As of 13 August 2026, AI can only partly identify which existing customers are upsell leads.

What to actually do

1. Use a tool built for this

   The route this page recommends

2. Hand it to a person

   Second choice

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

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 30 minutes until you can act on the result.

   How to actually do it

   1. Open your CRM, billing or subscription system, product usage reports, support records and recent call notes, then export the fields needed for customer identity, purchases, usage, engagement, complaints, opt-outs and account dates.
   2. Remove unnecessary personal data and replace customer names with account IDs, keeping the fields needed to reconnect a result to the correct account internally.
   3. Write down your current products or services, eligibility rules, known exclusions, pricing bands and any customers who must not be contacted for sales reasons.
   4. Paste the anonymised customer data and the product rules into the prompt, then ask the chatbot to produce the ranked list, evidence and exclusions.
   5. Give the ranked output to a sales operations colleague and have them compare every high-priority recommendation with the current CRM record, support status, opt-out record and account owner’s knowledge.
   6. Remove recommendations with stale, conflicting or weak evidence, then ask the account owner to confirm the timing and suitable next action before any outreach is sent.
   7. Record the accepted leads and reasons in the CRM, and send a human-written or human-approved message that reflects the customer’s actual situation.

   Prompt

   Act as a sales operations analyst. Identify which existing customers are the strongest upsell leads from the customer data below.
   
   Use only the information supplied. Do not invent customer needs, budgets, decision-makers, contract dates or buying intent. Do not use demographic or other sensitive personal information. Rank the customers by evidence of a relevant upsell opportunity, not by account size alone.
   
   For each customer, provide:
   1. A priority of high, medium or low.
   2. The specific product or service they may be suitable for, only where the data supports it.
   3. The evidence, quoting the relevant fields and dates supplied.
   4. Any missing or contradictory information.
   5. The risk of contacting them now.
   6. One sensible next action that gathers information rather than assuming interest.
   
   Exclude customers with recent complaints, unresolved support issues, explicit opt-outs or no evidence of a relevant need. Put excluded customers in a separate list with the reason. End with the three data fields that would most improve this assessment.
   
   Customer data:
   [PASTE ANONYMISED CRM, PURCHASE, USAGE, SUPPORT, ENGAGEMENT AND CALL-NOTE DATA HERE]
   
   Available products or services and eligibility rules:
   [PASTE YOUR CURRENT PRODUCT LIST, PRICING BANDS AND ELIGIBILITY RULES HERE]

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