Support Intelligence Cabinet

Two agents scan every inbound ticket, cluster recurring issues by theme, detect emerging problems before they hit the queue at scale, and ship a weekly Support Insights Report with CSAT trends, deflection rates, and concrete recommendations. Every cluster, every tagged ticket, and every report is a file you can open, share, and act on.

Keep Zendesk or Intercom. Replace the manual tagging, the Explore dashboards nobody interprets, and the "why do customers keep asking about X?" Slack thread nobody resolves.

Why this template

Most support teams have ticket volume data but not insight data. They know they're busy; they don't know why. This cabinet makes the ticket analysis a standing operation: clusters are rebuilt daily, trends are surfaced weekly, and the product and engineering teams always have a sourced view of the top pain points — not a survey, a living analysis of real tickets.

The team

• [[.agents/ticket-cluster-analyst]] — Ticket Cluster Analyst. Runs the daily clustering pass, tags tickets by theme, detects volume spikes, and flags emerging issues before they become incidents.
• [[.agents/insights-synthesizer]] — Insights Synthesizer. Owns the weekly Support Insights Report: pulls the week's clusters, CSAT, deflection metrics, and draft product pain recommendations. Orchestrates the final publish.

Recurring rhythm

Cadence	Job	Owner	Output
Daily (Mon–Fri 09:00)	[[.jobs/daily-ticket-cluster-scan]]	Ticket Cluster Analyst	Updated cluster list in /clusters/, emerging issues flagged
Weekly (Mon 08:00)	[[.jobs/weekly-insights-report]]	Insights Synthesizer	Support Insights Report in [[dashboard]]

How to run the demo

1. Open the [[dashboard]] — ticket volume trend, top issue clusters with counts and sentiment, emerging issues, CSAT, and deflection rate at a glance.
2. Browse /clusters/ — each file is a labelled ticket cluster with representative tickets, volume, and trend.
3. Open /reports/ — the weekly insights report the Synthesizer writes each Monday.

Connectors

Required: Zendesk, Intercom, or Freshdesk (ticket export via API). Recommended: Jira or Linear (link clusters to product backlog), Slack (#support-insights channel), CSAT/NPS survey tool (for satisfaction data overlay).

Demo data

Cluster	Volume (7d)	Trend	Sentiment
Billing — invoice confusion	142	↑ +34%	Negative
Onboarding — SSO setup	87	↑ +18%	Mixed
API — rate limit errors	64	↑ +52%	Negative
Export — CSV format issues	41	→ Stable	Neutral
Mobile — iOS login loop	38	↑ New	Negative

Every cluster is a file in /clusters/. Every insight in the dashboard links to its cluster file.