Clear the support queue, keep nothing

Solo founders and 5-20 person SaaS teams whose founders still handle support personally or have 1-2 junior agents drowning in volume · Assemble in ~10 min; first triage batch runs in under 5 min once installed.

ProductEdit flowon X● golden 87

step 1connector

Pull and normalize inbox

⚡ bring your own tool

step 2ready

Classify and tag each ticket

triage-support

step 3ready

Surface and prioritize what needs attention

triage-messages

step 4partial

Route tickets to owner

taskflow-inbox-triage

step 5ready

Draft contextual reply for each ticket

dm-respond10 alts

step 6ready

Polish and finalize drafts

email-drafter

step 7connector

Send approved replies

⚡ bring your own tool

Did this recipe work for you?

What you get

▸Prioritized ticket digest (type, urgency, sentiment) ready to skim in under 2 minutes
▸Routed assignment list — each ticket assigned to founder, agent, or auto-reply lane
▸Polished draft replies for every ticket, human-approved before sending
▸Closed ticket list with send timestamps for support SLA tracking

tools: Gmail / Freshdesk / Zendesk account (existing) — no extra cost; helpdesk API access may require paid tier

tokens: ~$1-4 per 50-ticket batch depending on ticket length

≈ $5-20/mo in tokens at 50 tickets/day; helpdesk SaaS cost separate

What stays yours

· Review and approve each draft before sending — one-click approve is realistic, but a human must stay in the send loop
· Flag edge cases (legal complaints, billing disputes, churn escalations) that need personal attention before reply
· Tune the ticket taxonomy and routing rules every 2-4 weeks as ticket types shift

✓ Last verified 2026-06-11claude-sonnet-4-6 (subagent, test-mode fixtures)

✓ step 1 · Pull and normalize inbox

✓ step 2 · Classify and tag each ticket

✓ step 3 · Surface and prioritize what needs attention

✓ step 4 · Route tickets to owner

✓ step 5 · Draft contextual reply for each ticket

✓ step 6 · Polish and finalize drafts

✓ step 7 · Send approved replies

view sample outputs from this run

step 2 · Classify and tag each ticket

# Classified Ticket List
Generated: 2026-06-11T09:05:00Z | Skill: triage-support | Batch: 8 tickets

---

## T-001 — Cannot log in / locked out
- **Type:** Bug
- **Urgency:** P0 — CRITICAL (time-sensitive: campaign live TODAY)
- **Sentiment:** Distressed / Frustrated
- **Evidence:** "big sales campaign going live TODAY", "urgently", "ASAP"; password reset not resolving — likely account state issue not user error
- **Tags:** `auth` `login` `P0` `time-sensitive`

---

## T-002 — Webhook `order.completed` not firing
- **Type:** Bug
- **Urgency:** P1 — HIGH (blocking production pipeline)
- **Senti…

step 3 · Surface and prioritize what needs attention

# Prioritized Attention Digest
Generated: 2026-06-11T09:10:00Z | Skill: triage-messages | Batch: 8 tickets

---

## IMMEDIATE — Respond within 1 hour

### 🔴 T-006 · James Whitmore / Orbitek — CHURN SIGNAL — Pro plan cancel intent
- **Why first:** Explicit "cancel end of month" — 3-month Pro customer leaving. Revenue at risk. Window to retain is open but closing.
- **Action required:** Founder call. Do not send a template reply. Offer a 30-min Zoom, ask what success would look like, identify the exact features they're not using, propose a tailored onboarding session.
- **Route to:** Founder / c…

step 4 · Route tickets to owner

# Routed Ticket Assignments
Generated: 2026-06-11T09:15:00Z | Skill: taskflow-inbox-triage | Batch: 8 tickets

Routing rules applied:
- P0 bugs → Engineering on-call
- P0 churn signals → Founder (non-delegatable)
- Billing disputes → Finance/billing ops
- Legal/compliance → Legal/ops
- How-to + feature requests → Junior support agent / auto-reply

---

## QUEUE: FOUNDER

| Ticket | Sender | Subject | Due | State |
|--------|--------|---------|-----|-------|
| T-006 | james.whitmore@orbitek.com | Thinking of canceling — not getting value | TODAY 12:00 | WAITING: founder to schedule call |

**Ro…

The problem

When I get 50+ support emails or tickets daily as a solo founder or tiny team, I want an AI that classifies each ticket by type, urgency, and sentiment, routes it, and drafts a contextual reply, so that I stop spending 3+ hours a day on support and can focus on product and growth.

Cybernetic layers

Memorybyo

Ticket history, per-customer context, and known issues list. Use a local markdown KB file, Notion page, or Freshdesk KB — no harvested memory artifact yet. Adding a `support-kb.md` and passing it into each session is the highest-leverage DIY step.

Background / schedulingbyo

The batch should run 1-2x daily unattended. Wire a Claude Code cron or a simple shell script on a schedule. No library artifact yet for support-specific scheduling.

Harnessready

Claude Code is the runtime. Drop each ready skill into ~/.claude/skills/ and invoke by name. The batch runs in a single session without persistent agent loops.

Click any step to see what fills it — and the ranked alternatives you can swap in.