Prompt Chain: AI-Powered Lead Scoring and Priority Follow-Up System

What This Builds

A system that evaluates incoming leads and prioritizes which ones to call first — based on urgency signals, buying readiness, and profile fit — so you spend your highest-value calling time on the most-ready borrowers. LOs who implement lead scoring report 2–3x improvement in conversion rates because they stop treating all leads equally.

Prerequisites

• Claude Pro or ChatGPT Plus ($20/month)
• Google Sheets (free)
• Basic familiarity with your lead sources (website, Zillow, Realtor.com, referral, etc.)
• Optional: Zapier for automation ($20/month Starter)

The Concept

Not all leads are equal. A borrower who says "I need to be in a home by April 30 and my Realtor just put an offer in" is worth a call in the next 10 minutes. A borrower who fills out a form and says "just thinking about buying sometime in the next year" can wait. Without a system, most LOs call back in the order leads arrive — missing hot prospects because they're tied up with cold ones. This system adds a quick AI scoring step that sorts your lead queue automatically.

Build It Step by Step

Part 1: Define Your Lead Scoring Criteria

Before building anything, define what a "hot" lead looks like for your business. Use AI to help:

I'm a mortgage loan officer. Help me create a lead scoring rubric (1–10) based on these signals:

- Timeline: How soon do they need to close? (Under 30 days = highest priority)
- Pre-approval status: Already pre-approved, starting process, or just curious?
- Down payment situation: Have funds ready vs. still saving?
- Referral source: Realtor referral vs. online form vs. cold website lead?
- Credit self-assessment: Good (700+), average (640-699), or they're not sure?
- Employment: Stable W2, self-employed, or recently changed jobs?

Create a scoring guide: score each factor 1-5, weight them, and output a total score out of 10 with priority tiers (Hot: 8-10, Warm: 5-7, Cold: 1-4).

Use the scoring guide Claude produces as your foundation.

Part 2: Create Your Lead Intake Sheet

Set up a Google Sheet with these columns:

When a new lead comes in (from website, referral, phone call), add a row and fill in the signals you know.

Part 3: Build the AI Scoring Prompt Chain

This is a 2-step prompt chain:

Step 1 — Score the Lead:

Open your Claude Project or ChatGPT (use your LO Assistant if you set one up):

Score this mortgage lead on a 1-10 urgency/readiness scale based on our scoring rubric.

Lead details:
- Name: [Name]
- Source: [Realtor referral / website form / Zillow / etc.]
- Timeline: [what they said]
- Pre-approval status: [starting / already approved / just curious]
- Down payment: [amount or "saving still"]
- Credit self-assessment: [their words]
- Employment: [type]

Scoring rubric: [paste your rubric from Part 1]

Output:
1. Score (1-10)
2. Priority tier (Hot/Warm/Cold)
3. Top reason for score
4. Recommended first action (call within X hours/days, send email, add to drip)

Step 2 — Generate the Outreach Message:

Feed the score output back into a second prompt:

This is a [Hot/Warm/Cold] lead scored [X/10].

Lead context: [paste key details]

Write a [phone call opener script / personalized email / text message] for my first contact with them. Tone should match urgency: Hot = immediate and direct; Warm = helpful and educational; Cold = low-pressure value delivery.

My key differentiators: [e.g., 21-day close, no overlays on FHA, same-day pre-approval letters]

Part 4: Add Zapier Automation (Optional)

If you want to automate the intake:

1. Trigger: New row in Google Sheet (or new lead form submission via Typeform/website)
2. Action: Send lead data to OpenAI API with your scoring prompt
3. Action: Write score back to Google Sheet in the AI Score column
4. Action: If score >= 8 (Hot): send you a text alert immediately
5. Action: If score 5–7 (Warm): add to a 3-day email sequence
6. Action: If score <= 4 (Cold): add to a long-term drip campaign

This means: every new lead is automatically scored, and you only get interrupted for Hot leads.

Real Example: Full Workflow

Incoming lead (website form):

• Name: David Chen
• Source: Google Ads (website form)
• Timeline: "Looking to buy in the next 2–3 months"
• Pre-approval: "Not yet — want to understand the process first"
• Down payment: "We have about 5% saved"
• Credit: "I think around 700"
• Employment: "Both W2, been at our jobs for 3+ years"

Step 1 prompt output:

Step 2 prompt output — Email:

Time spent: 3 minutes to score and generate the outreach. Instead of calling David first and spending 20 minutes on an educational call while a hot lead waits — you send David an email and call the 8/10 lead immediately.

What to Do When It Breaks

• AI scores every lead as Hot: Your rubric may be too generous — recalibrate the scoring instructions with stricter criteria; test with 10 past leads you know closed vs. didn't close
• Scoring takes too long to be practical: Simplify to 3 factors (timeline, source, pre-approval status) and score manually using AI just as a sanity check
• Zapier step fails: Check that your Google Sheet columns are named exactly as referenced in Zapier; re-authenticate connections

Variations

• Simpler version: Score leads manually using the rubric (no AI needed) — just having the rubric improves prioritization immediately
• Extended version: Track lead scores over time; correlate with close rates to refine your rubric based on real data (what signals actually predict a closed loan?)

What to Do Next

• This week: Define your scoring rubric using the AI prompt; score your last 20 leads to calibrate it
• This month: Apply scoring to all new leads for 30 days; see if your time-to-call on hot leads improves
• Advanced: Connect to your CRM's lead assignment logic so Hot leads automatically get tagged for immediate callback

Advanced guide for loan officer professionals. Lead scoring is a judgment tool — treat AI scores as helpful guidance, not absolute decisions. Always review lead context before deciding on outreach priority.