Review a supplied cash forecast against actual bank movements using explicit timing assumptions. Reconcile opening cash, receipts, payments and closing cash before asking AI to explain the variances.
AI cash flow forecasting is reliable only when receipts, payments and timing assumptions can be checked against actual cash movements. Keep the forecast arithmetic in an explicit financial workbook and use the external AI model to explain verified differences. Start the review with opening cash and finish with a closing balance that reconciles.
Save the forecast version that existed before the period began. Keep its expected receipt dates, payment dates, account scope and currency. Later updates belong in separate versions; otherwise the review replaces a missed estimate with information that arrived afterward.
This is a review method for a supplied forecast. The existing cash flow forecast guide explains the build, and the 13-week cash flow template provides the longer layout. Use those pages for the schedule itself rather than creating a second, incompatible workbook just for an AI explanation.
Define what cash means for the decision. A treasury review might include selected operating bank accounts while presenting restricted balances separately. Apply the same definition to forecast and actuals, and identify any overdraft treatment explicitly.
Cash is also different from profit. The SEC's financial statement guide distinguishes income statements from cash flow statements and separates operating, investing and financing cash flows. A profitable sale is not evidence that the customer has paid.
Bring the saved forecast, bank movements for the completed periods and the supporting receivables and payables schedules. Useful columns include transaction reference, account, currency, expected or actual settlement date, category, amount and forecast version. Keep customer or supplier references consistent enough to investigate a difference.
Use positive receipts and positive payments in separate columns. The calculation subtracts payments. If the bank export instead uses signed amounts, map the signs once and check a known receipt and payment before aggregating.
Exclude internal transfers only when both sides belong to the cash scope being consolidated. A transfer to an excluded account is a real outflow from the selected scope. A credit facility limit is not a bank receipt; only an actual draw belongs in actual cash movements.
Aggregate by the same settlement periods as the forecast. Do not compare a forecast based on expected bank availability with actuals grouped by invoice date. Keep unmatched or uncategorized movements in a visible review category so the total still ties to the bank.
A US distributor starts with $100,000 in the selected USD bank accounts. Its saved forecast and actual bank movements produce the following comparison. All amounts are USD, there are no currency translation effects, and internal transfers within scope have been removed from both sides.
| Week | Forecast opening cash (USD) | Forecast receipts (USD) | Forecast payments (USD) | Forecast closing cash (USD) | Actual opening cash (USD) | Actual receipts (USD) | Actual payments (USD) | Actual closing cash (USD) | Closing variance (USD) |
|---|---|---|---|---|---|---|---|---|---|
| Week 1 | 100,000 | 120,000 | 100,000 | 120,000 | 100,000 | 90,000 | 110,000 | 80,000 | -40,000 |
| Week 2 | 120,000 | 80,000 | 90,000 | 110,000 | 80,000 | 100,000 | 85,000 | 95,000 | -15,000 |
Closing cash = Opening cash + Receipts − Payments
Closing cash variance = Actual closing cash − Forecast closing cash
With this table in columns A through J and Week 1 in row 2, enter these formulas in E2, I2 and J2, respectively:
=B2+C2-D2
=F2+G2-H2
=I2-E2
For Week 2, enter the following formulas in B3, F3, E3, I3 and J3, respectively:
=E2
=I2
=B3+C3-D3
=F3+G3-H3
=I3-E3
The links between weeks matter. Resetting Week 2 actual opening cash to the forecast opening would erase the previous week's miss. Retain the separately rolled forecast and actual balances.
The first week's receipt variance is 90,000-120,000, or negative $30,000. Payments exceed forecast by $10,000, which has a negative cash effect. Together they explain the $40,000 closing shortfall.
| Cash effect | Week 1 (USD) | Week 2 (USD) | Two-week total (USD) |
|---|---|---|---|
| Receipts: actual minus forecast | -30,000 | 20,000 | -10,000 |
| Payments: forecast minus actual | -10,000 | 5,000 | -5,000 |
| Net cash movement variance | -40,000 | 25,000 | -15,000 |
Week 2 recovers $25,000 of the earlier shortfall, leaving closing cash $15,000 below forecast. Across the full window, forecast receipts total $200,000 and payments $190,000. Actual receipts total $190,000 and payments $195,000.
The cumulative proof is therefore 100,000+190,000-195,000=95,000 actual closing cash, against 100,000+200,000-190,000=110,000 forecast closing cash. The negative $15,000 difference equals the summed movement variances.
Check the Week 1 and Week 2 bridges with these formulas, respectively:
=(F2-B2)+(G2-C2)-(H2-D2)-J2
=(F3-B3)+(G3-C3)-(H3-D3)-J3
Both checks must equal zero. Also tie actual opening and closing cash to the bank balance records. The arithmetic identity cannot prove that the export contains every bank movement.
A receipt that appears in Week 2 may settle an invoice forecast for Week 1. Establish that using the invoice or remittance reference before calling the variance temporary. The table alone shows that receipts recovered partially; it does not prove which customer paid late or whether the remaining gap will be collected.
Ask for a checked explanation: which expected receipts did not arrive, which payments moved, and what evidence supports the revised dates? Treasury or finance owns the update. Sales and accounts payable can supply timing evidence, but an external AI model's suggestion is not a customer commitment.
NIST's AI Risk Management Framework includes trustworthiness considerations in AI use and evaluation. For this review, retain a clear distinction between a calculated variance, a documented cause and an unconfirmed explanation. See AI analytics for that broader calculation-and-explanation approach.
If the comparison spans currencies, use the separate multi-currency cash flow forecasting guide to distinguish transaction movements from exchange-rate effects.
Bring a sample of your own saved forecast and actual cash-movement exports; Covirage maps the columns, and its tools compute and check the comparison figures. The external AI model explains the results with every figure traced to its rows. Create a free Covirage account, set up your business and bring your files.
See forecast analysis for the supporting approach to reviewing estimates against actual results.
Bring the forecast version saved before the review period, actual bank movements, opening and closing bank balances, and supporting receipt and payment schedules. Keep account scope, currency and date rules consistent so that the forecast and actual totals are comparable.
Revenue recognition and customer payment are different events. Customers can pay later than expected, inventory can be purchased earlier, and capital spending or debt repayments can use cash without reducing current-period revenue. Reconcile bank movements rather than treating revenue as receipts.
After tools calculate and reconcile the differences, the external AI model can explain which receipt or payment categories changed. An assertion that a payment was delayed still needs a bank reference, revised settlement date or other supporting evidence from the business.