[your_name].dev
Demo B · AI Workflow Automation

InboxTriage

Every inbound email lands in one queue. This pipeline reads each one and classifies it, assigns a priority, routes it, and drafts a reply — an AI decision step wired into a real automation, not just node-plumbing.

Demo mode: the results below are pre-computed and replayed in your browser — no API key, no cost, nothing leaves this page. The live version connects to the Claude API to produce the same category / priority / draft fields per email. All 20 emails are synthetic data for a fictional SaaS, "Lumina Notes".
01 · TRIGGER

New email

Inbound hits a shared inbox / webhook (Gmail, Front, a form, an IMAP poll).

02 · AI STEP

Classify + score

Claude reads it, returns a category, priority, confidence & sentiment as strict JSON.

03 · ROUTE

Branch

Rules on the AI output send it to the right queue — finance, engineering, sales, or archive.

04 · AI STEP

Draft reply

A second, guard-railed call drafts a reply in your brand voice (or skips it for spam).

05 · ACT

Log + notify

Write to CRM/sheet, save the draft, and ping the right Slack channel. A human approves & sends.

Processed 0/20 P1 urgent 0 Auto-archived 0 Est. handling 0s vs ~8 min/email manual
Run the pipeline, then select an email to see the AI's classification, reasoning and drafted reply.

The AI step is one structured call. The prompt pins the company voice, forces a fixed category set, and demands JSON so the automation can branch on it deterministically — the model decides, but never freestyles the schema.

SYSTEM You are the triage assistant for Lumina Notes, a collaborative note-taking SaaS. For each email, return JSON only: { category: one of [Billing, Bug Report, Feature Request, How-to, Sales, Spam], priority: one of [P1, P2, P3], confidence: 0.0–1.0, sentiment: short label, needs_reply: boolean, draft: reply in our voice, or null } Rules: - If confidence < 0.6, do NOT guess the issue — set draft to a clarifying question. - Never invent policy, prices, or facts. - P1 only for: outage, billing incident, legal/GDPR, or explicit time-pressure. - Spam & cold outreach → needs_reply:false.

Guardrails are what make it safe to put in front of real customers. Each is visible in the demo data:

  • CONFIDENCE GATELow confidence → ask, don't guess. Email #15 ("help / it doesnt work") scores 0.55, so the draft is a clarifying question instead of an invented fix.
  • NO HALLUCINATED FACTSVoice & policy pinned in the prompt. Prices, limits and legal timelines come from your knowledge base, not the model's imagination.
  • PRIORITY RULESP1 is reserved. Only outages, billing incidents, legal/GDPR (#16) and explicit time-pressure (#7) escalate — so urgent things aren't buried.
  • HUMAN IN THE LOOPDrafts, never auto-sends. Every reply lands as a draft for a human to approve — the AI proposes, a person ships.
  • SILENT ON SPAMneeds_reply:false. Scams (#5) and cold outreach (#19) are archived with no draft and no human time spent.

Point it at your real inbox (Gmail, Outlook, Front, Help Scout, or a contact form) and the pipeline above runs on every message as it arrives. Your team stops triaging by hand: urgent billing and outage emails jump to the top, spam disappears, and a ready-to-edit draft reply is waiting for each real one. A person reviews and hits send — so quality and tone stay yours, but the sorting and first draft are done.

TYPICAL FIRST-PASS TIME
~8 min → ~3 s
DELIVERED AS
Importable n8n / Make workflow you own
YOUR EFFORT
Review & send — not sort & write