Purpose
A single source of truth for the terms used throughout these docs. Definitions are intentionally plain, one or two sentences each, so product, sales, Customer Success, and engineering share one vocabulary. Referenced terms link to the page where they’re explained in full.| Term | Meaning | Where used |
|---|---|---|
| Agent | A specialized, pre-built worker accountable for one job. Agents are the building blocks of an AI Employee. | Agent Library |
| Agent Inbox | The one queue where work that needs a person lands. It holds two item types: reports to view and schedules to accept. Upcoming. | Agent Inbox |
| Agent Library | The catalog of pre-built agents on Smaya. Upcoming in Exterview: an Agent Catalog in Settings that lists them. The Gradientflo agents are Upcoming. | Agent Library |
| Agent OS | A governed environment for building, deploying, supervising and improving a digital workforce. Smaya is Exterview’s. | Smaya |
| Agent Session | A live, real-time interaction run by an agent: phone, voice, avatar, panel, HR, assessment, or post-hire. | Deploy |
| Agent Studio | Where an administrator builds, deploys, monitors, evaluates and improves AI Employees. | Agent Studio |
| Agent swarm | One run per role, plus one sub-run per candidate, built from pre-built agents. Upcoming, including starting a role run. Screening through to a ranked Leaderboard is Available today. | Agent swarm |
| AI Employee | A named AI role on your team, staffed with agents. Available today: Role Designer, Screener, Coordinator and Report Specialist, plus Smaya. They run with your access rights and ask for approval before creating a role, publishing a job or contacting a candidate. | AI Employees |
| AI feature | A capability powered by Exterview’s models, such as screening, interviewing, matching, and scoring. Marked with an AI label in the guides. | Modules |
| AI tutor | A Gradientflo agent that teaches — carrying project, team and company knowledge to a new hire over voice or avatar, and answering from your own material with citations. | Gradientflo Agents |
| Application | A vertical workload built on Smaya, carrying the expertise of one business function. Exterview and Gradientflo are the first two. | Applications |
| Approval gate | The point at which an AI Employee stops and waits for a named person before an action takes effect. | Governance & Control |
| Assessment | A scored evaluation of a candidate’s skills or responses. | Session Types |
| Audit Trail | A traceable record of every action, decision, and result. | Auditability |
| Available | Usable today. Contrast with Upcoming. | Agent Library |
| Band | A seniority or responsibility level defined in Foundation. | Foundation |
| Blind screening | Hiding names, photos and personal details so the first review is based on capability alone. Upcoming. | Blind Screening |
| Bounded authority | The rule that an agent’s effective authority is the intersection of its own declared scope and the permissions of the person who invoked it, never the sum. | Governance & Control |
| Brand voice | The organization-level tone every agent uses when speaking or writing to candidates. | Voice and Policy Setup |
| Calibration | Tracing a score back to its rubric version and re-baselining under change control when rules change. | Calibration |
| Calibration window | The period after go-live during which scoring sharpens as sessions accumulate. | How to Go Live |
| Candidate Invite | The link or message that brings a candidate into a session: email, SMS, WhatsApp, or link. | Invite Flows |
| Candidate-first matching | A candidate uploads one resume and Exterview surfaces the best-fit open roles, each with a reason. Upcoming. | How to Find Jobs |
| Change Control | The reviewed, approved process for altering a frozen configuration; scoring changes re-baseline to a new version. | Configure |
| Charter | The mandate, scope of authority, decision authority, escalation path, operating window and three named owners every AI Employee carries. | AI Employees |
| Confidence-gated fallback | An AI tutor that is not confident enough to answer defers and logs the gap rather than guessing or sending the person to their manager. | Gradientflo Agents |
| Configuration | Your organization’s structure, rules, and governance, frozen into a signed baseline before go-live. | Configure |
| Configuration Contract | The signed, versioned, frozen record of all setup decisions that every evaluation score traces back to. | How to Freeze the Configuration Contract |
| Configuration Tier | How much you configure: Starter, Growth, or Enterprise (land-and-expand). | Configure |
| Content-based scoring | Scoring what a candidate says, not their face, tone, or biometrics. | Responsible AI |
| Core feature | A standard platform capability such as jobs, users, billing, and configuration. Marked with a Core label in the guides. | Modules |
| Data residency | The region where your organization’s data is stored, chosen during setup. | How to Set Compliance and Privacy |
| Decision Center | The area where you review candidates, scores and interview reports and record decisions. Formerly Reports. Available. | Monitor |
| Decision Record | The record behind a decision. Available today: the model and prompt behind every score, AI score overrides with a mandatory reason, and a stated reason for every decision change. Upcoming: Decision Room, one permanent record of who decided what, when and on what evidence. | Auditability |
| Deploy | Launching active agent sessions or document workflows. | Deploy |
| Digital Twin | Customer-specific customization of the AI interviewer persona. Upcoming. | How to Configure the AI Digital Twin |
| Document Workflow | A non-live task such as role creation or resume match. Offer generation is Upcoming. | Core Concepts |
| Enterprise | The tier with the full configuration set and per-organization hyper-calibration. | Configure |
| Escalation | Automatic routing of work outside an AI Employee’s mandate to its named human owner. | AI Employees |
| Evaluation | A scored assessment of a candidate against a role’srubric. | Core Concepts |
| Fairness Register | Cohort-level measurement of outcome rates by stage and group, with alerts and an exportable package. Upcoming. | Governance & Control |
