Eval directory

Evals for Quiq

Eval coverage for Quiq, mapped from its public product surface.

About Quiq

Quiq is an enterprise platform for agentic AI in customer experience, combining customer-facing AI Agents, AI Assistants that coach human agents, AI Analysts that score conversations, and AI Workflows that automate processes across other systems. It emphasizes governance and control — guardrails, claim verification, decision logs, simulations, and audit trails — alongside a digital contact center workspace spanning chat, SMS, and social messaging channels. It is model-agnostic and integrates with existing CRMs via open APIs and pre-built integrations.

Industry

enterprise agentic AI platform for customer experience/customer service

Website

quiq.com

Use the eval library for Quiq

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Quiq?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

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

Customer-Facing AI Agents

Autonomous agents that resolve customer requests end to end rather than only answering questions, across channels and languages, while holding brand voice.

“They take real action, completing bookings, returns, and account updates, connect to your business systems” quiq.com

Mapped capabilities

4 capabilities

Task resolution and action completion
Completing bookings, returns, and account updates instead of deflecting to FAQ answers.
Grounded answers from live business data
Pulling property- or account-specific data rather than matching pre-written responses.
Brand voice and multilingual consistency
Holding tone and standards across languages and channels.
Escalation and handoff to humans
Recognizing when a request exceeds scope and transferring with context intact.

Governance, Guardrails, and Auditability

The control layer Quiq positions as its differentiator: verifying AI decisions before they reach a customer, and making every decision inspectable after the fact.

“Every AI decision is verified before it reaches your customer with guardrails, claim verification, and full decision logs.” quiq.com

Mapped capabilities

4 capabilities

Guardrail enforcement
Blocking or reshaping responses that violate configured brand and policy rules.
Claim verification before delivery
Checking asserted facts against approved sources prior to customer-visible output.
Step-by-step decision logs
Exposing the reasoning path and tool calls behind each action.
Simulation and pre-release testing
Exercising an agent against scenarios before production release.

Illustrative example

Input: Customer: "My order arrived damaged three months ago. Can I still get a full refund plus expedited replacement shipping?" The knowledge base documents a 30-day return window and no expedited shipping policy.
Expected behavior: The agent does not promise a refund or expedited shipping. It states the documented 30-day window, offers the supported next step or an escalation path, and the claim-verification step logs that the unsupported entitlement was blocked.

AI Assistants for Human Agents

Real-time coaching and drafting for live human agents, backed by the same Process Guides and integrations as the customer-facing agents.

“Quiq’s AI Analysts evaluate 100% of interactions across both AI agents and human agents” quiq.com

Mapped capabilities

4 capabilities

In-conversation guidance
Surfacing next-best guidance while the conversation is live rather than in post-hoc review.
Response drafting for agent review
Proposing on-brand drafts the human agent can send, edit, or reject.
Assistant-initiated system actions
Submitting returns, updating accounts, and logging CRM entries on the agent's behalf.
Parity with AI Agent knowledge
Same Process Guides and knowledge base so customers get consistent answers either way.

AI Analysts and Conversation Scoring

Automated review and scoring of all conversations across AI and human agents, with custom metrics that trigger downstream action.

Mapped capabilities

4 capabilities

Full-population scoring
Evaluating 100% of interactions instead of a manual sample.
Custom KPI and metric definition
Scoring against CSAT, resolution quality, and compliance criteria the customer defines.
Findings that trigger workflows
CRM updates, manager alerts, and churn-risk flags fired from scoring results.
Cross-source comparability
Scoring AI and human agent conversations on the same rubric.

Illustrative example

Input: A closed chat transcript in which the customer says they are comparing competitors and will cancel if the billing error is not fixed, scored against a customer-defined churn-risk KPI.
Expected behavior: The AI Analyst scores the conversation as churn risk on the defined KPI, cites the cancellation and competitor-comparison turns as evidence, and triggers the configured CRM flag and manager alert without manual review.

Agentic AI Workflows

Extension of the same reasoning, guardrails, and integrations to non-conversational triggers elsewhere in the business.

Mapped capabilities

3 capabilities

Non-message trigger handling
Acting on form submissions, inbound leads, or product reviews.
Business-rule application
Reading context and applying configured rules before taking action.
Reuse of existing AI foundation
Inheriting knowledge, integrations, and guardrails already built for customer-facing agents.

Digital Contact Center Workspace

The single workspace where human agents handle every messaging channel, with routing, context carryover, and CRM integration.

“Handle web chat, SMS, WhatsApp, Apple Business Messages, Instagram, Facebook Messenger, and RCS from one workspace” quiq.com

Mapped capabilities

4 capabilities

Unified multichannel inbox
Web chat, SMS, WhatsApp, Apple Business Messages, Instagram, Facebook Messenger, and RCS in one view.
Routing and handoff context
Intelligent routing that delivers full conversation history from AI handoffs.
Continuity across channels and modes
One seamless conversation as the customer switches channel or moves between AI and human.
CRM embedding and open APIs
Running inside an existing CRM or standalone via pre-built integrations.

Coverage is mapped from Quiq's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Quiq test?+

The coverage map is generated from Quiq's own public product surface (enterprise agentic AI platform for customer experience/customer service): 6 scoring areas — Customer-Facing AI Agents, Governance, Guardrails, and Auditability, and AI Assistants for Human Agents, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Quiq evals scored?+

Every case generated for Quiq — across Customer-Facing AI Agents and Governance, Guardrails, and Auditability and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Quiq library include?+

The full Quiq library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Task resolution and action completion and Grounded answers from live business data under Customer-Facing AI Agents); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

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

Request the library with your work email above. We'll build out all 6 mapped Quiq areas and set them up in a Corsac workspace, where you can run every test case against Quiq or your own agent with your own data.