As of 13 August 2026, AI can only partly assign probabilities to your pipeline stages.

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 and export historical opportunities with stage history or entry dates, final outcome, close date, segment, sales team and opportunity type.
   2. Write down the exact meaning of every pipeline stage, including whether an opportunity can move backwards and how closed-lost, withdrawn and still-open opportunities should be treated.
   3. Remove duplicate records, test opportunities and records with missing or contradictory outcomes, and keep a separate list of every removal.
   4. Paste the cleaned export, stage definitions and business rules into the prompt, keeping customer names and other unnecessary personal data out of the file.
   5. Ask the model to calculate historical conversion rates and suggested probabilities separately for materially different segments, rather than using one average for the whole pipeline.
   6. Compare the model's counts and rates with a pivot table or CRM report made from the same export, correcting any difference before using the suggested probabilities.
   7. Apply the probabilities to the current open pipeline, label the result as an estimate, and ask a sales operations colleague to challenge the assumptions and data exclusions.
   8. Save the probability table and compare it with later closed-won and closed-lost outcomes so the stages can be recalibrated when the evidence changes.

   Prompt

   I need estimated probabilities for each stage of a sales pipeline.
   
   Use the pipeline data and stage definitions below. Do not invent missing figures, outcomes or assumptions. First check the data for duplicate opportunities, missing close outcomes, inconsistent stage names, impossible dates and opportunities that are still open. Report those issues before calculating anything.
   
   Calculate, for each stage:
   1. The number of historical opportunities that entered the stage.
   2. The number that eventually became closed-won.
   3. The observed conversion rate, clearly labelled as historical rather than predictive.
   4. A suggested probability for forecasting, only where the sample is large and consistent enough to support one.
   5. The limitations and data gaps affecting that probability.
   
   Separate new-business and expansion opportunities, and separate customer segments or sales teams where the data supports it. Do not combine groups with materially different sales motions without saying so. Explain which opportunities were included and excluded. If the data is insufficient, say that no reliable probability can be assigned rather than filling the gap.
   
   Then produce a concise table with stage, included opportunity count, closed-won count, historical conversion rate, suggested forecast probability, confidence level and reason. Include a short method for testing the probabilities against later outcomes. Do not present the suggested probabilities as guarantees or as a substitute for deal-specific judgement.
   
   Stage definitions:
   [PASTE STAGE DEFINITIONS]
   
   Historical pipeline export:
   [PASTE CRM EXPORT OR TABLE]
   
   Business rules and segments:
   [PASTE SALES MOTION, SEGMENTS, EXCLUSIONS AND DATE WINDOW]

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