# Lead qualification scorecard

A fill-in scorecard plus a copy-paste prompt that grades every inbound lead the same way, before anyone books a call. It works with any capable AI assistant: ChatGPT, Claude, and Gemini all handle it fine.

The problem it solves: most teams qualify leads by gut feel. Whoever reads the inquiry first decides whether it deserves a call, and two people will make two different calls on the same lead. A scorecard makes the criteria explicit. The AI applies them consistently, at any hour, without getting bored.

## Prerequisites

- Any capable AI assistant with a chat interface (ChatGPT, Claude, or Gemini)
- 10 or more past inbound leads whose outcomes you already know (closed, ghosted, bad fit)
- 30 minutes to define your criteria honestly

## Step 1: Fill in your scorecard

Pick five criteria and weight them so they total 100. Steal this structure, but set your own weights — they encode what actually matters in your business.

| Criterion | What to look for | Weight |
|---|---|---|
| ICP fit | Company size, industry, and location match your best customers | ___ |
| Problem match | They describe a problem you actually solve, in their own words | ___ |
| Authority | The person writing can sign, or clearly reports to someone who can | ___ |
| Budget signal | They ask about pricing, mention budget, or their company visibly spends on this category | ___ |
| Urgency | A deadline, a trigger event, or "we need this by" language | ___ |

Then define three score bands and what happens in each. For illustration only: 70 and up books a call today, 40-69 goes into a nurture sequence, under 40 gets a polite decline. Set your own thresholds in Step 4, using your own data.

## Step 2: Load the prompt

Open a new conversation with your AI assistant. Paste the prompt below and replace every [BRACKETED] section with your details before sending.

## Step 3: The prompt

```
You are a lead qualification analyst for [COMPANY NAME], which sells
[WHAT YOU SELL] to [WHO YOU SELL IT TO].

Score the lead below against this scorecard. Total possible: 100.

1. ICP fit (weight: [X]) — our best customers look like:
   [DESCRIBE: size, industry, geography]
2. Problem match (weight: [X]) — problems we solve:
   [LIST 2-4 PROBLEMS IN PLAIN LANGUAGE]
3. Authority (weight: [X]) — decision-makers are usually: [TITLES/ROLES]
4. Budget signal (weight: [X]) — signals we look for:
   [e.g., asks about pricing, mentions budget, company size implies spend]
5. Urgency (weight: [X]) — signals we look for:
   [e.g., a deadline, a trigger event like new funding or a new hire]

Rules:
- Use ONLY the information provided. Do not guess or invent facts about
  the lead or their company.
- For each criterion, give a score out of its weight plus one sentence of
  justification that quotes or references the lead's actual words.
- If the lead gives no information on a criterion, score it 0 and mark it
  "UNKNOWN — ask on the call."
- End with:
  - TOTAL: X/100
  - BAND: PURSUE / NURTURE / DECLINE (bands: [YOUR THRESHOLDS])
  - Two specific questions a human should ask in the first five minutes
    of a call, targeting the UNKNOWN or lowest-scoring criteria.

Lead:
[PASTE THE FULL INQUIRY: form fields, email text, LinkedIn message —
everything you have]
```

## Step 4: Calibrate before you trust it

Run the prompt on 10 past leads whose outcomes you know. Then check three things:

1. **Ranking sanity.** Did your actual best customers score near the top and the bad fits near the bottom? If not, your weights are wrong, not the AI. Adjust and re-run.
2. **Evidence discipline.** Read every justification. Each score must point to something the lead actually wrote. If the model assumed facts, tighten the "use only the information provided" rule or switch to a stronger model.
3. **Threshold fit.** Set your bands where your real data says they belong. Bands you didn't calibrate are decoration.

Re-test after any change to weights or criteria. Recalibrate whenever your offer or pricing changes.

## Day-to-day use

Keep the filled-in prompt in a saved note, or as a custom instruction or project inside your assistant. When a lead comes in, paste the inquiry, get the grade, and let the band decide the next action. The human still makes the final call — the scorecard just guarantees every lead got the same first look.

## When this outgrows DIY

Copy-paste scoring works fine at a handful of leads a week. It stops working when leads arrive faster than someone remembers to paste them, when scores need to land in your CRM automatically, or when the grade should trigger the next step by itself: booking links for high scores, nurture sequences for the middle band. At that point it is no longer a prompt. It is a system with intake, scoring, routing, and logging. Building those systems is what we do at agentclaw. If you are at that point, book a free AI opportunity audit at agentclawhq.com/book.