AI in treasury: Great power and great responsibility

This webinar explores the evolving role of AI in treasury and cash management, highlighting its shift from basic operational automation to strategic decision-making in highly volatile markets. Experts Ernie Humphrey and Sefi Itzkovich discuss the distinct differences between traditional ML and generative AI, emphasizing how these advanced technologies can provide actionable insights, improve complex forecasting scenarios, and mitigate operational risks. The session concludes with actionable advice on adopting AI safely, specifically addressing concerns like data leakage and AI hallucinations.

Watch Webinar

Summary

Navigating Market Uncertainty: Today's VUCA (Volatility, Uncertainty, Complexity, Ambiguity) environment, managing cash flows and understanding counterparty risks is more critical than ever. Scenario analysis and real-time data visibility are no longer optional, but essential capabilities for maintaining business agility and mitigating the impacts of geopolitical shifts and tariffs.
AI Opportunities in Treasury: AI and ML can dramatically streamline financial operations. While machine learning is highly effective at optimizing specific tasks like anomaly detection and cash categorization, generative AI unlocks broader strategic insights by allowing users to query data using natural language and run complex, edge-case forecasting scenarios without requiring extensive model training.‍
Risk Mitigation and Strategic Adoption: AI must be adopted safely and responsibly. Data leakage and AI hallucinations are identified as the primary risks when implementing these tools. To mitigate these threats, the speakers advise cross-referencing AI outputs, establishing a robust, highly-structured data infrastructure, and partnering closely with IT and secure platforms to ensure compliance and protect sensitive financial data.

Target Audience: Finance professionals, treasury professionals, CFOs.

Meet Your Speakers

Sefi Itzkovich

CTO @ Panax

Ernie Humphrey

CEO @ Treasury Webinars

The Risks and and Rewards of AI in Treasury

The current global market is characterized by severe volatility, uncertainty, complexity, and ambiguity (VUCA), making cash management and visibility more critical than ever. Fueled by shifting trade policies, tariffs, and geopolitical unrest, companies are facing unprecedented edge cases and financial turmoil. To navigate this environment, treasury professionals need clear visibility into the "how" and "why" of their cash movements, allowing them to effectively manage counterparty risks and dynamically adapt their business agility.

Artificial Intelligence, particularly Generative AI has emerged as a game-changer for treasury operations by moving beyond the limitations of traditional ML. While traditional ML is highly effective at specific, narrow tasks like anomaly detection and basic transaction categorization, Gen AI provides broader "zero-shot" capabilities that do not require task-specific training. Gen AI can analyze complex interrelationships across markets, enabling multi-scenario forecasting that helps teams predict the impact of sudden events (like doubled tariffs on specific countries) on their supply chain and cash flows.

By automating the heavy lifting of day-to-day operations, AI gives lean finance teams the agility to react to real-time alerts and capitalize on investment opportunities, turning market volatility into a competitive advantage. Furthermore, implementing this technology reduces the manual burden on internal departments, fostering stronger, more strategic relationships with IT, as well as with external suppliers and customers.

However, the adoption of AI in treasury comes with significant risks that must be carefully managed, chief among them being data leakage and AI hallucinations. Financial data is incredibly sensitive, and feeding it into public LLMs can expose proprietary information to other users. To safely harness AI's power, companies must build robust, structured data infrastructures, implement fact-checking layers, and establish secure permission constraints to ensure they are driving decisions with accurate, compliant insights.

‍

Key Takeaways

Cash visibility is paramount in a VUCA environment to understand changing counterparty risks and capitalize on real-time liquidity opportunities.
Generative AI unlocks advanced scenario analysis, allowing finance teams to quickly model complex, unexpected market events and understand interconnected market risks.
Traditional ML is still highly valuable for specific efficiency tasks, such as anomaly detection and labeling historical cash flows.
Data leakage is the biggest risk of AI adoption; sensitive financial data must never be pasted into unsecured, public models where it can be leaked to other users.
Insights are only as good as the underlying data, meaning companies must become data-centric before they can successfully become AI-centric.

‍

Action Steps

Establish a robust data infrastructure: Ensure your financial data is high-quality, properly structured, and continuously validated, as AI models rely entirely on accurate, contextual data.
Implement strict data privacy guardrails: Protect sensitive information by securely managing connections to external data sources and strictly avoiding the use of open, public LLMs for proprietary data.
Deploy fact-checking mechanisms: Actively mitigate AI hallucinations by cross-referencing AI outputs, applying contextual constraints, and using responsible prompting to ensure business decisions are based on accurate insights.
Automate repetitive treasury tasks: Deploy ML and Gen AI to handle cash flow categorization and anomaly detection, freeing up your team to focus on strategic analysis and relationship-building.
Develop a unified company strategy for AI: Partner directly with your IT leaders and specialized technology providers to align on adoption frameworks, compliance requirements (e.g., SOC 1, SOC 2, ISO), and security protocols.

