Playbook · Lead qualification

AI Lead Qualification: A Playbook for Small Business

Published by Nexus AgentWorks · Playbook series

Every inbound form fill feels like good news until you triage it by hand. A two-person services firm can easily collect a dozen inquiries a week, each a different mix of "ready to buy," "just researching," and "wrong industry entirely." Someone reads every message, guesses who matters, and lets the rest rot in the inbox. If you want to automate lead qualification without turning your funnel into a spam cannon, this playbook shows how a rules-first scoring agent does it — and why ours never contacts anyone.

The manual triage tax

Inconsistent judgment. The lead that gets a fast reply on Monday gets ignored on Thursday, depending on who's tired. Your close rate quietly depends on mood.
Slow first response. Leads go cold in hours. Manual triage means your best-fit prospects wait behind obvious tire-kickers.
No audit trail. When sales asks "why didn't we chase this one?", nobody can reconstruct the reasoning.
The over-correction trap. Most automation tools "fix" this by blasting every lead with sequences — which burns your reputation with the very prospects you wanted.

Scoring rules you can actually read

Our Lead Qualification Agent uses a deterministic rule scorer — the same logic ships in our template code (LeadScorer) — that produces a score from 0–100 plus a plain-language reason list for every decision:

Service-fit keywords (up to +40). The agent scans the inquiry message and requested service for terms like agent, automation, support, inbox, chatbot, voice, workflow. Each hit adds points, capped at 40, so keyword-stuffing can't dominate.
Budget threshold (+30). A stated budget at or above your minimum (default $2,000) adds 30 points. Below it earns no points but is recorded as a reason, not silently dropped.
Business domain (+15). A real company email domain scores; free/test domains don't. Banned domains (e.g., competitors) are hard-disqualified regardless of everything else.
Detailed-brief intent signal (+15). Messages longer than ~120 characters signal genuine intent — serious buyers describe their problem; one-liners rarely do.

The total maps to three tiers: hot (70+), warm (40–69), cold (below 40). Because every point traces to an explicit rule, you can tighten or loosen thresholds in one config file instead of retraining anything.

What happens per tier

Hot → route_to_sales_queue. The lead lands at the top of your sales queue with its full reason list attached, so a human calls knowing exactly why it scored well.
Warm → nurture_content. Not ready yet — flagged for your existing newsletter or content track rather than a sales call.
Cold → disqualify. Archived with the recorded reasons. Nothing is deleted, so you can spot patterns (a pricing page attracting the wrong audience, say) later.

The hard no-contact guarantee

This is deliberate and non-negotiable: the agent never messages anyone. Its output is limited to routing decisions — route_to_sales_queue, nurture_content, or disqualify. There is no outbound email step in the scorer, no SMS, no auto-reply. A human always makes first contact. That keeps you compliant with anti-spam expectations, protects your sender reputation, and means the worst case of a misconfigured rule is a mislabeled row in a queue — never an embarrassing message sent to a prospect.

A worked example

Simulated example — illustrative numbers only, not a real customer. An inbound lead arrives with service "voice support agent", message "We run a 3-location dental practice and miss after-hours calls constantly. Looking for an automated voice agent with escalation to staff, budget around $4,500.", and a business-domain email address:

Service fit: "agent", "voice", "support" → +30 (three keyword hits)
Budget $4,500 ≥ $2,000 minimum → +30
Business domain → +15
Message length > 120 chars → +15

Total: 90 → tier "hot" → next_action: route_to_sales_queue, with all four reasons logged. A one-line "how much?" from a personal Gmail would have scored cold and been archived instead of eating a founder's afternoon.

Start small, measure honestly

Pick one intake channel (usually the website contact form). Run the scorer in shadow mode alongside manual triage for two weeks and compare calls. Track time-to-first-human-touch on hot leads, the hot/warm/cold distribution, and how often the human disagrees with the score — disagreements are how you tune thresholds. Lead qualification automation pays off precisely because it removes a repeated judgment call, not because it sends more email.

Want to see the scorer make decisions live? Watch it route simulated leads in our live demo, then check pricing to see what a deployment costs.