MODULE 02 · LEAD ENGINE AGENT

Stop chasing leads. Build a fleet that hunts.

Sales is a pipeline problem. A single AI assistant won’t fix it. You need a coordinated fleet of specialized agents: one to hunt, one to score, one to enrich, one to reach out, one to track. Each does its job. The pipeline never sleeps.

The 5-Agent Fleet

One agent = a chatbot. Five specialized agents working together = a sales operating system. Each link does one job exceptionally well, then hands off to the next.

Hunter

Finds raw companies matching ICP signals daily

Scorer

Ranks 0-100 across firmographic + intent traits

Enricher

Pulls decision-maker + context from open data

Outreach

Drafts personalized sequence, you review & send

Tracker

Logs replies, schedules follow-up, flags hot

Build Your ICP Scorer

Don’t outreach to everyone — qualify first. Tap the traits that describe your ideal customer. Watch the score build. 70+ = drop the outreach agent on it.

Click traits that match a hypothetical prospect. Each one weighs the score. Use this as a template — swap signals for your business.

FIT SCORE

out of 100

SELECT TRAITS

The Pipeline, Visualized

Here’s what a real Lead Engine output looks like over a week. The funnel narrows. The volume scales. The cost stays flat.

Companies scanned (Hunter)2,400/wk

↓ score ≥ 50

Qualified prospects (Scorer)320/wk

↓ enriched + verified

Enriched contacts (Enricher)280/wk

↓ drafted & approved

Outreach sent (Outreach)140/wk

↓ replies + booked

Meetings booked12/wk

The KPIs That Matter

Vanity metrics waste calories. Track these four — they tell you whether the engine is firing or stalling.

Hunter Yield

2.4k

↑ companies/wk

Qualified Rate

13%

↑ +4pts vs manual

Reply Rate

8.5%

↑ +3.2pts vs cold

Cost / Meeting

$47

↓ −68% vs SDR

Your Fleet Config Template

Drop this into your agent runtime. Adjust signals to your ICP. Start with the Hunter — get its yield right before bolting on the others.

lead-engine.fleet.yaml

# Hunter — finds raw matches daily hunter: sources: [“crunchbase”, “linkedin”, “news_feed”] filters: [“hiring_signal”, “funding_recent”, “icp_industry”] cadence: “daily_at_07:00” # Scorer — 0-100 fit ranking scorer: weights: {size: 25, industry: 20, intent: 30, recency: 25} threshold: 50 # promote to Enricher # Outreach — drafts, you approve, sends outreach: channel: “email_first_linkedin_followup” sequence: [“intro”, “value_prop”, “case_study”, “break_up”] human_review: true # always — agent drafts, human ships

What to Automate vs. What to Keep Human

Step	Automate	Why / Why Not
Find companies	✓ FULL	Pure pattern-matching. Agent crushes it.
Score fit	✓ FULL	Deterministic weights. Reproducible. Audit-able.
Enrich contact	✓ FULL	Open data lookups. No judgment needed.
Draft message	✓ DRAFT	Agent writes. Human reviews. Always.
Hit send	✗ HUMAN	Brand voice. Tone. Accountability. You own it.
Negotiate / close	✗ HUMAN	Trust transfers through people, not bots.

⏱ THE 30-DAY RAMP

Don’t build all five at once.

Week 1: stand up Hunter alone. Get to 1k+ companies/wk scanned. Week 2: add Scorer. Tune threshold until top decile feels right. Week 3: layer Enricher + Outreach drafting. Hit send manually. Week 4: connect Tracker, automate follow-ups, measure cost per meeting. By day 30 you have a Lead Engine that runs while you sleep.

Common Failure Modes

1Hunter casts too wide → garbage in, score noise out. Fix: tighten ICP filters first.
2Scorer trusts ML over rules → unexplainable verdicts. Fix: start with deterministic weights, layer ML later.
3Outreach auto-sends → brand burns. Fix: human-in-loop on every send for the first 90 days.
4Tracker silent on losses → no learning loop. Fix: log every “no” with reason code, retrain monthly.

Module Recap

A Lead Engine is not a tool. It’s a fleet of agents wired into a deterministic workflow with human checkpoints at every send. Build Hunter first. Measure yield. Add layers only when the previous one is producing clean output. The goal isn’t replacing sales reps — it’s giving them 10x the qualified at-bats.