As of 13 August 2026, AI can build a sales forecast from your CRM.

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 CRM pipeline view and export the relevant opportunities as CSV or XLSX, including opportunity ID or name, owner, value, currency, stage, forecast category, probability, expected close date, created date, last activity date and closed status.
   2. Remove unrelated customers or opportunities from the export, keep the column headings, and replace sensitive notes with short factual fields where the notes are not needed for forecasting.
   3. Write down the forecast period, the meaning of each stage and forecast category, the treatment of renewals or expansion deals, and whether values are gross, net or recurring revenue.
   4. Paste the export and those definitions into a chatbot using the prompt, then answer its clarification questions from your CRM rules rather than allowing it to guess.
   5. Compare every deal-level value, stage, owner and close date in the result with the current CRM, and recalculate the displayed category totals from the supplied rows.
   6. Ask your sales manager or sales operations colleague to challenge the scenario rules, stale-deal flags and exclusions against recent team knowledge and the forecast process.
   7. Send the checked base forecast and its assumptions to the people responsible for targets, staffing or cash planning, keeping the conservative and upside scenarios visibly separate.

   Prompt

   Build a sales forecast from the CRM data below. Use only the supplied data and do not invent deal values, close dates, probabilities, activity or customer information. First identify the columns and any missing, duplicated or contradictory records. Then produce: 1) a deal-level forecast with opportunity name or ID, owner, value, stage, expected close date, current CRM probability if supplied, and your forecast category; 2) totals by forecast category and expected close period; 3) a conservative, base and upside scenario, explaining the rule used for each; 4) deals that are overdue, stale, missing key fields or have weak evidence of progressing; 5) the assumptions, exclusions and data-quality limitations. Do not present invented probabilities as facts. If the data does not support a reliable calculation, say exactly what is missing and provide the most defensible partial result instead. Show the arithmetic for every total and label all dates and currencies clearly. Use the following definitions unless the data says otherwise: closed won is booked revenue, closed lost is excluded, open opportunities are included only in the relevant scenario, and expected close period is based on the supplied expected close date. Ask concise clarification questions before calculating if a required field is absent or ambiguous. CRM export: [PASTE CSV OR TABLE HERE]. Forecast period: [PASTE PERIOD HERE]. Currency: [PASTE CURRENCY HERE]. Company definitions for forecast categories: [PASTE DEFINITIONS HERE].

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