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

# Create an AI agent

> Walk the four-step wizard (targeting, pace, goal and review) and understand the numbers it puts in front of you.

export const Screenshot = ({id, alt, caption, frame = 'browser', url, marks = [], lang = 'en', workspace = 'northwind-outbound'}) => {
  const src = `/images/screenshots/${lang}/${id}.png`;
  const chrome = url ? `app.pipelime.ai/${workspace}${url}` : 'app.pipelime.ai';
  return <figure className={`pl-shot pl-shot--${frame} not-prose`}>
      <div className="pl-shot__frame">
        {frame !== 'bare' && <div className="pl-shot__bar">
            <span className="pl-shot__dots">
              <span className="pl-shot__dot" />
              <span className="pl-shot__dot" />
              <span className="pl-shot__dot" />
            </span>
            <span className="pl-shot__url">{chrome}</span>
          </div>}
        <div className="pl-shot__media">
          <img src={src} alt={alt} loading="lazy" />
          {marks.map(mark => <span key={mark.n} className="pl-shot__mark" style={{
    left: `${mark.x}%`,
    top: `${mark.y}%`
  }} aria-hidden="true">
              {mark.n}
            </span>)}
        </div>
      </div>
      {caption && <figcaption className="pl-shot__caption">{caption}</figcaption>}
    </figure>;
};

By the end you'll have a new AI agent: its targeting, pace, goal and instructions all set, and either sending already or waiting for you to connect a channel.

<div className="pl-availability">
  <div className="pl-availability__row">
    <div className="pl-availability__label">Where</div>
    <div className="pl-availability__value"><span className="pl-path">Outreach<span className="pl-path__sep">→</span>New AI Agent</span></div>
  </div>

  <div className="pl-availability__row">
    <div className="pl-availability__label">Your role needs</div>
    <div className="pl-availability__value">Create access to New AI Agent <code>create-campaigns</code>. Admin and Member have it by default.</div>
  </div>

  <div className="pl-availability__row">
    <div className="pl-availability__label">To create or change</div>
    <div className="pl-availability__value"><code>create-campaigns</code> to add one, <code>update-campaigns</code> to change one, on top of the permission above.</div>
  </div>

  <div className="pl-availability__note">Not listed in the sidebar today; reachable from the surfaces that link to it, and by URL. If you cannot find this in your sidebar, your workspace may have a custom menu configuration. Contact support and we will check it for you.</div>
</div>

<Note>
  You don't need a connected email account, LinkedIn account or WhatsApp account to start this
  wizard. An agent created with no channel launches paused instead of being blocked. Connect one
  afterwards and switch it on.
</Note>

<Steps>
  <Step title="Targeting: who should it look for?">
    The wizard opens on a suggestion rather than a blank form. While it reads your company profile
    and pre-fills the filters, a panel headed "Suggesting your targeting…" names the stages one at
    a time: reading your company profile, drafting your ideal customer profile, sizing the
    reachable market, picking the decision-makers to target, pre-filling your targeting filters.
    Those lines advance on their own timer: they are narration, not a progress bar. If the
    suggestion fails, a message says so and the fields are left blank for you to set by hand.
    Either way you land on the same screen next.

    <Screenshot id="ai-agents/create-an-ai-agent--ai-prefill" url="/ai-agents/new" alt="The Targeting step while the suggestion runs: a panel headed 'Suggesting your targeting…', a line of explanation under it, and one narrated stage ('Reading your company profile') standing in for the filter fields" caption="This runs once, on its own, the moment the wizard opens. There is nothing to click to start it. The state exists only while the suggestion is in flight, so this capture holds the request open: a representative example rather than a live screen." />

    Once it settles, adjust or add to what it found. **Industry** takes one or more terms: you
    can leave it empty once the hiring filters below it are configured, because a company hiring
    for a named role is a complete target on its own. Add a second term and a match toggle
    appears: **Any term** widens the search, **All terms** narrows it. **Company locations** is
    where the companies are headquartered; empty means worldwide. **Decision-maker roles** is the
    people inside each company the agent should reach.

    **Advanced targeting** stays collapsed until you open it, and holds company size, by
    employee-count range, and the locations to exclude. See
    [Decision-Maker Targeting: filters explained](/en/finding-leads/decision-maker-targeting) for
    what every filter here actually widens or narrows, hiring filters included.

