Review contacts and find neglected opportunities

Why it matters

The highest-value contacts in a CRM are rarely the newest. They are the buyer who texted "can we see two this weekend?" three weeks ago, the seller whose "after the school year" date passed in June, the renter whose lease just ended. A sort by Last Activity cannot see those. A model reading the notes can, as long as you make it show its work.

Step by step

1. 01

   Start a fresh chat and upload this week's export

   One chat per review keeps old files from mixing in. Attach the CSV with + and send the upload check from chapter 02.05 first. Fix any missing columns before ranking.

   Check: Grok's row count matches your file.

2. 02

   Run the ranking prompt

   Copy the Neglected opportunities ranking prompt below, set the report date to today, and send it. Expect a table, then excluded counts, duplicates and data problems.

   Check: Every row has a reason that quotes a field or note, and no row suggests a channel the opt-out fields forbid.

3. 03

   Compare with the rules demo

   Open the neglected-lead finder in the demonstration section. It applies fixed, visible rules to the same columns. It is not AI. Use it as a baseline:

   • Contacts in both lists are your safest first calls.
   • Contacts only in Grok's list should have a note-based reason (a date passed, a life event). Check the quote is real.
   • Contacts only in the rules list may have notes that explain the silence ("not until 2027"). Grok should have said so.

   Check: You can name one contact Grok ranked higher than the rules did, and say which note caused it.

4. 04

   Spot-check the top five in your CRM

   Open each of the top five records in your CRM. Confirm the stage, the last message and the opt-out status match what Grok reported. If any fact is wrong, send the Explain one ranking prompt for that contact and correct the source data.

   Check: Five of five facts match the CRM, or you have fixed the mismatches and re-run.

5. 05

   Act on the findings yourself

   Merge the duplicate pairs in your CRM, fix stages that contradict notes, then move to chapter 02.08 to turn the ranked list into tasks. Delete the export from grok.com/files when you are done.

   Check: Duplicates merged, and the export no longer appears in grok.com/files.

Copyable templates

Use the CRM export I uploaded in this chat. Report date: YYYY-MM-DD.

Step 1. Exclude, and count separately: Stage = Trash; Stage = Closed in the last 60 days;
anyone with SMS Opt-Out = Y AND Email Opt-Out = Y, or whose notes say stop, remove me or
do not contact; anyone with Last Activity in the last 14 days.

Step 2. Find likely duplicates (same email ignoring case, or same last 10 phone digits).
Keep the record with the newest Last Activity. List the pairs. Rank only the kept record.

Step 3. Rank the remaining contacts by how much they deserve attention now. Weigh, in order:
  a) They wrote last (Last Inbound after Last Outbound) and nobody answered
  b) No outbound contact ever (Last Outbound blank)
  c) A date or life event in Last Note that has arrived or passed as of the report date
     (lease ending, school year, move, baby, retirement, "list in the fall")
  d) Stage Hot Prospect or Active Client with 30+ days of silence
  e) Timeframe 0-3 or 3-6 months with no Next Task Due
Do not use names, neighborhoods, or anything that suggests race, color, religion, sex,
disability, familial status or national origin as a ranking factor.

Step 4. Return a table of the top 15:
Rank | Contact ID | Name | Stage | Days quiet | Reason (cite fields and quote the note) |
Permitted channels (Call / Text / Email, using opt-out and Consent Source) | Suggested next step

Step 5. Below the table: "Excluded" (counts by reason), "Duplicates" (pairs and record kept),
"Data problems" (missing phone and email, stage contradicting notes), and
"Needs your approval:" (every action I must take myself).

Explain why Contact ID ____ is ranked where it is. Quote every field and note you used.
Then tell me what would move it up or down. If anything you said was an inference rather than
a field value, label it (inferred).

Try it

How to test it

Run the ranking prompt on the sample export with report date 2026-10-20. A correct answer includes Avery Collins (C001, unanswered weekend showing request) and Elena Petrova (C004, never contacted) near the top; lists Frank Delgado (C005) and Wesley Turner (C022) only as excluded; pairs C006 with C030, C031 with C032, and C037 with C038 as duplicates; and should notice notes the rules cannot read: Carlos Mendez's June 2026 selling plan and Henry Brooks's lease ending August 2026. Chapter 02.11 has the full answer key.

Troubleshooting

Advanced technical details

Optional. You can follow every step above without these.