Lead Classifier
Score and classify an inbound sales lead from its profile/message, then suggest the single best next action so reps work the right deals first.
What it does
Given one inbound lead, the skill scores four dimensions (0–25 each) and totals them to a 0–100 score:
- Authority — buying power inferred from job title.
- Intent — buying signals mined from the message and lead source (demo/pricing/budget/timeline phrases).
- Fit — ICP match from company size and industry.
- Quality — contact validity (business vs. free/disposable email, name, phone, country).
From the total it assigns a tier — HOT (75–100), WARM (50–74), COOL (25–49), or COLD/SPAM (0–24) — flags red flags (disposable email, empty message, job application, competitor/test submission, domain mismatch), and recommends exactly one next action sized to the tier.
What to feed it
A single lead as JSON. All fields are optional; missing fields are scored as neutral, never invented:
{
"name": "Jane Doe",
"email": "[email protected]",
"company": "Acme Industrial",
"title": "VP of Operations",
"phone": "+1-555-0100",
"companySize": "500-1000",
"industry": "Manufacturing",
"source": "Pricing page form",
"message": "Evaluating automation platforms for Q3, need a demo this week. Budget approved.",
"country": "US"
}
What it returns
Two parts, in order:
- A human summary — lead line, tier + score, per-dimension scores, recommended next action, red flags, and a 2–3 sentence rationale.
- A machine JSON block with
tier,score,dimensions,nextAction,redFlags, andrationale— safe to parse and route into a CRM or a follow-up job.
Empty or unparseable input returns tier COLD/SPAM, score 0, and the red flag unparseable-input.
Tips
- Feed the rawest profile + message you have; the skill is most accurate when the
messageandsourceare present. - Pair it with an InTouch job: classify on inbound, then branch — route HOT leads to an AE notification, drop COOL leads into a nurture list.