Key Concepts & Glossary

A comprehensive reference for every term and concept used across the Operanix platform. Bookmark this page and refer back whenever you encounter unfamiliar terminology.

Glossary

Agent

A specialized AI assistant configured for a specific business domain (e.g., Sales, Support, HR). Each agent has its own identity, knowledge bindings, system prompt, and deployment settings. Agents are the core unit of work in Operanix.

Agent Builder

The visual interface for creating and customizing agents. The Agent Builder lets you configure an agent’s name, persona, system prompt, knowledge assignments, autonomy level, and deployment channels without writing code.

Agent Certification

A formal readiness status assigned to an agent after it passes evaluation. Certification levels range from Needs Training (not ready) to Production Ready (fully certified for customer-facing use). Certification is required before production deployment.

Agent Marketplace

A library of pre-built, one-click installable agents and agent bundles. The marketplace lets you skip configuration and instantly deploy agents that have been optimized for common use cases like customer support, sales outreach, or HR onboarding.

Agent Readiness

The section of the platform dedicated to evaluating whether an agent is ready for deployment. Agent Readiness runs automated test suites that score agents across multiple quality dimensions and produces a pass/fail verdict.

Agent Workforce

The collective set of AI agents deployed within your workspace. The Agent Workforce view shows all installed agents, their deployment status, trust scores, and recent activity — giving operators a single dashboard for their entire AI team.

Approval Queue

A governance feature that holds agent actions for human review before they are delivered to end users. Actions are routed to the approval queue when they fall below the auto-approve trust threshold or match a policy rule requiring manual review.

Audit Trail

An immutable, chronological log of every action taken on the platform — including agent responses, governance decisions, configuration changes, and user activity. The audit trail supports compliance requirements and enables root-cause analysis.

Autonomy Level

A configuration setting that determines how independently an agent can act. Levels range from Full Approval (every response requires human review) to Autonomous (responses are delivered automatically without review). Autonomy is scoped per agent and can be adjusted at any time.

Bundle

A pre-packaged collection of agents from the Marketplace designed to work together for a specific use case. For example, the “Customer Experience Bundle” includes Sales, Support, and Marketing agents pre-configured for customer-facing scenarios.

Certification Level

The formal quality tier of an agent based on its evaluation results. The four levels are: Needs Training (below 50%), In Development (50–69%), Certified (70–89%), and Production Ready (90%+).

Chat Widget

An embeddable UI component that can be added to any website to provide a customer-facing chat interface powered by an Operanix agent. The widget is customizable with your brand colors, logo, and welcome message.

Citation Coverage

The percentage of claims in an agent’s response that are backed by references to specific knowledge base sources. High citation coverage indicates that the agent is grounding its answers in verified information rather than generating content from its general training data.

Command Center

The primary dashboard in Operanix that provides an executive overview of your AI workforce. The Command Center displays key metrics, active missions, agent health scores, recent activity, and alerts — serving as the operational hub for platform operators.

Compliance

The system of rules, policies, and checks that ensure agent behavior meets regulatory and organizational standards. Operanix supports compliance frameworks including SOC 2, GDPR, HIPAA, and custom enterprise policies.

Compliance Gate

An automated checkpoint in the agent response pipeline that evaluates whether a response meets compliance requirements before delivery. Responses that fail the compliance gate are blocked and flagged for review.

Continuous Evaluation

A scheduled, recurring process that automatically re-evaluates agent quality over time. Continuous evaluation detects quality drift, knowledge staleness, and performance degradation — alerting operators before issues impact end users.

Cost Intelligence

The analytics module that tracks the cost of every AI interaction, broken down by agent, model, department, and time period. Cost Intelligence helps operators optimize LLM spending and calculate ROI for their AI workforce.

Crawl

The process of automatically scanning a website to extract and ingest its content into the Operanix knowledge base. Crawling follows links, extracts text, and organizes content into structured knowledge entities.

Department

An organizational unit within a workspace that groups agents, team members, and knowledge for scoped access control. Departments allow enterprises to isolate AI resources by business function (e.g., Sales, Engineering, HR).

Deployment

The process of making an agent available to end users through one or more channels. Deployment progresses through environments (Staging, Production) and can be rolled back if issues are detected.

Deployment Channel

A method through which end users interact with a deployed agent. Operanix supports five channels: Chat Widget (embedded on websites), Hosted Chat (standalone page), API (programmatic access), Slack, and WhatsApp.

Evaluation

An automated quality assessment that tests an agent against a suite of questions and scores its responses across multiple dimensions. Evaluations produce a pass rate, individual dimension scores, and a certification-level verdict.

Evaluation Case

A single test question within an evaluation suite. Each case includes a question, expected answer (optional), and scoring criteria. Cases are automatically generated from the agent’s knowledge base or manually authored by operators.

Golden Set

A curated collection of high-quality evaluation cases that serve as the benchmark for agent quality. The golden set typically contains the most critical questions for a given domain and is used for certification-level evaluations.

