Disqualification Handling

Disqualification Handling

A composite: categorize.py tags each lead's handling category deterministically; the agent drafts the right response per category (via email-drafting); then launch_campaign.py sends/routes (one-off decline or nurture) — or stops at draft-only with no send key.

When to use

• "Handle the disqualified leads" / "draft rejection emails" / "what do we do with the ones that didn't qualify?"
• After an inbound-lead-qualification step produces disqualified/near-miss leads.

How it works (scripts + agent)

1. Categorize — categorize.py maps each lead's sub_verdict / mismatch_type to a handling category + action:
   • referral (right-company/wrong-person) -> referral ask (highest value)
   • nurture (close-but-not-quite) -> nurture sequence
   • decline (clearly outside ICP) -> polite decline + do-not-contact
   • competitor (competitor employee) -> LOG only, NEVER email
   • existing_customer -> route to CS/upsell, never a decline
2. (agent) For referral, pick the referral-target persona from ICP (or pull from CRM / Apollo if known). Draft the appropriate response per category with email-drafting, matching the configured tone; include a resource link if set.
3. Route / send — nurture -> launch_campaign.py --tool lemlist/instantly; one-off decline -> --tool sendgrid/resend; competitor/CS flags + do-not-contact -> store write + slack CS handoff (agent). Draft-only mode (no send key) just returns the drafts.

How to run

# 1 — tag handling categories
python3 ${SKILL_DIR}/scripts/categorize.py \
  --input ${WORKSPACE}/disqualified.json --output ${WORKSPACE}/tagged.json

# 3 — send a one-off decline (only the 'decline' leads; agent drafts seq.json first)
python3 ${SKILL_DIR}/scripts/launch_campaign.py \
  --tool resend --campaign-name "polite declines" \
  --leads ${WORKSPACE}/declines.json --sequence ${WORKSPACE}/decline_msg.json

# nurture routing
python3 ${SKILL_DIR}/scripts/launch_campaign.py \
  --tool instantly --campaign-name "nurture" \
  --leads ${WORKSPACE}/nurture.json --sequence ${WORKSPACE}/nurture_seq.json

Outputs

• tagged.json — each lead {handling_category, handling_action, referral_target_persona}, the agent's drafted responses per category, and the routing/send results; summary to channel.

Credentials / env

• env.required: none. The default is draft-only (categorize + draft, no send), and --tool generic_csv exports a tool-ready CSV with no key.
• env.optional (all degrade): Send/route — if a send key (SENDGRID_API_KEY / RESEND_API_KEY for one-off declines, LEMLIST_API_KEY / INSTANTLY_API_KEY for nurture) is set → send/route the response; else → draft-only or export a CSV (--tool generic_csv, the keyless default). HUBSPOT_API_KEY — if set → CRM referral lookup + routing/tagging; else → manual. APOLLO_API_KEY — if set → find the referral contact; else → ICP persona only. SUPABASE_URL + SUPABASE_SERVICE_ROLE_KEY — if set → do-not-contact list / competitive-intel log; else → workspace CSV ledger.

Notes & edge cases

• NEVER send a rejection to a competitor employee — log for competitive intelligence and stop.
• Existing customers are not disqualified — route to CS / upsell, never a decline email.
• Every decline must preserve the relationship; add declined leads to the do-not-contact list.
• Right-company/wrong-person is the highest-value path — a referral ask is warmer than cold.