Skip to main content
Apragya AI
Docs · Enterprise Apps

Support & Helpdesk overview: tickets, SLA, knowledge base

The customer-support slice - inbound ticket triage, SLA tracking, KB-backed answers, and the AI agents that triage + draft responses.

Support & Helpdesk handles the inbound side of the customer relationship: tickets in, answers out. Multi-channel intake (email, in-app, web form, Slack, WhatsApp), routing rules, SLA policies, knowledge-base-backed answers. Agents triage, classify, and draft responses; humans handle the cases that need judgement.

The ticket flow

  • Intake - email-to-ticket, web form, in-app help drawer, Slack-Connect channel, WhatsApp Business
  • Triage - classifier assigns category + priority + initial route
  • Route - to a team / queue / individual based on the routing-rules matrix
  • Respond - first reply within SLA (or escalation fires)
  • Resolve - mark resolved, customer satisfaction survey, post-resolution analytics

Bundled AI agents

  • Ticket Classifier - assigns category + priority on inbound; fires on ticket.created
  • Auto-Responder - drafts the first reply using KB-backed answers; agent posts as a draft for human approval, then sends
  • KB Suggester - while a human is composing a reply, surfaces relevant KB articles in the side panel
  • Sentiment Watcher - flags negative-sentiment trends in a customer's ticket history; escalates to a CSM

Configure on day one

  • SLA policies - per priority + per customer tier + per category (first reply, next reply, resolution)
  • Routing rules - the matrix that decides which queue a ticket lands in
  • Knowledge Base - the corpus the AI draws on (see the RAG and Knowledge Bases guide)
  • Macros and Templates - tenant-authored canned responses
  • Customer satisfaction (CSAT) survey - global default + per-resolution category overrides
SLA breaches write to support_sla_breaches + fire the sla.breached event. Bound an escalation agent on that event and you'll never silently miss an SLA again - the agent pings the responsible manager + reassigns + drafts an apology to the customer all in one shot.

Common workflows

Email lands in support@ -> Classifier categorises + sets priority (sub-second) -> KB Suggester finds three matching articles -> Auto-Responder drafts a reply pulling from the top article -> human reviews + sends -> ticket marked resolved -> CSAT survey fires 24h later. Average first-reply time on tenants using the bundled agents: ~12 minutes.

Need help building this? Talk to a real engineer

← All docs
Ask Vippy anything