Classify Intercom conversations and alert sales on buying questions
This agent reads each new chat in the support inbox, sorts it by type and urgency, and alerts sales when it is really a buying question. Sales leads stop waiting in line behind bug reports.
Paste this into ChatGPT, Claude or your agent tool. See the prompt
- CS / Expansion
- beginner
- CS
- SDR / BDR
- Category
- CS / Expansion
- Difficulty
- Beginner
- Est. cost
- Not stated by source
Problem
Sales inquiries arrive in the support inbox and wait in the same queue as bug reports. Manual triage is slow and inconsistent.
Setup
An n8n webhook receives each new Intercom conversation and sends it to a small OpenAI model with a structured prompt. The model returns category (support, product, sales, other), sentiment, urgency, reasoning and tags; a Code node merges this with customer details. Support and product requests become ClickUp tasks, and sales inquiries post to a Slack sales channel with the context and AI reasoning.
Prompt
Starter prompt by The GTM Agent Weekly, written from the setup described on this page. It is not the source's own prompt; see the source link below for their write-up.
# ROLE
You triage each new customer conversation from the support chat and flag buying questions to sales.
# INPUTS
- {{CONVERSATION}}: the messages so far.
- {{CUSTOMER}}: name, email, company, plan, account owner if known.
# TASK
Classify the conversation and return structured data:
1. category: support, product (feedback or feature request), sales (pricing, upgrade, new
purchase, contract, demo request) or other.
2. sentiment: positive, neutral or negative.
3. urgency: low, medium or high, with a short reason.
4. tags: up to three short tags.
5. reasoning: one or two sentences explaining the category.
# OUTPUT FORMAT (JSON only)
{"category": "", "sentiment": "", "urgency": "", "urgency_reason": "", "tags": [], "reasoning": ""}
# ROUTING (done by the workflow)
- support and product: create a task in {{TASK_TOOL}}.
- sales: post to {{SALES_CHANNEL}} with customer details, a quote of the question and your reasoning.
# GUARDRAILS
- When unsure between support and sales, choose sales only if they asked about price,
plans, seats, contracts or buying.
- Never reply to the customer. Replace every {{PLACEHOLDER}} before running. Works with any capable LLM and runner.
Tools
Steps
- Point an Intercom webhook for new conversations at the n8n workflow.
- Classify with structured JSON output: category, sentiment, urgency, reasoning and tags.
- Parse the output in a Code node and merge it with customer info.
- Route support and product to task creation and sales to a Slack alert.
- Review a sample of classifications weekly to tune the prompt (editor note).
Results
Workflow template; no outcome data.
What goes wrong
Gotchas
- A first message often lacks enough detail to classify; consider reclassifying after the customer replies (editor note).
- Sales alerts with no owner get ignored, so route to a named rep or rotation (editor note).
- Malformed model JSON breaks the Code node unless you validate it (editor note).
Get 5 like this every week, free.
Real agent setups with sources and failure modes, in one short weekly email.
Free. One email a week. Unsubscribe anytime.
Source and credit
Avkash Kakdiya, n8n template · vendor
Source dated 2026-10-11 · last reviewed 2026-10-11. Summarized in our own words; see the source for details.