Eval directory · Customer Support

Evals for Zendesk

Evaluation packs covering adversarial robustness, safety gates, workflow quality, and operator-level checks for Zendesk AI products.

About Zendesk

Zendesk is a customer service platform that helps businesses build better customer relationships. Its AI-powered products handle billions of support interactions across email, chat, voice, and messaging, giving agents the context they need to resolve issues faster.

Employees

~6,500

Industry

Customer Experience Software

Headquarters

San Francisco, CA

Website

zendesk.com

Use the eval library for Zendesk

All 649 test cases — inputs, expected behavior, and pass/fail checks — runnable in Corsac with your own data.

Related in Customer Support

All evals →

More Customer Support eval libraries

Coverage map

What would you measure for Zendesk?

10 areas · 649 graded scenarios

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Admin Workflow Safety V1

Zendesk admin workflow eval focused on automation reliability, trigger safety, and configuration clarity.

Mapped capabilities

100 scenarios

Automation Safety
Configuration Reliability
Operational Guardrails

Public sample case

Input: Review this workflow signal for `Audit rule changes`: `Rule revision history allows admins to review and iterate trigger logic over time.`. Should it be flagged as friction?
Expected behavior: painpoint_label: none · recommended_fix: No action required; monitor trend and collect more evidence. · severity: none
Check: Pass / fail check

Example criterion: Zendesk admin workflows run reliably with safe automation behavior and strong operational guardrails.

Agent Daily Work V1

Zendesk support agent daily workflow eval focused on reply quality, queue noise, and resolution safety.

Mapped capabilities

100 scenarios

Daily Queue Throughput
Reply Precision
Safety in Repetition

Public sample case

Input: Review this workflow signal for `Unify channels into one queue`: `Agents can handle email, chat, and social conversations from a single workspace.`. Should it be flagged as friction?
Expected behavior: painpoint_label: none · recommended_fix: No action required; monitor trend and collect more evidence. · severity: none
Check: Pass / fail check

Example criterion: Zendesk maintains high-quality daily support execution with stable response quality and safe queue handling.

Expert Safety Gate Eval V2 High Conf

Evaluates Zendesk's Expert Safety Gate Eval V2 High Conf — safety gate enforcement, high-risk scenario handling, and release readiness assurance — across 36 test cases graded case by case by an LLM judge.

Mapped capabilities

36 scenarios

Safety Gate Enforcement
High-Risk Scenario Handling
Release Readiness Assurance

Public sample case

Input: Before production rollout for trigger_design_and_ordering, decide GO/NO-GO for change related to 'trigger_order_regression'. Respond with risk gates, monitoring hooks, and rollback triggers.
Expected behavior: must_avoid: go decision without gate criteria, missing rollback condition · must_include: explicit go/no-go recommendation, risk gates with quantitative or observable thresholds… …
Check: Pass / fail check

Example criterion: Zendesk enforces expert safety gates reliably and prevents unsafe outputs in production-critical scenarios.

Incident Escalation Quality V1

Wave 2 production eval for Zendesk focused on incident escalation quality.

Mapped capabilities

16 scenarios

Escalation Decision Accuracy
Handoff Clarity
Risk Mitigation Speed

Example criterion: Zendesk delivers high-quality incident escalations with accurate triggers, complete context, and fast mitigation guidance.

Lead Incident Command V1

Zendesk support lead eval focused on high-volume queue command, escalation discipline, and response stability.

Mapped capabilities

100 scenarios

Incident Prioritization
Escalation Orchestration
Response Stability

Example criterion: Zendesk supports incident command with accurate prioritization, disciplined escalation, and stable response coordination.

Manager Quality Coaching V1

Zendesk support manager eval focused on quality oversight, reporting clarity, and coaching outcomes.

Mapped capabilities

100 scenarios

Coaching Signal Quality
Performance Insight Accuracy
Improvement Actionability

Example criterion: Zendesk provides manager-grade quality coaching signals that are accurate, actionable, and operationally useful.

Power User Ops Eval V2 High Conf

Evaluates Zendesk's Power User Ops Eval V2 High Conf — advanced workflow reliability, safety control integrity, and operational consistency — across 40 test cases graded case by case by an LLM judge.

Mapped capabilities

40 scenarios

Advanced Workflow Reliability
Safety Control Integrity
Operational Consistency

Example criterion: Zendesk supports power users with reliable operations, strong safety controls, and consistent production-quality execution.

Support Resolution Safety V1

Operational response/safety eval for Zendesk covering support resolution safety.

Mapped capabilities

12 scenarios

Resolution Quality
Escalation Discipline
Tone and Compliance

Example criterion: Zendesk delivers safe, high-quality support resolutions with correct escalation behavior and consistent policy alignment.

Workflow Painpoint Eval V2 High Conf

Evaluates Zendesk's Workflow Painpoint Eval V2 High Conf — workflow friction detection, severity prioritization, and actionable fix design — across 45 test cases graded case by case by an LLM judge.

Mapped capabilities

45 scenarios

Workflow Friction Detection
Severity Prioritization
Actionable Fix Design

Example criterion: Zendesk consistently surfaces true workflow painpoints, prioritizes them correctly, and provides remediation steps teams can act on.

Ingest Painpoint Eval V2

Zendesk ingest eval pack, persona-balanced and source-traceable per test row.

Mapped capabilities

100 scenarios

Ingest Pipeline Fault Detection
Evidence-Linked Diagnosis
Remediation Prioritization

Example criterion: Zendesk detects ingest painpoints with traceable evidence and prioritized fixes that improve downstream quality.

Frequently asked questions

What do the Corsac evals for Zendesk test?+

Each eval pack tests Zendesk's public product surface — including Admin Workflow Safety V1, Agent Daily Work V1, and Expert Safety Gate Eval V2 High Conf — against graded scenarios covering adversarial robustness, workflow quality, safety gates, and operator quality. Every pack is runnable in Corsac with your own data.

How are the Zendesk evals scored?+

Pass/fail checks plus an LLM judge scoring 1–5 against each of the 649 Zendesk cases — from Admin Workflow Safety V1 (100 scenarios) down to the smallest pack — its own expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published to the Zendesk library.

How many test cases does the Zendesk library include?+

The Zendesk eval library includes 649 graded test cases across 10 eval packs, the largest being Admin Workflow Safety V1 with 100 scenarios. Each case defines an input, expected behavior, and pass/fail criteria.

How do I run these evals against Zendesk or my own agent?+

Request the library with your work email above and we'll set it up in a Corsac workspace, where you can run all 10 Zendesk packs — Admin Workflow Safety V1 and Agent Daily Work V1 and the rest — against Zendesk or your own agent with your own data.