As of 13 August 2026, AI can only partly forecast your business sales.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Do it yourself

   Second choice

   A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open your sales, order or accounting records and export a dated table covering as many comparable past periods as you have, including sales value, refunds, cancellations, product or service and channel where available.
   2. Add a separate list of known future changes, such as price changes, promotions, contracts, lost customers, capacity limits and seasonal events, and mark each item with the date or period when it applies.
   3. Remove duplicate transactions and check that the sales totals in the export agree with your accounting records for each period before sharing the data with a chatbot or Julius AI.
   4. Paste the data, business context and known changes into the prompt, replacing every bracketed slot with your information and stating the forecast horizon and weekly or monthly frequency.
   5. Ask the model to produce the baseline and alternative scenarios, then copy the forecast table into a spreadsheet and recalculate its totals and formulas from the stated inputs.
   6. Compare the model's historical comparison with your actual sales records, challenge every assumption about future demand, and change or remove assumptions you cannot support.
   7. Send the checked forecast and its assumptions to your accountant or financial adviser if it will determine borrowing, major spending, tax planning or other material business decisions.

   Prompt

   Create a sales forecast for my business using the data below. Do not invent missing figures, customers, trends or events. First check the data for missing periods, duplicate rows, inconsistent dates, refunds, cancellations and changes in product or service categories, and list every issue you find. Then produce a forecast for [FORECAST PERIOD] at [WEEKLY OR MONTHLY] level, showing the calculation method, assumptions, baseline forecast and at least two alternative scenarios. Separate historical actuals from estimates. Explain which patterns in the data support each assumption, identify where the forecast is least reliable, and show the results in a table that can be copied into a spreadsheet. Include a comparison of the forecast method against recent historical periods where the necessary data exists. Do not present a forecast as certain and do not use external market data unless I provide it and you label it clearly.
   
   Business context:
   [BUSINESS TYPE, PRODUCTS OR SERVICES, SALES CHANNELS, LOCATION IF RELEVANT]
   
   Known changes during the forecast period:
   [PRICE CHANGES, PROMOTIONS, CONTRACTS, LOST CUSTOMERS, CAPACITY LIMITS, SEASONAL EVENTS OR OTHER KNOWN FACTS]
   
   Sales data:
   [PASTE A TABLE OR DESCRIBE THE ATTACHED FILE, INCLUDING DATE, PRODUCT OR SERVICE, QUANTITY, NET SALES, REFUNDS AND CHANNEL WHERE AVAILABLE]
   
   Return the result in this order: data-quality issues, method, assumptions, historical comparison, forecast table, scenarios, limitations, and a short list of decisions or extra data that would most improve the forecast. This is not professional advice, so flag decisions that need review by my accountant or a qualified financial adviser.

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

3. Hand it to a person

   The distant third

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