Governance

The comprehensive system of policies, approvals, auditing, and controls that ensures AI agents operate safely and within organizational boundaries. Governance encompasses trust scoring, compliance gates, approval workflows, content policies, and audit trails.

Groundedness

An evaluation dimension that measures whether an agent’s response is factually supported by its assigned knowledge base. A highly grounded response uses information from verified sources rather than hallucinating or inventing facts.

Guardrails

Safety mechanisms built into the agent pipeline that prevent harmful, inappropriate, or off-brand responses. Guardrails include content filters, topic boundaries, PII detection, and custom safety rules defined by administrators.

Hosted Chat

A standalone, Operanix-hosted web page that provides a chat interface for a specific agent. Each deployed agent gets a unique hosted chat URL that can be shared with anyone — no website embedding required.

Knowledge Base

The organized collection of information that agents use to answer questions. A knowledge base consists of crawled web content, uploaded documents, and manually entered entities. Knowledge is versioned, searchable, and assignable to specific agents.

Knowledge Coverage

A metric that measures what percentage of expected topics are represented in an agent’s assigned knowledge base. Low knowledge coverage indicates gaps that may lead to poor responses or “I don’t know” answers.

Knowledge Operations

The platform section dedicated to managing the knowledge lifecycle — from ingestion (crawling, uploading) through processing (chunking, embedding) to assignment (binding knowledge to agents). Also called “Knowledge Ops.”

Lead Capture

A feature of the chat widget that collects visitor information (name, email, company) before or during a conversation. Captured leads are stored in the platform and can be exported or pushed to CRM integrations.

Marketplace

See Agent Marketplace.

MFA (Multi-Factor Authentication)

A security feature that requires users to verify their identity with two or more factors (e.g., password + authenticator app) when signing in. MFA can be enforced organization-wide by administrators.

Pass Rate

The percentage of evaluation cases that an agent answers correctly, as determined by automated scoring. Pass rate is the primary metric for agent readiness and certification. A pass rate of 90%+ qualifies an agent as Production Ready.

Pipeline

An automated, multi-stage process that runs on a schedule. The primary pipeline in Operanix is the Knowledge Pipeline, which processes raw content through ingestion, extraction, embedding, evaluation, and publication stages.

Production Ready

The highest certification level for an agent, indicating it has passed evaluation with a 90%+ score and is safe for customer-facing deployment. Production Ready agents have verified knowledge, high groundedness, and passing governance checks.

Prompt

The instruction text sent to an AI model to generate a response. In Operanix, each agent has a system prompt that defines its personality, rules, and behavior, plus per-conversation prompts assembled from user input and retrieved knowledge.

RAG (Retrieval-Augmented Generation)

An AI architecture that enhances model responses by first retrieving relevant information from a knowledge base, then using that information as context for the model’s answer. Operanix uses a two-layer RAG system combining structured entities with vector-searched document chunks.

RBAC (Role-Based Access Control)

A security model that restricts platform access based on assigned roles. Operanix supports 10 roles including Owner, Admin, Manager, Operator, Agent Builder, Knowledge Manager, Compliance Officer, Analyst, Member, and Viewer — each with specific permission sets.

Readiness Score

A composite score (0–100) that represents how ready an agent is for deployment. The readiness score combines evaluation pass rate, knowledge coverage, governance compliance, and configuration completeness into a single metric.

ROI (Return on Investment)

The business value generated by your AI workforce relative to its cost. Operanix calculates ROI by tracking conversations handled, estimated time saved, cost per interaction, and customer satisfaction improvements.

Rollback

The ability to revert an agent’s deployment, knowledge, or configuration to a previous version. Rollback is available for deployments (restore a previous live version), knowledge (revert to an earlier snapshot), and agent settings.

SSO (Single Sign-On)

An authentication method that allows users to sign in to Operanix using their existing corporate identity provider (e.g., Google Workspace, Microsoft Entra ID, Okta). SSO eliminates the need for separate Operanix passwords and enables centralized access management.

Tenant

An isolated organizational unit within the Operanix multi-tenant architecture. Each tenant (typically one per company) has completely separated data, configuration, agents, and knowledge. Cross-tenant data access is impossible by design.

Trust Score

A composite metric (0.00–1.00) that quantifies the reliability of an agent or a specific agent action. The trust score is calculated from five weighted dimensions: Agent Quality (30%), Citation Score (20%), Compliance Score (20%), Governance Score (15%), and Knowledge Score (15%).

Version

A point-in-time snapshot of an agent’s configuration, knowledge bindings, or deployment state. Operanix tracks versions for agents, knowledge bases, and deployments, enabling comparison, auditing, and rollback.

Widget

See Chat Widget.

Workflow

An automated sequence of steps that executes when triggered by a specific event or condition. Workflows can chain agent actions, governance checks, notifications, and integrations into repeatable processes — for example, automatically routing a support ticket through triage, response generation, review, and delivery.