10 AI prompts for treasurers: exposure, term sheets, committee — and the check to run every time
Ten copy-ready prompts for the treasury tasks where AI genuinely saves time, each with what to provide and the check you must not skip. The first two are here in full; the other eight are for subscribers.
A treasury prompt is not judged by how elegant its answer is, but by the time it saves once the answer is checked. The ten below were chosen on that basis: each targets a recurring task, states what to give the tool, and ends with the check without which the answer must not be used.
Three rules before you start
- No market number comes from AI. Rates, forward points, volatility and premiums come from a time-stamped market source; none of the prompts below asks the tool to supply them.
- Sensitive data is anonymised. Replace customer, supplier and bank names with codes, and never enter bank details, except in an enterprise tool that excludes data reuse.
- The answer is a draft. It is reviewed, corrected and approved by the person accountable for it, like the work of a brilliant intern who does not know the company.
The structure of a good treasury prompt
| Element | What it contains | Example |
|---|---|---|
| Role | The expected point of view | You are the deputy treasurer of a mid-sized industrial exporter |
| Context | What the tool cannot guess | Functional currency euro, customers in USD and GBP, 12-month hedging policy |
| Data | The file or extract, anonymised | Export of open invoices: date, currency, amount, due date |
| Task | A verb and a deliverable | Compute net exposure by currency and by month |
| Format | The exact shape of the answer | A table, then three one-sentence points of attention |
| Guardrail | What the tool must not do | Do not invent any data; flag any ambiguous line instead of correcting it |
Prompt 1 — Measuring FX exposure from invoices
What to provide: an export of open customer and supplier invoices, with currency, amount, due date and direction (receipt or payment).
You are the deputy treasurer of a company whose reporting currency is the euro. Here is the export of our open invoices. 1) For each currency other than the euro, compute total receipts, total payments and the net position, month by month over the next twelve months. 2) Present the result as a table, one row per currency and month. 3) List separately the lines you could not classify (missing currency, inconsistent date, ambiguous direction), without correcting them. Do not use any exchange rate and do not convert anything into euros.
Check that the table's totals by currency match those of your export — a simple sum in the spreadsheet will do — and deal with the flagged lines one by one. Conversion into euros is done afterwards, in the spreadsheet, with a dated market rate.
Prompt 2 — Reviewing an FX policy and finding its gaps
What to provide: the text of the FX or treasury policy, anonymised.
You are auditing a small company's FX risk management policy. Here is our policy. Check whether it clearly answers the following questions: what scope is covered, what hedge ratios by horizon, which instruments are allowed and which are forbidden, who decides and up to what amount, how and how often hedges are monitored, what happens when a limit is breached. For each question, quote the sentence that answers it or write "not addressed". End with the three most important gaps, ranked by risk. Do not rewrite the policy.
Check every quotation against the original document: a model can attribute to your text a sentence it does not contain. The gaps flagged are leads, not conclusions; some are dealt with in other company documents.
The other eight prompts
The full version of this article contains the following eight prompts, each with what to provide and its check, plus a summary checklist to keep next to the tool:
- Prompt 3 — Extracting a term sheet's features into a table
- Prompt 4 — Preparing the monthly treasury committee memo
- Prompt 5 — Classifying the lines of a multi-currency bank statement
- Prompt 6 — Building the skeleton of a 13-week forecast by currency
- Prompt 7 — Comparing two hedging proposals clause by clause
- Prompt 8 — Summarising a central bank decision and its effects on the company
- Prompt 9 — Preparing the renegotiation of your FX terms
- Prompt 10 — Building a margin stress scenario without letting AI calculate
The rest of this article is for subscribers — free, in exchange for an email address.
Get all 10 prompts in full
The remaining eight prompts — term sheet, committee, bank statement, forecast, renegotiation… — each with what to provide and its check, plus the checklist to keep next to the tool.
- One field only
- One-click unsubscribe
- No data resold
Market numbers, without AI
Prompts prepare the documents; the price is checked with market data. The pricer recomputes a hedge with the current spot rate, forward points and volatility.
Open the pricerFrequently asked questions
The questions finance departments ask most often on this topic
What is the best prompt for analysing FX risk?
One that gives the tool the context it cannot guess — reporting currency, horizon, hedging policy — and forbids it from inventing data or using exchange rates. Conversion into euros and valuation are then done with dated market data.
Can ChatGPT be used for corporate treasury?
Yes for reading, classifying, summarising and drafting, provided data is anonymised or an enterprise offering excluding reuse is used. No for obtaining rates, premiums or a hedging decision: those require time-stamped market data and human approval.
How can I avoid AI errors on financial figures?
By explicitly forbidding it to invent or correct data, asking it to flag ambiguous lines, and checking the totals of its answer against the source. Market numbers must never come from the tool itself.
Do I need a specialised AI tool for treasury?
Not to start with: a general-purpose assistant in its enterprise version covers reading, classifying and drafting. A specialised tool becomes useful once you need reliable market data — to value hedges or check a price — which general-purpose assistants do not have.
Read next
AI and currency risk: what it does well, what it does badly, and where it becomes dangerous
9 min read
Reading an FX term sheet: the eight lines that set the price
10 min read
Treasury reporting to the executive committee: the seven indicators that matter
9 min read
Multi-currency cash forecasting: building a reliable 13-week forecast
15 min read