As of 13 August 2026, AI can only partly forecast demand for your products.

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 sales or stock system and export dated product-level records covering units sold, selling price, promotions, returns, stock availability and stockouts.
   2. Create one clean spreadsheet with one row per product and period, and mark periods where demand was limited because the product was unavailable.
   3. Add known future events such as promotions, price changes, launches, closures, bank holidays and supplier constraints, then remove duplicate rows and clarify any product name changes.
   4. Paste the business context and the cleaned table into the prompt, and ask the chatbot to identify missing or unreliable fields before it produces a forecast.
   5. Paste the resulting forecast into a spreadsheet and compare its backtested predictions with the actual sales in the held-back periods, checking that it has not treated stockouts as low demand.
   6. Discuss the central, lower and upper scenarios with the person responsible for purchasing, then record which scenario you will use for stock and supplier decisions and when you will update it with actual sales.

   Prompt

   Forecast demand for the following product using the data provided below. Treat the forecast as a planning aid, not a certainty.
   
   Business context:
   - Product: [product name]
   - Market and sales channel: [market and channel]
   - Forecast horizon: [period to forecast]
   - Forecast frequency: [weekly or monthly]
   - Known future events: [promotions, launches, closures, price changes, holidays or supplier constraints]
   
   Historical data:
   [paste a dated table with units sold, selling price, promotions, stock availability and any known stockout periods]
   
   Instructions:
   1. Check the data for missing dates, duplicate rows, changing product definitions and periods when sales were limited by stock availability.
   2. Do not invent missing figures. List the assumptions and data problems before forecasting.
   3. Produce a forecast for each period, with a central estimate and a clearly labelled lower and upper planning scenario. Explain how each scenario was chosen without presenting it as a probability unless the data supports that.
   4. Separate observed patterns from assumptions about future demand. Discuss seasonality, promotions, price changes, trend, one-off events and stockouts where the data supports them.
   5. Hold back the most recent complete periods as a test set if the data allows. Compare the forecast with actual sales and report the error in plain English. If meaningful backtesting is not possible, say so.
   6. Show the result in a table that I can paste into a spreadsheet.
   7. Recommend what additional data would most improve the forecast.
   8. End with a short list of decisions that still require my judgement, including stock, promotions and supplier commitments.
   
   Use only the information I provide. Ask concise questions before forecasting if a missing input would materially change the result.

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