Corporate treasury professionals are moving beyond experimentation with artificial intelligence to real use cases. Current AI adoption ranges from basic process automation to advanced machine learning models and custom AI agents.

To facilitate peer learning on applications of AI in treasury, Tom Hunt, CTP, Director of Treasury Practice at the Association for Financial Professionals, convened a discussion group. Below is a look at how treasury professionals are deploying AI and automation across foreign exchange, cash forecasting, fraud detection and reporting.

Core AI Applications in Treasury

• Executive copilots: Dedicated tools that give leadership self-service access to real-time cash, FX and treasury data without interrupting daily operations.
• Fraud and anomaly detection: Machine learning running on consolidated TMS payment data to flag unusual vendor changes, transaction sizes or timing shifts in real time.
• On-demand analysts: Queryable AI-enabled solutions that synthesize data from multiple systems to calculate risks and hedging scenarios.
• Agentic process execution: Autonomous agents that handle scheduled monthly reporting, build slide decks, log meetings and draft calendar responses.
• Multi-source data synthesis: Continuous background scanning of emails, notes, market news and internal datasets to maintain live intelligence maps and power decision-making.

Use Case 1: Give Executives Direct Access to Treasury Information

Another treasury team is developing what its treasurer calls a “treasury copilot.” The goal is to allow the CFO to ask the tool routine questions about cash, FX exposure and other treasury-related data directly rather than going directly to the team, who would then have to stop what they’re working on to assemble an answer each time.

Getting there, however, started well before an AI tool was introduced. The organization first focused on its data strategy because, as the treasurer put it, “AI is only as good as your data is.”

Use Case 2: Look for Payment Anomalies and Potential Fraud

Treasury teams are using AI for payments and fraud detection. At one global company, the treasury team previously managed payments through separate regional bank portals — a fragmented setup that made it hard to spot patterns.

They consolidated everything through a single treasury management system, creating “one source of truth” for all payment data. With that foundation in place, they're now using AI to flag unusual payment activity — things like unexpected vendor shifts, atypical transaction sizes or timing patterns that deviate from normal behavior — to catch potential fraud before it happens.

Use Case 3: Build an On-Demand Treasury Analyst

One treasury practitioner built an AI-enabled solution that brings together data from multiple internal systems and makes it queryable through Microsoft Teams. “I don't have a team member who does foreign exchange, so I kind of created my own analyst,” he explained.

Instead of manually gathering the inputs for an exposure analysis, he can ask the system about value at risk, cash-flow risk, EBITDA risk or hedging scenarios and receive an answer that’s based on company data. And he is now applying the same approach to interest-rate exposure, credit agreements, covenants and compliance.

This use case isn’t about replacing a treasury role; it’s about compressing the time between question and analysis. Information that had to be assembled before treasury could use it is now available on demand.

Use Case 4: Turn Recurring Work into an Agent-Run Process

Treasury practitioners are moving recurring work from one-off AI prompts into repeatable, autonomous workflows. One treasury professional uses an agent for monthly business-review presentations.

“It knows that by x day of the month I need the slides. I already have the code that builds the PowerPoint presentation. It just follows the steps and up pops the PowerPoint deck. All I have to do is review and double-check that the colors are right,” the treasury professional said.

But the real scale comes from stacking agents for different tasks. Another treasury practitioner said they run more than a dozen agents simultaneously. For example, one agent summarizes and logs project meetings, and another monitors incoming emails to draft meeting responses with availability checks and timezone conversions already built in.

Use Case 5: Pull Scattered Information Together for the Next Action

More advanced applications of AI in treasury are being developed to combine information previously living in separate places. One treasury practitioner described it as pulling information from datasets and connecting it with related emails, meeting transcripts, notes and chats. Rather than simply answering questions, the AI agent assembles information to help the team prepare for the next action.

Compilation tasks could address a familiar problem: The information needed to make a decision often exists, but gathering and connecting it manually consumes a lot of time. For example, one team built a detailed market map of their industry, tracking who has capital with whom and what they're building. The system continuously scans news and deal announcements, updating the map automatically. What used to be a manual research project now runs in the background, so decision-makers always have current intelligence at hand.

This AI use case helps treasury teams make better use of the information they already have — and do so faster.

A Practical Way to Think About AI in Treasury

The examples above point to a common pattern: The most practical AI applications in treasury start with information or processes that already exist but require significant manual effort to gather, analyze or execute.

Treasury teams can look for opportunities to:

• Create a reliable data foundation so your data is AI-ready.
• Make information easier to access, as with the virtual analyst and treasury copilot examples.
• Identify patterns that are difficult to spot manually, such as unusual payment activity and potential fraud.
• Turn recurring tasks into repeatable workflows, including monthly reporting and other routine administrative work.
• Connect information from multiple sources, bringing together datasets, emails, meeting transcripts, notes and market information to prepare for the next action.

For treasury teams considering where to start, a useful question is: What recurring task takes the most time because someone has to gather or connect information that already exists?