once those two layers are sufficiently controlled. Otherwise it can produce confident answers from weak inputs and simply accelerate an inefficient process.
AI works well as an automated analyst
The near-term role of AI in treasury is easier to understand when it is treated as an analyst rather than an autonomous decision-maker.
It can prepare exposure data, gather information, generate an initial view and identify areas that deserve attention. The treasurer remains responsible for evaluating that work and making the material decision.
As confidence in the technology increases, some lower-risk and more standardised decisions may move further toward automation. Event-driven and judgement-heavy decisions are likely to remain human-led for much longer.
Most of the effort in FX risk management happens before the decision
Sander van Tol offers a useful ballpark for FX exposure management: roughly half of the effort can sit in getting the data right, another large share in forecasting future exposures, and a much smaller share in the final decision.
The exact percentages will vary by company, but the direction is useful. AI is strongest in the parts that consume the most preparation time: retrieving data, processing it and supporting a forecast.
The final hedge decision may take less time, yet it carries more responsibility. That is precisely where context and human judgement still matter most.
Operational treasury skills will become less central
A large part of today's treasury workload still involves reconciliation, payment reformatting, exposure gathering, reporting and copying information between systems.
Those tasks are natural candidates for further automation. The skill of being the person who knows how to assemble a complex spreadsheet may therefore become less valuable as systems handle more of the preparation.
That does not reduce the value of treasury expertise. It changes which expertise matters.
Data literacy and judgement become core treasury skills
The treasurer of an AI-enabled finance function does not need to become a data scientist. Data literacy is more useful: understanding where information comes from, how it is structured, which definitions are being used and what can reasonably be inferred from it.
Structured analytical thinking also becomes more important because users need to break a business problem into the decisions, data and process steps behind it.
The third skill is influence. Treasury still has to communicate with senior management, banks and business stakeholders and make decisions where the answer depends on context rather than calculation alone.
Some decisions remain irreducibly human
Risk appetite is one example. A model can estimate exposures and simulate outcomes, but the organisation still has to decide which risks it is prepared to accept.
Relationship decisions are another. The number and type of banking relationships, key suppliers and strategic partners cannot be determined by a model without reference to broader corporate priorities.
Crisis situations add a further limit. When a war, pandemic, market failure or other event changes the context itself, historical patterns become less reliable and management judgement becomes more important.
Start with the decisions treasury makes every day
A practical AI roadmap does not need to begin with a large programme across every treasury workflow.
A better diagnostic is to identify the three to five decisions the team makes most often. For each one, ask what data is required, how that data is gathered, which manual steps slow the process and where judgement actually enters.
That makes it easier to separate a data problem from a process problem and identify smaller use cases that can create value without placing the entire function inside one experimental project.
The treasury team may stay similar in size while the work changes
Automation can reduce some roles in individual organisations, but it does not automatically follow that treasury as a function becomes smaller everywhere.
Companies can use the capacity created by automation in different ways. Some will reduce operating cost. Others will redirect people toward financial risk, business partnership and decisions that previously received too little attention.
The more durable change is therefore likely to be the composition of treasury work: less manual processing, more interpretation, judgement and stakeholder influence.
About the interview
This article is based on Nomentia's interview with Sander van Tol, Partner at Zanders and co-responsible for its Treasury Advisory Group. With nearly 30 years of experience, van Tol advises global organisations on corporate treasury, risk management and corporate finance.