> ## 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.

# The Hybrid Workforce

> How people and AI Employees work together: one person supervising a portfolio of AI Employees, how the working relationship is structured, and how a function is run by a small human team with a large digital one.

## Purpose

The organizational change Smaya enables is not "AI helps with tasks." It is a **change in what one person can be accountable for.**

A recruiter today runs a pipeline. A recruiter on Smaya **supervises a team of [AI Employees](/smaya/ai-employees) that runs the pipeline** — and is accountable for their judgment rather than for their keystrokes. The work does not disappear. The unit of the job moves up a level, from doing to directing.

<Info>
  **The metric that matters is the supervision ratio, not headcount.** Exterview
  does not sell headcount subtraction. It sells how much a single accountable
  person can competently oversee.
</Info>

## What the ratio looks like

One person can supervise a portfolio of AI Employees. How large depends entirely on how much judgment the work carries and how mature the configuration is.

| Stage           | Typical ratio per person | What the person is doing                                                     |
| --------------- | ------------------------ | ---------------------------------------------------------------------------- |
| **Starting**    | 1–5 AI Employees         | Reviewing most output, calibrating criteria, building confidence             |
| **Established** | 10–20 AI Employees       | Reviewing exceptions and escalations; routine output is spot-checked         |
| **Mature**      | 20–50+ AI Employees      | Owning the criteria and the escalation queue; reviewing only what is flagged |

The ratio rises as calibration proves itself — not as a license tier. An organization moves along it by demonstrating that its criteria produce judgments its people agree with, which is measurable and visible in the record.

## How the working relationship is structured

```mermaid theme={null}
%%{init: {'theme':'base','themeVariables':{'primaryColor':'#EEF2FF','primaryBorderColor':'#4B6CFE','primaryTextColor':'#0F172A','lineColor':'#64748B','clusterBkg':'#F8FAFC','clusterBorder':'#E2E8F0'}}}%%
flowchart TD
  H([Function lead — accountable for the outcome])
  M1([Manager · portfolio of AI Employees])
  M2([Manager · portfolio of AI Employees])
  A1([AI Employee])
  A2([AI Employee])
  A3([AI Employee])
  A4([AI Employee])
  H --> M1
  H --> M2
  M1 --> A1
  M1 --> A2
  M2 --> A3
  M2 --> A4
```

*A reporting line: each AI Employee has one named human owner. Upcoming: a consolidated portfolio view for each manager.*

It is a reporting line, deliberately. Each AI Employee has **one named human owner**, the way a person has one manager.

<CardGroup cols={2}>
  <Card title="You set the mandate" icon="target">
    What the role is for, and the outcomes it is accountable for. Written in a sentence you would sign off on for a person.
  </Card>

  <Card title="It works, and reports" icon="activity">
    It does the work and answers when you ask, with citations. Upcoming: reports
    brought to you in the [Agent Inbox](/smaya/agent-inbox), so you do not chase
    status updates.
  </Card>

  <Card title="It escalates rather than guesses" icon="bell">
    Anything outside its mandate comes to you automatically. An AI Employee that
    is unsure raises it; it does not improvise.
  </Card>

  <Card title="You approve what matters" icon="user-check">
    Creating a role, publishing a job and contacting a candidate wait for your
    approval. Routine reading and answering proceed.
  </Card>

  <Card title="You review and correct" icon="pencil">
    Your corrections are recorded, and a pattern of corrections is the signal that
    criteria need recalibrating.
  </Card>

  <Card title="You can suspend instantly" icon="power">
    **Upcoming.** One action pauses or revokes a role, exactly like removing a person's access, without touching the others.
  </Card>
</CardGroup>

## A day in the hybrid model

<Steps>
  <Step title="Morning: read the Agent Inbox, not your email">
    You open the [Agent Inbox](/smaya/agent-inbox) and see the reports to view and the schedules to accept. What is waiting on you is the agenda. The Agent Inbox is Upcoming; today you start from the Decision Center and the Leaderboard.
  </Step>

  <Step title="Answer the reports">
    Each report carries the evidence and a recommendation. You answer Proceed,
    Drop or Talk. Your reasoning is captured.
  </Step>

  <Step title="Accept the schedules">
    Slots that need your team's time wait for you to accept or decline. Actions
    you asked an AI Employee to take wait on an approval card.
  </Step>

