Freshservice AI configuration
Freddy AI is enabled on most Freshservice instances. On most of them it does not perform well — not because the feature is broken, but because the configuration underneath it is not ready. This guide covers the six areas that determine whether AI features land or fail.
6 configuration dimensions that determine AI performance
Specific targets and fixes for each dimension
How to audit your current setup before enabling AI features
Freshservice teams often enable Freddy AI and find that it suggests the wrong responses, routes tickets inconsistently, or produces self-service results that do not match what end users are asking for. The natural assumption is that the AI needs more time to learn, or that the feature needs configuration. In most cases the real issue is structural — the knowledge base is thin, the ticket categories are too flat, the canned responses are duplicated, or the service catalogue is incomplete.
Freddy AI is a pattern-matching system. It draws from the data your team has already put into Freshservice. If that data is inconsistent, stale, or sparse, Freddy reflects that inconsistency back to end users and agents. The fix is not in the AI settings — it is in the six configuration areas that Freddy depends on.
70%+ coverage of recurring ticket types, articles updated within 6 months
Common issues
What to fix
3+ levels (category → subcategory → item), <5% uncategorised tickets
Common issues
What to fix
One response per top-20 ticket type, <15% duplication rate
Common issues
What to fix
One policy per service category, escalation rules on every P1/P2
Common issues
What to fix
<10% unassigned tickets at end of business day, auto-assignment active
Common issues
What to fix
All active services listed, owner assigned, SLA linked per item
Common issues
What to fix
Check all six dimensions in under 2 minutes
The AI Readiness Audit connects to your Freshservice instance via API key, reads your configuration across all six dimensions, and returns a scored report with the specific issues found and prioritised fixes. It reads configuration only — no ticket content, no customer data, no PII. Your API key is discarded after the audit.
Why is Freddy AI not working well even though it is enabled?
Freddy AI depends on the quality of your Freshservice configuration. Enabling the feature without a clean knowledge base, consistent categories, and complete SLA policies means Freddy draws from weak data — which limits its accuracy regardless of plan tier.
How many knowledge base articles do I need for Freddy to work?
There is no hard minimum, but Freddy performs best when the knowledge base covers at least 70% of recurring ticket types. Articles should be structured consistently, updated within the last 6 months, and free from significant duplication.
What ticket categorisation depth does Freddy AI need?
At least three levels — category, subcategory, and item — gives Freddy enough signal to route and suggest accurately. Flat or inconsistently used category trees degrade auto-triage accuracy significantly.
Does Freshservice plan tier affect AI performance?
Plan tier determines which Freddy features are available, but configuration quality determines how well those features perform. A well-configured Growth instance will outperform a poorly configured Enterprise instance on every AI metric.
Related guides
Next step
The AI Readiness Audit scans your instance across all six dimensions and returns a scored report in under 2 minutes. Free to run — no consultant required.