    A right-hand panel keeps a live count of matching companies as you change the filters, with a
    handful of real company samples beneath it. **Continue** unlocks as soon as one *company*
    filter is set: an industry term, a location, an excluded location, a company size or a hiring
    filter. Decision-maker roles do not count towards it: they decide who is contacted inside a
    company, not which companies are found.

    <Warning>
      Continue will let you past on a location or a company size alone, but the agent cannot be
      created that way. Creating it needs at least one industry term, or one hiring filter. Set
      one before you reach the last step, or the launch fails partway.
    </Warning>

    <Screenshot id="ai-agents/create-an-ai-agent--targeting" url="/ai-agents/new" alt="The Targeting step with an industry term and a decision-maker role entered as chips, Advanced targeting still collapsed, and the estimated-reach panel in the right rail" caption="The right rail's count is what Total companies defaults to on the next step, until you edit it yourself." />
  </Step>

  <Step title="Pace: how fast should it go?">
    Three fields set how the agent sources. **Total companies** is where it stops: a chip offers
    to match it to the live count from Targeting in one click. **Companies per day** is the daily
    research pace; the field's own helper text names the ceiling your plan allows. **Contacts per
    company** is how many decision-makers it finds inside each one, and can't go below 2. A line
    beneath the three turns them into a daily number: how many new leads that pace produces a day,
    and (once Total companies is set too) about how many days it takes to work through every one
    of them.

    <Screenshot id="ai-agents/create-an-ai-agent--pace" url="/ai-agents/new" alt="The Pace step with Total companies, Companies per day and Contacts per company filled in, and the resulting daily-leads estimate below them" caption="Change any of the three numbers and the estimate below updates immediately." />

    **Also find phone numbers** sits below the pace numbers, off by default. Switch it on and each
    contact costs more credits for a chance at a direct number as well as the email: the extra
    credits are charged only when a number is actually found. A **Quality gate (optional)** comes
    after it: describe what you're looking for, and a slider sets how high a company's AI match
    score has to be before the agent contacts it at all. See
    [How many credits does a lead cost?](/en/finding-leads/credits-per-lead) for what a phone
    number adds to the price.
  </Step>

  <Step title="Goal: what should it drive toward?">
    Four cards set what every conversation is working toward: **Meeting Link** (leads book a call
    straight into your calendar), **WhatsApp** (moves interested leads into a WhatsApp chat),
    **Website Link** (sends leads to a landing page or website) and **Custom** (you describe the
    outcome and the AI adapts). **Meeting Link** is selected by default. Whichever you pick, the
    panel below asks for that goal's destination; switching cards clears whatever you'd filled in
    for the one before.

    <Screenshot id="ai-agents/create-an-ai-agent--goal" url="/ai-agents/new" alt="The Goal step with the Meeting Link card selected and a meeting link chosen in the panel below it" caption="Switching cards clears whatever was filled in for the goal you're leaving." />

    Below the goal, a collapsed **Instructions** accordion holds five fields: **Behaviour & tone**,
    **What you sell**, **Outbound**, **Reply** and **Follow-ups**. Open it and four of the five are
    already filled with default text, not a blank box. **What you sell** is the exception; it has
    no default, only a greyed-out example of what to write. Leave a field as it is and the agent
    runs on that exact text. Clear one back to empty and the agent doesn't launch with a blank
    prompt either: an empty field is not saved at all, and the built-in default applies at writing time,
    the same text these fields started from. See
    [Agent instructions](/en/ai-agents/agent-instructions) for what each field actually controls.

    A **Sending channels** accordion appears too, once the workspace has at least one connected
    email account, LinkedIn account or WhatsApp account. Email accounts rotate across all of them
    by default: this is where you'd hand-pick a smaller set instead. LinkedIn and WhatsApp
    accounts are the opposite: none are attached until you pick them here. A workspace with no
    channel at all doesn't see a warning on this step. See *What happens next* below for what
    that means at launch.
  </Step>

  <Step title="Review and launch">
    The name here (auto-generated from your targeting, editable with the pencil beside it) is
    the first time the wizard shows you a name at all; nothing earlier in it asks for one.