Full Transcript

Ernie Humphrey: Hello everyone. My name is Ernie Humphrey. I will be the moderator for our webinar today, "AI in Treasury: Great Power and Great Responsibility." I am incredibly excited for my conversation today with my co-speaker, Sefi Itzkovich, the CTO of Panax. I'm honored to have Panax as a thought leadership partner; they care about the professional success of those who leverage their solutions, just as I care about the professional success of those who consume my content. Before we jump in, I'm going to ask Sefi to come in and give us a little bit about his background and what Panax does.

Sefi Itzkovich: Hi Ernie, first of all, thank you. Great to be here. I'm Sefi, the CTO of Panax. I'm coming from a background of almost 20 years in tech, mainly in fields of data, big data, and AI, which is what I've been doing for the past 10 years mostly.

Ernie Humphrey: Thank you very much. Before we jump in, we have a few housekeeping items. We will be recording the webinar today, so you will be receiving a recording in your inbox. We have a lot of content to go over, so you'll have the opportunity to review the recording afterward. Feel free to ask questions via chat throughout the webinar, and if we don't get to them, we will reach out directly after. Finally, we are excited to hold a raffle today; we will announce the winner on the Panax LinkedIn page, so just follow Panax on LinkedIn to check if you won.

Now, a little bit about our agenda: we're going to have a conversation. We'll start with uncertainty in the markets and why AI is more important than ever, framed around the importance of cash management and managing counterparty risk. We'll talk about AI opportunities in finance and treasury, as well as AI adoption risks and mitigation. It's very important that we understand how we are deploying AI and all the associated risks, which is why partnering with CTOs like Sefi is crucial for company strategy. Then we'll wrap things up with actionable takeaways.

First of all, why is cash management more important than ever? The world is currently in financial turmoil with a lot of uncertainty in global markets fueled by trade policy, rhetoric, and geopolitical uncertainty. It's more important than ever that we understand our cash movements. We need visibility to understand the "how" and "why" of our cash flows to build relationships with our customers and suppliers. Anything you'd like to add here, Sefi?

Sefi Itzkovich: We see a lot of interest regarding market volatility, especially in the past few months. This is very top of mind for our prospects and customers. In a VUCA environment—where you have volatility and uncertainty—you have to be on top of things. This applies to complex scenarios like forecasting different outcomes, but also simpler day-to-day issues like getting visibility into your data, understanding transactions, and knowing the purpose behind bank movements. This is the value Panax brings: allowing you to make smarter decisions faster.

Ernie Humphrey: I agree wholeheartedly. I love the term "VUCA"—it speaks perfectly to the uncertainty and ambiguity of our current environment. I'll share a quick story: my wife works in procurement, and they were trying to import equipment from France to the US. Originally, there was rhetoric about tariffs on the aluminum parts, but then they decided to charge the tariff on the entire piece of equipment. Imagine having tight manufacturing margins and suddenly facing a 30% tariff. It's incredibly important to have visibility. We are looking at scenarios people never even thought of, and lacking the technology to do scenario analysis can be costly.

Let's level-set on AI. Can you please give us a high-level overview? Touch on the differences between traditional AI, machine learning, and generative AI.

Sefi Itzkovich: It's a very good question. In simple terms, AI is the ability of a system to autonomously make decisions and gain deeper insights. Five years ago, you'd see complex algorithms or ML, but not autonomous decision-making. This is the era of AI we are entering now.

Ernie Humphrey: When I think about machine learning, I think about efficiency—removing manual tasks to free up time for strategic activities. On the generative AI side, I think it's more about understanding the "why" and helping us make better decisions. Can you add some technical context to the Gen AI side?

Sefi Itzkovich: Machine learning is more task-specific. For the past 5-10 years, ML was used to solve narrow problems, like optimizing resource allocation. You needed a lot of data to train it for specific analytical tasks. Gen AI has a broader scope. It uses natural language, and most models don't require specific training; you get "zero-shot" capabilities out of the box. You just ask a general-purpose model a question and get an answer. The potential of Gen AI is that you don't have to solve one problem at a time; you can solve multiple problems with this new technology.

Ernie Humphrey: Does ML learn from a customer's specific data, whereas Gen AI learns from the wider universe of data it's pointed to?

Sefi Itzkovich: That's a very good way of thinking about it. Traditional ML is trained on specific data sources, while Gen AI is trained on up to a million times more data than a traditional ML model.

