编程

Inbound Call Triage Brief

试用

Triage inbound calls to an AI agent line — classify caller intent, screening questions, escalate-vs-handle rules, message-taking script, and spam signals bef...

它能做什么

Triage inbound calls to an AI agent line — classify caller intent, screening questions, escalate-vs-handle rules, message-taking script, and spam signals before the agent picks up.

技能文档

Inbound Call Triage Brief

Purpose

Produce an inbound triage brief when a call arrives (or is expected) on an AI agent phone line — PollyReach, Agent Phone Call, OpenPhone, Twilio, etc. Covers intent classification, first questions, escalate-to-human rules, message-taking, and spam detection.

This skill does not answer calls, provision numbers, or access telephony APIs. It gives the agent (or human supervisor) a playbook for this caller/context.

Pairs with phone-call-prep-brief (outbound) and agent-phone-call-debrief (after-call).

When to use

Use when:

  • Setting up how an AI line handles incoming calls
  • A call is ringing and the agent needs a quick routing script
  • Defining receptionist / front-desk behavior for an agent number
  • User asks to screen, triage, or route inbound callers

Safety and boundaries

Do not collect passwords, OTP codes, full card numbers, or SSNs — even if callers offer them.

Emergency: If caller describes immediate danger (medical, fire, violence), instruct transfer to emergency services (e.g. 911) — do not attempt AI-only handling.

Harassment/spam: Document disengage script; do not engage debate.

Regulated industries (health, legal, financial): default to human handoff for advice-giving.

Recording: Include disclosure line if calls are recorded (high level only).

Required inputs

  1. Line purpose — support, sales, booking, personal screening, mixed.
  2. Business/context — who the number represents (company, individual, service).
  3. Caller hint (optional) — known number, CRM name, or “unknown.”
  4. Hours — 24/7 AI vs business hours with after-hours behavior.
  5. Human backup — who can take escalations and how (warm transfer, callback).
  6. Must-capture fields — name, callback number, account ID, etc.

Workflow

  1. Classify expected caller intents for this line (≤6 categories).
  2. For each intent: handle / message / escalate.
  3. Draft opening script (≤20 seconds spoken).
  4. Define escalation triggers (keyword, emotion, VIP, legal threat, etc.).
  5. Output triage brief (format below).

Intent categories (starter set)

Adapt to line purpose; pick 3–6:

IntentTypical handle
Booking / schedulingAgent collects slots → confirm
Order / account statusVerify identity lightly → lookup or escalate
Support issueTroubleshoot tier-1 → escalate if unresolved
Sales inquiryQualify → book or escalate
Wrong number / spamPolite disengage
VIP / urgentFast-track human

Output format

Inbound triage brief — {line or business}

FieldValue
Line purpose
After-hours
Human backup
Recording disclosureyes/no + one-liner

Opening script

What the agent says when answering (friendly, states who they represent, sets expectation).

Intent routing table

IntentFirst questionsAgent can resolve?Escalate when
yes / partial / no

Screening questions (unknown callers)

Numbered list, minimal PII.

Message-taking template

If taking a message: fields to capture + read-back script.

Spam / robocall signals

Behaviors that trigger short disengage (silent line, obvious pitch, repeated hang-up).

Escalation pack

  • Trigger list
  • Warm handoff line: “I’m connecting you with …”
  • If human unavailable: callback promise script

Do not do on this line

Bullets (e.g. process payments, give medical advice, promise refunds without policy).

Quality bar

  • Inbound-first — opening script assumes caller initiated contact.
  • Escalation is explicit — not “use judgment.”
  • Short scripts — speakable in one breath where possible.
  • PII minimal — collect only must-capture fields.

Examples

Good opening: “Thanks for calling Acme Support — I’m the automated assistant. I can help with order status or scheduling a callback. What are you calling about today?”

Bad opening: “Hello, how can I help?” (no context, no boundaries)

Good escalate trigger: “Caller requests chargeback over $500 or mentions attorney.”

Bad escalate trigger: “If the caller is angry.” (too vague)

相关技能

Debrief after an AI agent phone call — compare outcome to goal, capture commitments, flag escalations, and produce follow-up tasks and CRM-ready notes.

2 次安装

Build a one-page call brief before an AI agent or human places or takes an important phone call — objective, talking points, consent/disclosure checklist, vo...

1 次安装

The phone is your last API. Give your agent a voice to call any US business and take action in the real world. PlaceCall handles reservations, inquiries, and quotes - navigating IVRs, holds, and transfers. You get a verified outcome + full transcript. First 250 calls free. Pay only for outcomes.

Make AI phone calls with OpenClaw. Use when the user says "make a call", "call this business", "book by phone", "call customer service", "schedule a phone call", or wants an AI receptionist. ClawCall makes confirmed outbound calls, finds public business numbers, receives inbound calls, schedules calls, and returns transcripts, summaries, recordings, and costs. Supports international E.164 numbers; use ai-calls-china-phone for mainland China numbers.

Triage an incoming insurance claim: check the coverage trigger against policy wording, band severity and complexity, screen for fraud indicators, set a first...

Make phone calls, view received SMS, provision numbers, manage agents, and track billing through the AgentLine telephony API. Use when the user asks to call...

19 次安装