Checklist

The service desk AI readiness checklist

Before you switch on Freshservice Freddy or Zendesk AI, work through this. It's the same set of checks the audit runs automatically — written as a manual list so you can self-assess. Every item maps to one of the seven dimensions that decide whether your AI features actually deliver.

Prefer to skip the manual work? The free AI Readiness Audit scores every item below from your real data in under two minutes — or take the 2-minute quiz for a rough estimate with no login.

1. Knowledge base

AI suggestions and deflection bots can only repeat what your knowledge base already says.

  • Every high-volume ticket type has at least one current article.
  • Articles are published, not stuck in draft.
  • No obvious duplicates competing to answer the same question.
  • Content has been reviewed in the last 6–12 months.
  • Articles are structured (steps, headings) rather than walls of text.

2. Ticket data quality

Triage and summarisation learn from your historical tickets — garbage in, garbage out.

  • Subjects are descriptive, not 'Help' or a forwarded email subject.
  • Descriptions carry enough detail to classify the request.
  • Intake isn't dominated by a single generic catch-all mailbox.
  • Resolved tickets have meaningful closure information.

3. Categorisation

Auto-classification needs a category structure that's actually used and consistent.

  • A defined category / subcategory (or ticket form / field) structure exists.
  • Categories are applied consistently, not left blank or defaulted.
  • Tags (Zendesk) or fields (Freshservice) are hygienic, not sprawling.
  • The structure reflects how work actually arrives, not an old org chart.

4. Canned responses / macros

These seed AI reply suggestions and agent-assist.

  • Common replies are captured as canned responses (Freshservice) or macros (Zendesk).
  • Stale or duplicate responses have been retired.
  • Responses are shared where appropriate, not buried in personal lists.

5. SLA configuration

Urgency detection and escalation lean on your SLA and priority signals.

  • SLA policies exist and cover your real priorities.
  • There isn't a single catch-all policy masking urgency differences.
  • Priority is set meaningfully rather than always 'medium'.

6. Routing & automation

AI classifies; your routing decides where the classified ticket actually goes.

  • Groups and assignment rules cover the categories you triage.
  • Triggers / automations / workflows route without manual reassignment.
  • There's an escalation fallback for unrouted tickets.

7. Service catalogue / self-service

Deflection only works if there's somewhere to deflect people to.

  • A service catalogue (Freshservice) or request forms (Zendesk) exist for common requests.
  • Self-service is reachable from where users actually start.
  • The most-requested items are represented, not just a token few.

Or do all of this automatically.

The audit checks every item on this list against your real Freshservice or Zendesk configuration and gives you a score, the weakest dimensions, and the highest-impact fixes — free.

Run your free audit

FAQ

Do I need to do all of this before turning on AI?

No — but the weakest dimensions are what cap your results. The checklist is ordered by how much each area tends to affect AI performance. Fix the biggest gaps first.

Is there a faster way than going through this manually?

Yes. The free AI Readiness Audit measures every item on this list automatically from your real Freshservice or Zendesk data and scores you across all seven dimensions in under two minutes.

Zendesk is a trademark of Zendesk, Inc. Service Desk Builder is not affiliated with or endorsed by Zendesk.

Freshservice and Freddy are trademarks of Freshworks Inc. Service Desk Builder is not affiliated with or endorsed by Freshworks.