Troubleshooting guide

Freshservice Freddy AI not working — why and how to fix it

Freddy AI is enabled on thousands of Freshservice instances that see little to no ROI from it. The problem is almost never the feature — it is the configuration underneath it. Here are the six reasons Freddy underperforms, and what to do about each one.

Skip the manual diagnosis

Connect your Freshservice account and get a scored breakdown of all six areas in under 2 minutes.

Run free audit

1. Knowledge base is too thin or stale

Freddy chatbot and article suggestions draw directly from your KB. If article coverage is below 50% of your common ticket types, articles are more than 6 months old, or there is significant duplication, Freddy cannot surface useful answers.

Symptoms

  • Chatbot responds with 'I couldn't find an answer' most of the time
  • Suggested articles are irrelevant or outdated
  • Users abandon chatbot and log tickets anyway

Fix

Audit which ticket types lack KB articles. Write or update articles for your top 10 recurring request types. Set a 6-month freshness review cycle.

2. Ticket categories are flat or inconsistent

Freddy auto-triage needs at least three levels of categorisation (category → subcategory → item) to route accurately. Flat trees, inconsistent naming, and unused legacy categories all erode triage quality.

Symptoms

  • Auto-assigned tickets land in the wrong queue
  • Freddy routing accuracy is below 60%
  • Agents frequently reassign Freddy-routed tickets

Fix

Build a 3-level hierarchy. Retire categories unused in the last 90 days. Standardise naming: Service Area → Function → Item.

3. Canned responses have gaps or duplicates

Freddy AI surfaces canned responses as agent suggestions. If there is no response for a ticket type, Freddy suggests nothing. Duplicate responses with similar wording create conflicting suggestions that agents stop trusting.

Symptoms

  • Freddy agent suggestions are blank for common ticket types
  • Agents see multiple near-identical response suggestions
  • Canned response adoption rate is low

Fix

Map one canned response per top-20 ticket category. Deduplicate responses with more than 85% content overlap. Create a tone and format standard.

4. SLA policies are incomplete or catch-all

Freddy AI prioritisation relies on SLA policies to determine urgency. A single catch-all SLA policy means Freddy cannot distinguish priority by service context. Missing escalation rules on P1 and P2 tickets leave critical issues without automated action.

Symptoms

  • Freddy AI triage ignores service context when setting priority
  • P1 tickets are not auto-escalated
  • SLA breach rates are high despite Freddy being active

Fix

Create service-specific SLA policies aligned to your service catalogue. Add escalation rules on all P1 and P2 policies. Review breach thresholds against actual business impact.

5. Routing relies on manual assignment

Freddy auto-assignment improves over time by learning from your historical routing decisions. If most tickets are manually assigned rather than routed through rules, Freddy has poor training data and assignment accuracy stays low.

Symptoms

  • Freddy keeps suggesting the same agent regardless of ticket type
  • High proportion of tickets are manually reassigned after auto-assignment
  • Assignment accuracy does not improve over time

Fix

Configure automated routing rules in Freshservice before relying on Freddy assignment. Reduce manual overrides. Freddy learns from rule-based routing data more reliably than manual decisions.

6. Service catalogue is sparse or incomplete

Freddy AI self-service and request fulfilment features need items in the service catalogue to route against. Admins with fewer than 10 catalogue items effectively block AI-assisted request fulfilment for end users.

Symptoms

  • Chatbot cannot offer self-service for common requests
  • Users cannot find relevant service items through Freddy
  • Request fulfilment relies on free-text tickets rather than structured catalogue items

Fix

Build out your service catalogue to cover at least your top 15 recurring request types. Link each item to an SLA policy. Use Freddy service item suggestions to guide end users.

How to diagnose which problem applies to your instance

The six issues above apply to most Freshservice instances — but typically two or three are the dominant cause of underperformance. Diagnosing which ones apply to your specific instance requires pulling data from your Freshservice account and scoring it against the thresholds Freddy needs.

1

Connect your instance

Enter your Freshservice subdomain and API key. Read-only — nothing is written to your account.

2

We scan 6 dimensions

KB coverage, categories, canned responses, SLAs, routing, and service catalogue — all in under 2 minutes.

3

Get your scored report

An overall AI readiness score plus the top issue in each dimension. Free. No credit card.

Run your free AI readiness audit
Why is Freshservice Freddy AI not working even though it is enabled?+

Freddy AI depends on what is underneath it in your Freshservice instance. If your knowledge base has thin coverage, your ticket categories are flat or inconsistent, your canned responses have gaps, your SLA policies are incomplete, your routing relies on manual assignment, or your service catalogue is sparse — Freddy will underperform regardless of which plan tier you are on. Enabling the feature and getting value from it are two separate things.

Why does Freshservice Freddy chatbot not deflect tickets?+

Freddy chatbot deflects tickets by surfacing relevant knowledge base articles in response to user queries. If your knowledge base has fewer than 50 articles, article coverage does not match your most common ticket types, articles are out of date, or there is significant duplication, the chatbot will fail to find useful answers and deflection rates will be very low.

Why is Freddy AI auto-triage not routing tickets correctly?+

Freddy auto-triage learns from your ticket category structure. Flat category trees (no subcategories), inconsistent naming, legacy categories no longer in use, and a high percentage of tickets left at the top-level category all degrade routing accuracy. Freddy needs clean, consistent category data to route reliably.

How do I fix Freddy AI in Freshservice?+

Run an AI readiness audit to identify the specific gaps. The audit checks your knowledge base coverage, ticket category structure, canned response library, SLA configuration, routing rules, and service catalogue — and scores each one against the thresholds Freddy needs to perform. The free audit shows the top issue in each area; the full Readiness Blueprint gives all 18 issues with specific fix instructions.

Find exactly which issue is blocking Freddy AI

The free audit connects to your Freshservice account, scores all six dimensions, and shows you the top configuration issue in each area — in under 2 minutes.

Run your free audit

Free · Read-only · Under 2 minutes