| Fairness testing | Adverse-impact measurement across groups. Upcoming. It is a measurement aid, not a certification. | Responsible AI |
| Foundation Setup | The part of Configure that defines org structure and builds the hyper-calibration layer. | Foundation |
| Gate | A pass/fail checkpoint that controls whether a candidate advances (contrast with a signal). | Operations |
| Governance Setup | The part of Configure covering roles and permissions, frameworks, retention, human review, and audit. | Governance |
| Gradientflo | The post-hire application on Smaya: onboarding, assimilation, ramp, role fit, development and performance. Upcoming. | Gradientflo |
| Growth | The tier with light, department-based configuration and dynamic variables. | Configure |
| Human Review | A person reviews and approves outcomes; nothing adverse is decided autonomously. | Responsible AI |
| Hybrid workforce | An operating model in which a small human team supervises a much larger set of AI Employees, each with a named human owner. | The Hybrid Workforce |
| Hyper-calibration | The per-organization scoring layer built once in Foundation, taken through staging, then frozen. Enterprise-only. | Foundation |
| Interpretation vs calculation | The rule that AI may interpret evidence but never computes a score or applies a threshold; that arithmetic is deterministic and reproducible. | Governance & Control |
| MCP (Model Context Protocol) | The open standard through which approved assistants such as Claude, ChatGPT and Microsoft 365 Copilot reach Exterview. Available: more than 45 tools behind OAuth 2.1 with Microsoft sign-in, scoped to the person asking. No delete tools. | Microsoft & Interoperability |
| Microsoft Work IQ | Microsoft’s enterprise context layer. Smaya is designed to be grounded in it, and never requires it. | Microsoft & Interoperability |
| Module | What you enable and configure in Exterview. Underneath, the work is done by one or more agents. | Modules |
| Monitor | Reviewing completed sessions, reports, analytics, and outcomes. | Monitor |
| Operations Setup | The part of Configure covering channels, pipeline, gating, communication, and connectors. | Operations |
| Organization | Your isolated organization environment; all data and configuration are scoped to it. | Core Concepts |
| Outcome Intelligence | Post-hire signals at D30, D60, and D90 fed back to improve evaluation quality. Upcoming. | Outcome Intelligence |
| Pipeline Template | A reusable hiring-journey structure you apply to roles. | Operations |
| Policy guardrails | Customer-authored boundaries on agent behavior, bound per pipeline and role. Upcoming. | How to Configure Policy Guardrails |
| Portfolio | The set of AI Employees a single person is accountable for supervising. | The Hybrid Workforce |
| Prompt governance | Controlled, versioned governance over how agents are instructed, with auditability. Upcoming. | How to Govern Prompts |
| Quality & Fairness Agent | The agent that will independently check every other agent’s output before it enters the record. Visible to you, and not customer-configurable. Upcoming. | Governance & Control |
| Report / Takeaway | The structured output every completed session or workflow produces. | Session Reports |
| Report to view | An Agent Inbox item that shows what was found about a candidate. You answer Proceed, Drop or Talk. Drop is a human decision on record and never marks a candidate rejected. Upcoming. | Agent Inbox |
| Reproducible scoring | The same candidate and inputs always produce the same score, and nothing drifts. | Calibration |
| Role | A job definition. The structure a candidate is evaluated against. | Foundation |
| Role Fit | Re-evaluation of a hire at day 60 and 90 against the criteria they were originally hired on. Upcoming. | Gradientflo Agents |
| Rubric | The set of dimensions a candidate is scored against. | Operations |
| Schedule to accept | An Agent Inbox item for a slot that needs someone’s time. You answer Accept, Decline or Talk. Upcoming. | Agent Inbox |
| Scorecard | The structured set of scores a rubric produces for an evaluation. | Session Reports |
| Scoring weights | Stage, resume, and competency weightings that drive a score. Each block totals 100 percent. | How to Set Scoring Weights |
| Signal | An input that informs a decision without blocking progression (contrast with a gate). | Operations |
| Smaya | The enterprise Agent OS: the governed platform on which AI Employees are built, deployed, monitored and improved. Exterview is its first application. | Smaya |
| Smaya Assistant | The conversational surface of the Agent OS: asking questions and directing AI Employees. | Smaya Assistant |
| Starter | The entry tier: standard scoring, one team, no configuration set. | Configure |
| Supervision ratio | The operating model Exterview sells: one manager oversees a portfolio of AI Employees, so oversight scales with volume rather than headcount. | AI Employees |
| Talent Graph | Your organization’s private map of skills, evidence and outcomes, written and read by every application on the platform. | The Talent Graph |
| Upcoming | On the roadmap and not yet usable. Upcoming items carry no dates. Contrast with Available. | Agent Library |
| Workflow | See Document Workflow. | Core Concepts |
FAQs
A term I need isn't here.
A term I need isn't here.
Tell your Exterview contact and we’ll add it. This glossary is the shared vocabulary for product, sales, Customer Success, and engineering.
Do these definitions match the contract language?
Do these definitions match the contract language?
These are plain-language summaries for readers. Legal and configuration documents may use more precise wording.
Which terms are Enterprise-only?
Which terms are Enterprise-only?
Hyper-calibration and the full configuration set (Foundation, Operations, Governance) apply at Enterprise. Starter and Growth use standard and light configuration.
Related
- Core Concepts. The mental model these terms come from.
- Modules. The catalog of agents these terms describe.
- Enterprise Setup. Where tiers, foundation, and calibration are set.
- Overview. Start here if you’re new.