All templates
Deliverystarter
Support Ticket Triage
Inbound support ticket → AI categorises, drafts reply, routes to right human if needed.
Problem it solves
Support team spends 30% of time triaging vs answering. First-response SLA slips.
Trigger
webhook
Intercom/Zendesk webhook on new ticket
Time to ship
1 week
Data sources
3
Support inbox (Intercom/Zendesk/email)
Inputs
- 1.Support inbox (Intercom/Zendesk/email)
- 2.Past tickets + resolutions
- 3.Product docs (vault)
Outputs
- 1.Categorised ticket (bug / billing / how-to / feature-request)
- 2.Drafted reply (human approves)
- 3.Routed to right teammate
Scoring checklist
Every agent output is checked against these BEFORE a human reviews it. The agent self-loops on misses.
- Category accuracy ≥ 90% on backlog tickets
- Drafted reply needs <20% edits before send
- Routing matches the human team's intuition ≥ 95%
- First-response SLA dropped by ≥ 50%
Build this against your own data
The wizard's Step 7 walks through every command — schema probe, bootstrap SQL, Edge Function template, deploy, smoke test.
Open Step 7 — Ship First Agent