AI Lead Scoring: Claude AI Rates Every LinkedIn Prospect

The Problem with Manual Lead Qualification

You run a LinkedIn search. You get 30 results. Now what?

With every other LinkedIn tool, you either:

• Send to everyone — waste connections on bad-fit prospects, dilute your acceptance rate, risk account restrictions
• Manually review each profile — spend 2-3 minutes per person, 60-90 minutes for 30 profiles, every single day
• Use keyword filters — miss good prospects who don't have the exact keywords in their title, catch false positives who do

None of these approaches actually work at scale. You need something that can read a profile and reason about whether this person is a good fit — the way you would, but in seconds instead of minutes.

How LeadPilot Scoring Works

LeadPilot sends each scraped profile to Claude AI with your ICP definition and scoring criteria. Claude reads the full profile — not just the title, but the about section, experience, company info, education — and scores them on 5 weighted criteria:

1. Role Fit (30% weight)

Is this person a founder, CEO, CTO, or decision-maker? Or are they a VP Sales at a large enterprise who can't actually buy? Claude understands the difference between "CEO @ 10-person startup" and "VP at BigCorp" even when both have "leadership" in their title.

2. Industry Match (25% weight)

Does their company operate in your target industry? Claude reads the company description and about section to determine this — not just keyword matching. A "patient scheduling platform" gets correctly classified as healthtech even if the word "healthtech" never appears.

3. Company Stage (20% weight)

Is this an early-stage startup, a growth company, or an enterprise? Claude infers company stage from employee count, founding date, funding mentions, and company description. Early-stage founders building SaaS products score highest.

4. Geography (15% weight)

Are they in your target geography? LeadPilot can prioritize prospects from specific regions based on your ICP. Western markets typically get higher scores for B2B SaaS outreach.

5. SaaS Signals (10% weight)

Are there signals that they're building or buying software? Keywords like "SaaS," "platform," "digital transformation," or mentions of specific technologies all contribute to this score.

Real Terminal Output

Here's what AI lead scoring actually looks like when you run leadpilot filter:

Not Keyword Matching — Actual Reasoning

This is the critical difference. Traditional tools use keyword filters: if the title contains "CEO" or "Founder," it's a match. This breaks immediately:

• False positive: "CEO of my own freelance consulting" — not a startup founder
• False positive: "Former CEO @ ..." — not currently leading a company
• False negative: "Building the future of healthcare compliance" — founder who doesn't use the word "founder" in their title
• False negative: "Co-creating a new standard for clinical trials" — clearly building a healthtech startup

Claude AI reads the full profile context and reasons about these nuances. It understands that someone "Building a patient scheduling platform with 5 team members" is a founder even without the word in their title.

Customizable ICP

Your ICP definition is fully configurable in config.yaml. Change the industries, role types, company stages, geographies, and scoring weights to match your exact target market.

Targeting fintech founders instead? Just change the industries. Looking for CTOs instead of CEOs? Change the roles. The AI adapts its scoring to whatever ICP you define.

From Score to Connection

The scoring pipeline feeds directly into the connection pipeline:

1. leadpilot search — find and scrape profiles
2. leadpilot filter — AI scores each profile against your ICP
3. leadpilot connect --dry-run — AI writes personalized messages for qualified leads only
4. leadpilot connect --max 22 — send connections with human-like behavior

You never waste a connection request on someone who doesn't fit. Every message is personalized. Every prospect is pre-qualified by AI.