Array ( [20250723] => 194 [20250724] => 80 [20250725] => 24 [20250726] => 16 [20250727] => 20 [20250728] => 49 [20250729] => 16 [20250730] => 36 [20250731] => 27 [20250801] => 24 [20250802] => 19 [20250803] => 19 [20250804] => 28 [20250805] => 26 [20250806] => 46 [20250807] => 94 [20250808] => 41 [20250809] => 22 [20250810] => 25 [20250811] => 20 [20250812] => 20 [20250813] => 13 [20250814] => 18 [20250815] => 16 [20250816] => 16 [20250817] => 14 [20250818] => 19 [20250819] => 19 [20250820] => 18 [20250821] => 19 [20250822] => 19 [20250823] => 17 [20250824] => 35 [20250825] => 20 [20250826] => 22 [20250827] => 17 [20250828] => 24 [20250829] => 24 [20250830] => 13 [20250831] => 14 [20250901] => 15 [20250902] => 23 [20250903] => 15 )

“AI is not replacing us”: why treasurers are embracing AI on their own terms

Feature-image

Treasury teams are embracing AI to enhance cash visibility, liquidity planning, and FX forecasting—while staying firmly in control of the decisions.

by

Published: July 23rd 2025

Despite the excitement about artificial intelligence, few fields require as much realism about its limitations as corporate treasury. Treasurers are tasked with making judgments with serious financial consequences across currencies, time zones, and uncertain markets. At the heart of this pressure lies a foundational need: visibility. Where is the cash? Where will it be next week? How exposed are we to currency shifts?

Garima Thakur, Global treasurer and risk management leader at Creative Artists Agency (CAA), a global entertainment and sports agency representing artists, athletes and entertainers , sees AI not as a replacement, but as a long-overdue support system for answering those questions faster—and with more confidence.

“AI is not replacing what we do,” she said. “But it is definitely speeding up the parts that used to take days, even weeks as well as taking away the manual work so teams can focus on value-add activities. For lean treasury teams, that matters.”

One of the most promising applications of AI in treasury is improved real-time visibility into global cash positions. By layering machine learning on top of historical transaction data, treasurers can now forecast inflows and outflows with greater accuracy, assuming the data is clean. Treasury teams are using AI models embedded in their treasury management system (TMS) to anticipate liquidity needs across currencies and jurisdictions. However, the accuracy of these models is highly dependent on the accuracy of the historical data and the respective classifications of each cash flow type which is often referenced as ‘tagging’ of cash flows in TMS terminology.

Thakur highlights the risks of overconfidence in the tech. “AI’s only as good as the tagging behind it. If someone mislabels an M&A outflow as payroll, the system will assume you have a huge payroll every May. You’ll end up with projections that are way off, as a general example.”

That’s why many teams, including hers, still keep Excel models running in parallel-for validation and peace of mind. “We’re not ready to fully hand forecasting over to AI. It’s a useful second opinion, not a final answer and not a replacement for human judgement which is critical in cash forecasting.”

 When liquidity planning meets AI

Liquidity planning has historically been a time-consuming exercise involving fragmented systems and lagging data. AI promises a shift to faster, more automated liquidity insights. But the gap between potential and practice remains.

“AI can now offer real-time analysis of financials which is a helpful tool,” Thakur said.

Still, widespread adoption is slow because implementation requires more than just buying software. AI tools require both structured data and sustained oversight.

“Most companies don’t have the luxury of a dedicated data science team,” she pointed out. “Unless AI tools are built into systems we already use and don’t require a major lift to get going, adoption will remain patchy.”

 FX risk management

Nowhere is the balance between AI and human expertise more delicate than in FX hedging. With global exposures fluctuating constantly, treasurers are increasingly turning to AI to forecast and simulate exposure scenarios, especially for volatile currencies.

In Thakur’s experience, AI is most helpful in providing directional insight. “You upload historical exposure, and the system gives you a projection based on trends. That’s incredibly helpful—but not definitive,” she said.

In practice, AI’s output becomes one of several inputs in a broader hedging strategy. “No one’s hedging purely based on what the AI says,” she added. “We might take that forecast and use it to build a layered approach—putting in some hedges now, reassessing next quarter, and so on.”

In sectors where FX exposures change daily, human judgment is still better than machine logic. “I’ve worked in roles where we netted FX exposure down every single day. You had to think about day-of-week patterns, market holidays, even human behaviour to project the right daily exposure amounts. AI is great tool and an additional data point but may not always catch nuances around drivers of historical and projected exposures.”

“There’s still hesitation,” Thakur noted. “Some of it is about data quality, but some of it is just resistance to change. And then there’s the natural question around —what does automation mean for teams?”

She sees a leadership responsibility here. “It’s on treasury leaders to make clear: this is about leveraging technology as a tool while considering the risks and opportunities which will vary by industry and company. It’s about moving our teams away from manual work and toward real value-add analysis and decision-making.”

Even so, a treasurer’s ability to deploy AI depends on broader company policy. “In many firms, you can’t use external AI tools without legal and IT approval. And rightly so—no one wants sensitive data uploaded to the wrong place.”

Some organisations are building in-house versions of large language models to maintain control over their data environment, but this requires significant investment and support from senior leadership.

 What’s next, KYC?

In the future, Thakur sees potential for AI in areas that take time and require patience from treasury teams—especially KYC and bank account management.

“Managing signatories and filling out KYC forms is a huge administrative burden. If AI could assist with tracking and updating global signatories as well as KYC documentation, it would save a ton of time,” she said.

She’s also watching the development of more advanced scenario planning tools that could stress-test exposures under multiple market conditions and suggest a response strategy.

Treasury, not AI technology, leads the transformation

Despite all the sophistication, Thakur insists that treasury—not AI—should remain in control.

“AI can surface patterns we’d miss, but it doesn’t understand the why behind our decisions. That’s the treasurer’s job,” she said. “I don’t think that changes—no matter how good the AI technology gets.”

Is AI a trusted co-pilot in treasury, or still just a high-tech assistant?

We welcome your insights at [email protected] and may include your opinion in a follow-up article.