    Five boxes summarize the whole agent. **Target** repeats the estimated company count, the
    roles and the locations from **Targeting**. **Sourcing pace** repeats the daily formula from
    **Pace**, plus a line on what each lead costs in credits. **Channel** names which email
    accounts, and which LinkedIn or WhatsApp accounts, it sends from. **Goal** names what you
    picked and its destination. **Instructions** reads **Customized** the moment you've changed any
    of the four pre-filled fields, or typed anything into **What you sell**; otherwise it reads
    **Using built-in defaults**.

    <Screenshot id="ai-agents/create-an-ai-agent--review" url="/ai-agents/new" alt="The Review step with the agent's auto-generated name, and five summary boxes for Target, Sourcing pace, Channel, Goal and Instructions" caption="This is the only screen in the wizard that shows the agent's name: rename it here before you launch." />

    **Launch agent**, at the bottom of the page, creates a tag, a lead source and a campaign
    together and takes you to the new agent's dashboard.
  </Step>
</Steps>

## What happens next

<Check>
  A new agent appears in AI Agents with the targeting, pace, goal and instructions you just set,
  and the wizard drops you on its dashboard.
</Check>

Whether it starts sending depends on one thing: does the workspace have a connected email account,
LinkedIn account or WhatsApp account at that moment? With one, outreach is switched on at creation.
Without one, outreach is created switched off, and the confirmation message says so, telling you to
connect a sender to start outreach. Sourcing is started for you either way, so an
agent launched without a channel sits at **Outreach paused**: it keeps finding leads, and sends
nothing until you connect an account and switch outreach back on.

Anything you set here can be changed later from the agent's Settings tab. See
[Agent settings reference](/en/ai-agents/agent-settings-reference) for what every field there does.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Continue won't light up on the Targeting step">
    It's disabled while the AI suggestion is still loading, and again once that settles if no
    company filter is set: a tooltip says to add at least one targeting filter. An industry term,
    a location, an excluded location, a company size or a hiring filter each unlock it. A
    decision-maker role on its own does not: roles narrow who is contacted inside a company, not
    which companies are found.
  </Accordion>

  <Accordion title="Continue won't light up on the Pace or Goal step">
    On Pace, all three fields need a number, and Contacts per company can't be below 2. On Goal,
    it's either the destination your chosen card asks for (a meeting link, an international phone
    number, a URL or a written goal) or the sending channels: if you opened **Sending channels**,
    turned on **Choose specific senders** and then picked none, Continue waits until you select at
    least one.
  </Accordion>

  <Accordion title="Companies per day is red but Continue still works">
    The field turns red once you go past what your plan allows, and nothing in this wizard stops
    you there: Continue only checks that the three pace fields hold a valid minimum, and the number
    is saved as you typed it. Keep it under the figure the field's own helper text names: the form
    you use later to edit a lead source does refuse a value above that ceiling, so an over-cap
    agent is one you have to bring back down before you can change anything else about its
    sourcing.
  </Accordion>

  <Accordion title="Agent creation failed midway">
    The pieces are created in order (the agent tag, then the lead source, then the campaign), so
    an early failure can leave the first one or two behind. The error names what it managed to
    create and sends you back to the AI Agents list to finish setup there. One cause this wizard
    makes easy to hit: targeting with no industry term and no hiring filter. Continue accepts it,
    the lead source is rejected, and the tag is what gets left behind.
  </Accordion>

  <Accordion title="The agent launched paused">
    That's expected with no channel connected. See *What happens next* above. Connect an email
    account and switch outreach back on from the agent's dashboard. See
    [Add senders: connect, import or managed infrastructure](/en/email-accounts/add-senders).
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Agent settings reference" icon="sliders-horizontal" href="/en/ai-agents/agent-settings-reference">
    Every field from this wizard, and everything else, section by section.
  </Card>

  <Card title="Agent instructions" icon="messages-square" href="/en/ai-agents/agent-instructions">
    What Behaviour & tone, Outbound, Reply and Follow-ups actually control.
  </Card>

  <Card title="Decision-Maker Targeting: filters explained" icon="users" href="/en/finding-leads/decision-maker-targeting">
    Industry, company size, locations, roles and seniority: what each filter widens or narrows.
  </Card>

  <Card title="How many credits does a lead cost?" icon="coins" href="/en/finding-leads/credits-per-lead">
    What a revealed phone number adds to the price per contact.
  </Card>

  <Card title="Create a meeting link" icon="calendar" href="/en/meetings/create-a-meeting-link">
    Set one up first if Meeting Link is the goal you want on the Goal step.
  </Card>
</CardGroup>