  <Step title="Spend the rest of the day on judgment">
    Candidate conversations, hiring-manager alignment, offer strategy — the work
    that was always the job and never had enough hours in it.
  </Step>

  <Step title="Periodically: recalibrate">
    Where you keep overriding in the same direction, the criteria are wrong. You change them once, under approval, and the whole portfolio improves.
  </Step>
</Steps>

## Running a whole function this way

At maturity, a function is run by a **small human team with a large digital one**. A talent organization that ran on twelve recruiters might run on four people supervising sixty AI Employees — with the four doing the stakeholder, judgment and relationship work that was previously squeezed between coordination tasks.

<CardGroup cols={2}>
  <Card title="What people keep" icon="user">
    Judgment about people · relationships with candidates and hiring managers · setting the standard · every consequential decision · accountability.
  </Card>

  <Card title="What AI Employees take on" icon="bot">
    Volume · consistency · coordination · availability outside working hours · producing evidence · never getting tired at candidate two hundred.
  </Card>
</CardGroup>

## What does not change

<CardGroup cols={3}>
  <Card title="A person is still accountable" icon="shield">
    Every AI Employee has one named human owner. Accountability is never diffused across a system.
  </Card>

  <Card title="People still decide" icon="scale">
    Outcomes that materially affect a person are made by a named human, at every
    ratio. The ratio changes supervision, never authority.
  </Card>

  <Card title="Oversight stays real" icon="eye">
    A higher ratio is only legitimate when the record proves the judgment holds. Scale that outruns calibration is not efficiency.
  </Card>
</CardGroup>

## Adopting it

<Steps>
  <Step title="Start with one role and a low ratio">
    One AI Employee, one owner, heavy review. You are calibrating criteria, not saving time yet.
  </Step>

  <Step title="Measure agreement, not volume">
    Track how often your reviewers agree with the recommendation. Agreement is the
    signal that earns a higher ratio.
  </Step>

  <Step title="Raise the ratio deliberately">
    Add roles to a portfolio when agreement is consistently high and escalations
    are the genuine exceptions.
  </Step>

  <Step title="Reshape the job description">
    The role changes from executing a process to owning a standard and a portfolio. Say so explicitly — the people doing it need to know their job changed.
  </Step>
</Steps>

## FAQs

<AccordionGroup>
  <Accordion title="Is this about reducing headcount?">
    That is a decision for the organization, not a claim we make. What the platform changes is how much one accountable person can competently oversee. Most customers redeploy that capacity into hiring more, hiring better, or giving candidates and hiring managers a materially better experience.
  </Accordion>

  <Accordion title="How many AI Employees can one person really supervise?">
    It depends on how much judgment the work carries and how well-calibrated the
    criteria are. Teams typically start between one and five and grow into the
    tens as agreement between reviewers and recommendations proves out. There is
    no fixed license-imposed limit.
  </Accordion>

  <Accordion title="What stops a person rubber-stamping at a high ratio?">
    Override rates and approval patterns are visible in the record. A reviewer
    approving everything instantly at a high ratio is a measurable pattern, and it
    is the exact signal to lower the ratio or recalibrate.
  </Accordion>

  <Accordion title="Do our people need technical skill for this?">
    No. Configuring and supervising an AI Employee is a business task — a mandate,
    an approval boundary and a named owner. See [Agent Studio](/smaya/studio).
  </Accordion>

  <Accordion title="Is any of this available today?">
    Yes, in part. AI Employees, approval before writes and the record are Available in Exterview. The Agent Inbox, the consolidated portfolio view, suspending a role in one action, and AI Employees working on a standing goal are Upcoming.
  </Accordion>
</AccordionGroup>

## Related

<div className="kb-links">
  *   <span className="kb-ico">
        <Icon icon="id-card" />
      </span>
    [AI Employees](/smaya/ai-employees). What you are supervising.
  *   <span className="kb-ico">
        <Icon icon="git-branch" />
      </span>
    [Agent Studio](/smaya/studio). Where a portfolio is configured and watched.
  *   <span className="kb-ico">
        <Icon icon="shield-check" />
      </span>
    [Governance & Control](/smaya/oversight). Why a high ratio stays safe.
</div>