> ## Documentation Index
> Fetch the complete documentation index at: https://docs.exterview.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Employees

> AI Employees are named roles your team works with: Role Designer, Screener, Coordinator, Report Specialist and Smaya. Each does one kind of work, hands off to the others, and asks before it writes.

## Purpose

An **AI Employee** is a named role your team works with day to day. It is not a chatbot, not a workflow and not a raw agent. Each one does one kind of work, within your permissions, and brings anything consequential to you before it happens.

<Info>
  **Agents are the building blocks. AI Employees are the workforce.** Agents are
  pre-built by Exterview. AI Employees are the roles that put them to work for
  your organization.
</Info>

## The AI Employees in Exterview

**Available.** Five AI Employees work in Exterview today. You find them in **Settings → Setup → Operations → AI Employees**, with what each one does and how many tools it can use.

| AI Employee           | What it does                                                                                         | Writes?                      |
| --------------------- | ---------------------------------------------------------------------------------------------------- | ---------------------------- |
| **Role Designer**     | Drafts a role, its questions and its scoring, and publishes the job once a person approves it        | Yes, after your approval     |
| **Screener**          | Reads a job's applicants against its criteria and invites the ones worth screening                   | Yes, after your approval     |
| **Coordinator**       | Schedules, reschedules and cancels panel rounds, and refuses a real clash rather than double-booking | Yes, within your permissions |
| **Report Specialist** | Answers questions about any candidate report in your workspace, with citations to the evidence       | No                           |
| **Smaya**             | Answers anything, reading across jobs, candidates, interviews, reports and rounds                    | No                           |

You can ask any of them from the [Smaya Assistant](/smaya/assistant) or from a connected assistant over MCP. Upcoming: asking them from Slack.

## Handoffs between AI Employees

**Available.** You do not have to pick the right AI Employee. Ask one, and it hands the work to the colleague whose job it is. The reply names the AI Employee that did the work.

A handoff runs as you, with your permissions. Handing work to another AI Employee never widens what can be read or changed.

## Approval before writes

**Available.** Creating a role, publishing a job and contacting a candidate always show you an approval card first. Nothing happens until you approve it. Report Specialist and Smaya write nothing at all.

<Note>
  **Hiring outcomes stay with people.** No AI Employee sets a candidate to
  hired, rejected or withdrawn. It prepares the evidence and a recommendation. A
  named person decides.
</Note>

```mermaid theme={null}
%%{init: {'theme':'base','themeVariables':{'primaryColor':'#EEF2FF','primaryBorderColor':'#4B6CFE','primaryTextColor':'#0F172A','lineColor':'#64748B','clusterBkg':'#F8FAFC','clusterBorder':'#E2E8F0'}}}%%
flowchart TD
  ASK([You ask any AI Employee]) --> HAND([Handed to the colleague whose job it is])
  HAND --> RD([Role Designer drafts the role and pipeline])
  HAND --> SC([Screener picks who to screen])
  HAND --> CO([Coordinator proposes a panel slot])
  HAND --> RS([Report Specialist answers with citations])
  RD --> CARD([Approval card shown to you])
  SC --> CARD
  CO --> CARD
  CARD -->|Approve| DONE([Job published, invite sent or round booked])
  CARD -->|Decline| NONE([Nothing changes])
  RS --> READ([Read-only answer])
  DONE --> LOG([Recorded with the employee that did it])
```

*A handoff runs as you, with your permissions. Report Specialist and Smaya never write.*

## The agent's role for each job

**Available.** When you create a job, the **Agent Role** tab describes the agent's role for that job and the first message candidates receive. It is the mandate the AI Employees work to for that job, and you edit it like any other part of the job.

## The ceiling that never moves

Three rules hold for every AI Employee, in every configuration.

<CardGroup cols={3}>
  <Card title="Never exceeds its human" icon="shield">
    An AI Employee can never reach data or take an action the person invoking it could not. Authority is the intersection of the role's scope and yours, never the sum.
  </Card>

  <Card title="Never decides alone" icon="scale">
    Outcomes that materially affect a person — hiring, rejection, an offer, a
    performance judgment — are always made by a named human.
  </Card>

  <Card title="Never acts unrecorded" icon="clipboard-list">
    Every action and every approval is captured and attributed.
  </Card>
</CardGroup>

## Where AI Employees are going

**Upcoming.** Composing your own AI Employees from the [Agent Library](/smaya/capabilities), each with a full charter:

<CardGroup cols={2}>
  <Card title="Mandate and operating window" icon="target">
    What the role exists to achieve, the standing goals it is accountable for, and when it is active.
  </Card>

  <Card title="Scope of authority" icon="lock">
    An explicit allow-list of what it may read and write, bounded by module and
    data boundary.
  </Card>

  <Card title="Three named owners" icon="users">
    A business owner for the outcome, a technical owner for how it runs, and a
    data steward for what it may see.
  </Card>

  <Card title="A lifecycle, not on and off" icon="git-branch">
    Proposed, approved, active, suspended and retired, with the person who drafts a charter never the person who approves it. Suspending one role in one action leaves the others running.
  </Card>
</CardGroup>

AI Employees that work on a standing goal, without anyone starting the work, come with the [agent swarm](/smaya/agent-swarm) and are also Upcoming.

## What this changes commercially

Exterview does not sell headcount subtraction. It sells a **supervision ratio**: one person oversees a portfolio of AI Employees, each bringing anything outside its remit to the named person responsible. See [The Hybrid Workforce](/smaya/hybrid-workforce).

## FAQs

<AccordionGroup>
  <Accordion title="Is an AI Employee the same as an agent?">
    No. An agent is a specialized capability, pre-built and reusable. An AI Employee is a named role that puts agents to work. Your teams interact with AI Employees.
  </Accordion>

  <Accordion title="Can an AI Employee act without a person approving?">
    It reads and answers freely within your permissions. Creating a role,
    publishing a job and contacting a candidate always wait for your approval, and
    nothing adverse is ever decided by an AI Employee.
  </Accordion>

  <Accordion title="Can we create our own AI Employees?">
    Not yet. Five AI Employees are Available today. Composing your own from the
    Agent Library is Upcoming.
  </Accordion>

  <Accordion title="Do AI Employees work across applications?">
    Today they work in Exterview. As Gradientflo arrives, post-hire agents become available to AI Employees alongside talent agents.
  </Accordion>
</AccordionGroup>

## Related

<div className="kb-links">
  *   <span className="kb-ico">
        <Icon icon="network" />
      </span>
    [The Agent Swarm](/smaya/agent-swarm). Upcoming: how AI Employees will run a
    role together.
  *   <span className="kb-ico">
        <Icon icon="inbox" />
      </span>
    [Agent Inbox](/smaya/agent-inbox). Upcoming: where their reports will reach
    you.
  *   <span className="kb-ico">
        <Icon icon="shield-check" />
      </span>
    [Governance & Control](/smaya/oversight). The control model in full.
</div>