As of 13 August 2026, AI can only partly forecast cash flow from your invoices.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your invoicing system and export all outstanding and recently paid invoices for the forecast period, including invoice amount, VAT where available, due date, status and actual payment date.
   2. Open your business bank records and write down the opening balance for the forecast period, then list regular, one-off, VAT, payroll, tax and supplier payments with their expected dates.
   3. Put the invoice and outgoing-payment data into a spreadsheet with clear column headings, and remove duplicate invoices and cancelled items.
   4. Paste the spreadsheet data and the business context into the prompt, then ask the chatbot to calculate the base, cautious and optimistic forecasts without filling gaps with invented figures.
   5. Compare the forecast receipts with the invoice totals, check each opening and closing balance calculation, and trace every outgoing payment back to your records.
   6. Check the payment-timing assumptions against your actual customer payment history and change any scenario assumption that does not fit your customers.
   7. Paste the corrected data and assumptions into Julius AI or your spreadsheet, regenerate the forecast, and save the input data, assumptions and final forecast together.
   8. Use the cautious forecast to decide whether any payment or spending decision needs discussion with your accountant or finance adviser.

   Prompt

   Build a cash-flow forecast from the data below. Use only the figures I provide and label every assumption.
   
   Business context:
   - Forecast period: [START DATE] to [END DATE]
   - Currency: GBP
   - Opening bank balance: [AMOUNT]
   - Expected regular outgoing payments and dates: [PASTE DATA]
   - One-off outgoing payments and dates: [PASTE DATA]
   - Tax, VAT, payroll or other known payments and dates: [PASTE DATA]
   - Historical customer payment data, if available: [PASTE DATA]
   
   Invoice data, with one row per invoice:
   [PASTE A TABLE OR CSV WITH CUSTOMER NAME OR ID, INVOICE DATE, DUE DATE, AMOUNT INCLUDING VAT, AMOUNT EXCLUDING VAT IF AVAILABLE, PAYMENT STATUS, ACTUAL PAYMENT DATE IF PAID, AND ANY CREDIT NOTE]
   
   Produce:
   1. A weekly cash-flow table showing opening balance, expected customer receipts, each category of outgoing payment, net movement and closing balance.
   2. A separate list of invoices expected to be unpaid at each period end.
   3. A base case, a cautious case and an optimistic case. Derive payment timing from the supplied payment history where possible. If there is no history, do not invent a probability or delay. Instead, state the missing input and make the scenarios use clearly labelled timing assumptions for me to approve.
   4. A list of every missing or ambiguous input that could materially change the result.
   5. Reconcile the forecast receipts to the invoice list and show the arithmetic for each period.
   6. Flag negative closing balances and explain which assumption or invoice causes each one.
   
   Do not present the result as a guarantee. Do not invent invoices, payment dates, expenses, VAT treatment or customer behaviour. Ask questions before calculating if the data is too incomplete to produce a meaningful forecast.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.