AK
← All projects

AI Automation

AI Lead Triage & Follow-Up Assistant (Claude Skill)

A custom Claude Skill that reads a real-estate lead sheet through the Google Drive connector, scores each new lead against available inventory, and drafts follow-up messages for the hot ones.

Claude SkillsGoogle Drive connectorGoogle Sheets
Diagram — Claude Skill reads the Leads and Inventory sheet tabs, scores leads, and drafts follow-ups

How it flows

  1. 1Say "triage today's leads"
  2. 2Skill reads the Leads and Inventory tabs
  3. 3Each New lead is scored High / Medium / Low
  4. 4Sorted table with suggested actions
  5. 5Two-line follow-up drafted for High-priority leads

The problem

A real-estate agent adds leads to a Google Sheet all day. Which ones should be called first? Doing this by eye means good leads wait and weak leads waste time.

What I built

A custom Claude Skill named lead-triage, enabled in Claude’s settings and paired with the Google Drive connector.

What the skill does

  1. Finds the sheet Real Estate Lead Tracker and reads both tabs — Leads (name, phone, property, budget, source, date, status) and Inventory (property, min budget, max budget, availability).
  2. Uses “today” in Asia/Dhaka time and processes only rows where Status = New and Date Added = today.
  3. Scores each lead:
    • High — budget fits an available property and source is a referral or hot-listing inquiry.
    • Medium — budget fits, but source is an advertisement or cold lead.
    • Low — budget does not fit, or required information is missing.
  4. Sorts High → Medium → Low and outputs a table: Name, Priority, Phone, Suggested Action.
  5. Writes exactly two lines of follow-up for each High lead, in a natural call/SMS tone.

Safety rules built in — do not invent or repair missing data, do not edit the sheet, do not send anything.

Test run

Of eight rows, six were eligible. Two were the tricky ones:

  • A Hot Listing Inquiry lead with a 6,000,000 budget for a property whose minimum was 15,000,000 — correctly scored Low, because budget mismatch outranks the source.
  • A Referral lead with a blank phone cell — correctly scored Low for incomplete information, and the blank was reported, not filled in.

What I would improve

Add buying timeline, financing readiness, and location flexibility to the scoring, plus a confidence score for ambiguous sources — and connect the skill to a live inventory table so matching stays accurate.

More AI Automation projects