As of 13 August 2026, AI can only partly predict which leads will convert.

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 anonymised historical leads together with the outcome field, creation date and the fields that were available before a sales decision was made.
   2. Write down what counts as conversion, the conversion window, and which current leads are still open for follow-up.
   3. Remove names, personal contact details and fields created after conversion, then fix inconsistent labels and remove duplicate rows in a spreadsheet.
   4. Split the cleaned file into historical records and current leads, keeping the field names and definitions identical in both tables.
   5. Paste the field definitions, historical data, current leads and sales context into the prompt, then ask the model to produce the priority bands and recommended next actions.
   6. Ask a colleague who understands the CRM to check the exclusions, assumptions and sample rankings against the original records before anyone changes their outreach order.
   7. After the agreed conversion window, add the real outcomes to the current-lead table and compare them with the predicted priority bands before adopting the method again.

   Prompt

   You are helping me prioritise sales leads, not make certain predictions. I will paste an anonymised table of historical lead records below.
   
   Task:
   1. Define the conversion outcome using this exact field: [conversion field and value that counts as converted].
   2. Exclude any field that would only become known after the conversion decision, including later-stage activity, closed-won information and post-sale notes.
   3. Identify missing values, inconsistent labels, duplicate records and possible data leakage before ranking anything.
   4. Use the historical records to create a transparent lead-priority score for the current leads. Explain which observed fields contribute to the score and do not claim that any field causes conversion.
   5. Return a table with lead ID, priority band, evidence used, missing information and a recommended next action. Do not invent values or fill gaps with guesses.
   6. Separate the historical records used to learn patterns from the current leads being ranked. Suggest a simple time-based test using later outcomes, and state what result would count as useful without pretending the test proves future performance.
   7. Flag groups that may be disadvantaged by the available data, and list the checks a sales manager should complete before using the ranking.
   8. State clearly that this is a prioritisation aid rather than a guarantee of conversion.
   
   Data fields and definitions:
   [Paste the field list and definitions here]
   
   Historical lead data:
   [Paste anonymised historical records here]
   
   Current leads to rank:
   [Paste anonymised current lead records here]
   
   Business context:
   - Sales process: [brief description]
   - Conversion window: [for example, within the agreed sales period]
   - Capacity available for follow-up: [describe the team capacity]
   - Lead groups or territories that must be treated separately: [list them or write none]
   
   Use plain English and show your assumptions. Ask for clarification instead of guessing.

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