As of 13 August 2026, AI can only partly forecast your small business cash flow.

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, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open your business bank and accounting records and export the recent transactions, current balances, unpaid sales invoices and unpaid supplier bills as CSV files or readable tables.
   2. Gather the payment dates and amounts for payroll, VAT, PAYE, corporation tax or Self Assessment payments, loan repayments, rent, subscriptions, stock purchases and owner drawings that are expected during the forecast period.
   3. Remove duplicate rows and label each transaction as money in or money out, then add the account, date, amount, category and whether it is confirmed or estimated.
   4. Paste the prepared data and the business context into the prompt, including the forecast start date and the payment confidence for expected receipts and bills.
   5. Ask the model to produce the 13-week base, cautious and strong cash-flow forecasts, then answer its questions about missing or inconsistent data before accepting the calculations.
   6. Compare every opening and closing balance and each major receipt and payment against the bank and accounting records, correcting the prompt and regenerating the forecast when a figure or date does not match.
   7. Send the completed forecast and its assumptions to your accountant or a financially responsible colleague when a negative balance, borrowing decision, tax payment or major supplier commitment depends on it.

   Prompt

   Create a small business cash-flow forecast from the data below. Use cash received and cash paid, not accounting profit. Do not invent transactions, dates, balances or assumptions. First identify missing, duplicated or inconsistent data and list the questions that must be answered before relying on the forecast. Reconcile the opening balance to the supplied bank balance and show the calculation for every period. Separate confirmed items from estimates. For uncertain items, create a base case, cautious case and strong case, and state the assumption behind each change. Show opening cash, receipts, payments, net cash movement and closing cash for each week in a 13-week forecast. Flag any week where closing cash is negative or depends on an unconfirmed receipt. Keep VAT, tax, payroll, loan repayments and owner drawings separate where the data allows. Do not give regulated financial advice and do not claim that the forecast is accurate. Finish with a short list of source data and checks I must complete. Business context: [describe the business, payment terms, VAT status and known upcoming changes]. Current cash balances: [paste balances by account and date]. Historical transactions: [paste CSV or table with date, description, amount, money in or money out, account and category]. Expected invoices and receipts: [paste customer, amount, expected date, confidence and whether VAT is included]. Expected bills and payments: [paste supplier, amount, due date, confidence and whether VAT is included]. Other commitments and assumptions: [paste loans, payroll, tax, rent, subscriptions, stock purchases, owner drawings and planned one-off costs]. Forecast start date: [YYYY-MM-DD].

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