Treasury teams are under constant pressure to answer new questions from CFOs, fast. The challenge is that traditional systems generally aren’t built for the pace or the specificity of those requests. When priorities change, dashboards become stale, and building new reports can take days.
A recent anecdote from a Trovata Sales Engineer, Chris Brown, illustrates how this model is beginning to shift. With the help of Trovata AI and Intercom’s Fin AI assistant—embedded directly within the Trovata user experience—he was able to generate a key metric that a prospective client needed, days cash on hand, in just minutes. No SQL. No custom reporting. No development queue.
This example captures what’s becoming possible as treasury teams get AI-ready, using clean, real-time bank data and natural language interfaces to generate metrics on demand. It’s a practical step toward the agentic AI future that is being widely discussed in industry media and by the wider business community.
Here’s how it happened, and how any treasury department can adapt the principles for their own important metrics.
Watch this video for a quick explanation, or keep reading below!
The Metric in Question
The treasury analyst and prospective Trovata user needed a fast, accurate view of day’s cash on hand, which is a fundamental liquidity metric for any company burning cash or scaling aggressively. They wanted to communicate it to the CFO daily, which is not always easy for fast scaling businesses or those with complex banking and transaction data.
In many systems, answering that question requires building a custom report or dashboard that pulls balances from multiple accounts, normalizes the data, applies burn-rate assumptions, and calculates the runway. That means meetings, configuration work, validation, and, often, consultants. Trovata’s AI tooling offered a simpler way.
Step 1: Build the Prompt, Not the Dashboard
The first thing Chris did to help come up with this requested metric was draft a simple natural-language instruction on their behalf:
“Pull my balances, take the total cash available today, and divide it by my daily burn rate. Return the result as days cash on hand.”
Because Trovata is already aggregating real-time balances and transactions via direct APIs, Trovata AI can retrieve the numbers instantly. Then it performs the calculation and surfaces the result in plain English.
The key to this is that he didn’t need to pre-build a data model. The data already existed, structured and normalized in Trovata’s bank data lake. The AI layer simply bridged the gap between raw information and the specific metric the prospect needed.
This is exactly what we mean when we reference ‘clean, accessible, real-time data’ as a prerequisite for AI success in treasury. If that underlying data was a result of manual, error prone workflows or made up of information that was days old, the AI’s response has no chance of giving an accurate answer.
It’s like asking what you should wear outside today, based on the weather from three days ago. The data must be correct and up to date in order for AI to be usable and accurate.
Step 2: Use AI to Optimize the Prompt
Before sending the output back to the prospect, Chris wanted a more concise, reusable version of the prompt. So he opened Fin, Intercom’s AI assistant, which is synced to Trovata’s product documentation within the Help Center.
Fin understood both the business context (how days cash on hand should be calculated) and the technical context (how Trovata AI expects prompts to be structured). It helped refine the instruction into an even more clean, repeatable template:
This is where AI becomes a force multiplier. You don’t just use AI to get the result. You use AI to help create the prompt that gets the result.
Step 3: Validate Before You Trust
For treasury teams, trust is earned through validation. To verify that the output was accurate, Chris did a little bit of fact-checking by:
- Checking total cash on the Balances page
- Drilling into account-level metadata
- Verifying the math manually
After verifying, he found that the AI answer was indeed accurate, and the prompt was written effectively.
Tip: As another verification path, Trovata AI can show the database query used to generate its output, giving technically-minded users full transparency. Explainability is one of the key questions that treasury teams should ask any AI vendor.
Step 4: Repeat as Needed
Even though the AI prompt can’t yet be saved natively, Chris used a clever workaround, placing the prompt text directly in a dashboard report title. That allowed him to:
- Store the prompt in context
- Validate the underlying balances visually
- Copy and paste the prompt into Trovata AI whenever needed
What would have required a custom metric or a scoped dashboard in a legacy tool is now a lightweight workflow anyone can reproduce.
Why This Matters for Treasury Teams
This example isn’t just about calculating days cash on hand. It’s about a new way of working, where analysts generate bespoke metrics themselves, in minutes, without waiting for a dashboard or relying on IT.
This story highlights four shifts:
1. Prompts are Becoming More Flexible Than Dashboards
Dashboards are powerful, but they’re rigid. AI prompts adapt instantly to new priorities. If a CFO wants to monitor a new metric tomorrow, the analyst can draft a prompt on the spot.
2. AI Bridges the Last Mile Between Systems and Insight
Treasury data is rich. With the right data lake in place, it can also be easily categorized, tagged and searched. The bottleneck has always been turning that data into answers, but AI can now close that gap.
3. Validation Remains a Human Responsibility
AI acceleration doesn’t eliminate risk controls. It strengthens them by giving analysts more tools to confirm accuracy. This isn’t (and never will be) about giving AI the keys and letting it run your business, but it is about giving your real talent the time and bandwidth to use their expertise on high value work, no digging through statements and manually calculating spreadsheets.
4. This is a Preview of Agentic AI
Today, the user triggers the calculation. In tomorrow’s system, the prompt runs automatically, checks ranges, and alerts the user if the result deviates. This is exactly the direction outlined in the AI readiness framework we shared with Strategic Treasurer.
The Bigger Picture: AI-Ready Treasury Teams Move Faster
Most CFOs operate in environments where priorities can shift quickly and with little warning. Funding changes runway. M&A creates new reporting needs. Interest-rate moves drive daily questions about exposure.
The traditional model of building static reports for each new question can’t keep up.
AI-enabled treasury teams operate differently:
- They start with live, API-fed data
- They generate metrics in natural language
- They validate quickly
- They adapt prompts as the business changes
It’s a lightweight, analyst-led model that reduces reliance on BI teams, reduces friction, and expands the number of questions treasury can answer in real time.
What Comes Next
The days cash on hand example is one of many metrics that can be produced on demand when AI sits on top of clean, real-time data. Any calculation that blends multiple inputs is a strong candidate for prompt-based analysis.
What’s changing is who can create those insights. You no longer need a dashboard build, a modeling exercise, or a backlog request to answer a CFO’s follow-up question. Analysts can generate precise, situational metrics themselves, validate them quickly, and adapt the prompt as conditions shift. If you want to learn more about how this could work with your own data, and what types of metrics your team could generate instantly, book a Trovata demo.