Ernie Humphrey: Let's talk about smarter financial operations. Would you say categorization of cash flows is a machine learning task?

Sefi Itzkovich: Categorization or labeling is a classic ML task, but moving forward, we will get much better results using Gen AI for that as well. Gen AI shines in complex issues, like multi-scenario forecasting. Using your tariff example: Gen AI helps you quickly predict what happens if tariffs double for specific countries and how that impacts your suppliers and your company. You have to adjust to edge cases very quickly.

Ernie Humphrey: The risk insights are incredibly powerful. Gen AI can analyze the interrelationships between markets and show how a single tariff might expose you to multiple currency or counterparty risks. Data isn't useful unless it impacts decisions. What are some opportunities for AI in treasury?

Sefi Itzkovich: The first major advantage is agility—the ability to react very fast to a changing environment. We see use cases where companies have liquidity gaps or investment opportunities. Missing out on a 4-5% interest rate on millions of dollars is wasted revenue. Getting real-time insights allows you to react fast.

Ernie Humphrey: Volatility brings opportunity. Having cash management agility can give your company a competitive advantage and help build long-term relationships with suppliers and customers.

Sefi Itzkovich: Exactly. AI moves the heavy lifting of day-to-day operations out of the way, allowing lean finance teams to focus on strategic decisions. You get real-time alerts so your analysts don't have to manually grind through data day after day.

Ernie Humphrey: I like the idea of proactive alerts. When the CFO calls asking, "What are we doing?", you are ready. Technology helps us understand our changing risk exposures. We manage risk based on transactions, commodity prices, and counterparties. In a recent survey, I found DSOs are going up, DPOs are going up, and supplier costs are rising. You need agility.

Sefi Itzkovich: AI allows you to analyze risk in ways that usually can't be done manually. This takes a huge burden off our customers' minds. They are specifically concerned about FX exposure, gaining visibility into their AR and AP, and staying on top of hectic collections.

Ernie Humphrey: Investing in treasury technology also improves our relationship with IT. We don't need to constantly pull IT in to manage lockbox feeds and banking portals; instead, we collaborate.

Sefi Itzkovich: IT is often the most overwhelmed department. If you can pass some of that burden to a robust system, IT will definitely be supportive.

Ernie Humphrey: Visibility breeds better relationships. If you have visibility into your data, you can build trust with suppliers and understand the timing and costs of payments. Let's shift to mitigating the risks of new technology.

Sefi Itzkovich The main risks involve the models themselves, specifically AI hallucinations—thinking you are getting the right answer when you aren't. The other massive risk is data leakage and data protection. Finance data is incredibly sensitive. Finally, there is privacy, regulation, and compliance (like SOC 1, SOC 2, ISO).

Ernie Humphrey: What is MCP, and what is the risk there?

Sefi Itzkovichh: MCP gives you the ability to connect external data sources to the LLM. The risk is ensuring the model only receives data with the right permissions. Of all these risks, data leakage is the most important—you have to protect your data cautiously. AI hallucinations are the second biggest risk; if you take action based on wrong organizational insights, you miss the point.

To handle hallucinations, you need fact-checking layers, context constraints, and responsible prompting. To prevent data leakage, you cannot just copy-paste sensitive data into public models like OpenAI or Claude, as they train on your data. A model could remember what one user asked and leak that to another department.

Ernie Humphrey: Are customers pulling you into conversations with their auditors?

Sefi Itzkovich: Yes. Giving finance leaders tools to show and query their data allows them to comfortably answer auditors' questions. Furthermore, your insights are only as good as your data. You have to be a data-centric company before you can be an AI-centric company. You need high-quality, properly modeled data from day one before feeding it into an LLM.

Ernie Humphrey: Does ML automation help with data quality and validation?

Sefi Itzkovich: Definitely. Providers can sometimes send corrupted or missing data, and you have to identify and fix that in real-time. Most finance data is structured, and how you structure it really matters.

Sefi Itzkovich: Let's wrap up. Treasury is more important than ever. Business agility, scenario modeling, visibility, and mitigating risks are crucial. We need a company strategy for AI adoption, engagement with IT leaders, and the ability to ask the right questions of providers like Panax.

Sefi Itzkovich: AI moves fast. If you want to gain the most from it and stay on the cutting edge, you should have a partner to help you. This is where Panax shines—helping customers maximize their cash flow and treasury data.

Ernie Humphrey: Thank you for your amazing insights today, Sefi. Please follow the Panax company page for the raffle and visit their website. Have a great rest of your day, everyone.

Watch Webinar