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

# Agencies & marketing

> Many accounts, one team: briefs, production, reporting, and client status that don't depend on someone holding it all in their head — with anything client-facing always going through the account lead.

It's 6pm on a Friday. Account A wants a post ready for Monday's launch;
account B sent three WhatsApp voice notes with "a few brief changes"; account
C's monthly report was due yesterday; and nobody remembers the last time
anyone spoke to account D. Each account lives in a different manager's head,
and when that manager is buried in another one, the account cools off without
anyone noticing.

**What leaves your plate:** turning a client message into a brief and tasks,
pulling a first pass of content and reports, assembling each account's status,
and noticing who's gone quiet too long all start happening on their own — with
context, within budget, and with anything client-facing waiting on the account
lead's sign-off.

## How it works, in one line

An agency is, at bottom, **many client memories running in parallel.** The
agent doesn't replace whoever owns the account — it makes sure each client's
context, the first-pass production, and the follow-through exist even when that
account's manager is busy on another.

```mermaid theme={null}
flowchart LR
    A[Trigger<br/>client message,<br/>due date, routine] --> B[Agent pulls context<br/>from Company Brain + Boards]
    B --> C[Decides + uses tools<br/>Boards, Docs, Scout, Routines]
    C --> D{Goes to<br/>the client?}
    D -->|No| E[Creates task / drafts /<br/>organizes internally]
    D -->|Yes| F[Proposes + calls the lead<br/>with the piece ready]
    E --> G[Becomes memory<br/>the next deliverable knows]
    F --> G
```

## Concrete flows

### 1. Turn a client message into a brief and tasks

**Trigger:** a client request lands — an email, a
[WhatsApp](/en/integrations/whatsapp) message, a meeting note — something like
"I need a Mother's Day campaign, aimed at a younger audience."

[Athena](/en/agents/athena), the organization's operator, pulls what the org
already knows about that account from the
[Company Brain](/en/features/brain) — tone of voice, past pieces, brand
constraints, what worked last time. She drafts a **structured brief** in a
[document](/en/features/documentos) (goal, audience, deliverables, deadline,
references) and opens the matching [tasks](/en/features/boards) on the
account's board, already assigned — design, copy, approval — with the due date
on the card.

The brief doesn't become truth on its own: it's ready for the lead to review
and adjust before it turns into work. What changed is the starting point —
instead of a blank page at 6pm, the manager opens a draft that already
understood the client.

```mermaid theme={null}
flowchart TD
    A[Client request lands] --> B[Agent pulls account<br/>history from the Brain]
    B --> C[Drafts brief<br/>in a document]
    C --> D[Opens board tasks<br/>design, copy, approval]
    D --> E[Lead reviews and adjusts]
    E --> F[Becomes account memory]
```

### 2. Research a client's market with Scout

**Trigger:** a new account joins, or a client asks for a campaign and you need
to understand the terrain before you pitch.

[Scout](/en/agents/scout) is the public-web researcher. You ask *"pull this
brand's three main competitors and how they position"* or *"what's being said
about this sector in the last few months,"* and it searches via
[Tavily](/en/integrations/tavily), returning **cited snippets** — every claim
tied to the URL it came from. No citation means it isn't asserting; it stays an
explicit hypothesis.

The result feeds into the [brief](/en/features/documentos) and is stored in the
[Company Brain](/en/features/brain): next time that account needs a campaign,
the market context is already there, and nobody researches the same thing
twice.

<Note>
  Scout researches what lives **outside** your org — markets, competitors,
  news. It doesn't talk to the client or touch the account: it delivers cited
  material so a person decides the angle.
</Note>

### 3. Draft content and reports for human review

**Trigger:** a production task lands on the board — a carousel, a blog piece,
an account's monthly report.

The agent pulls the brief, the tone of voice, and past pieces from the
[Company Brain](/en/features/brain) and produces a **first pass** in a
[document](/en/features/documentos): the caption, the script, the copy, or the
report with the month's numbers organized and a summary of what happened. It
edits and refines the document as you ask for changes in the
[chat](/en/features/chats).

That pass **doesn't go to the client on its own.** It arrives ready for the
lead to review, cut, and approve — the creative work and the judgment stay with
people. What changes is that the blank page becomes a draft, and the
end-of-month report stops eating an afternoon of copy-and-paste.

<Info>
  **Why this matters:** the expensive part of production isn't writing the
  final version — it's getting off zero. A draft grounded in the account's
  history shortens that start, and human review keeps the quality the client
  signed up for.
</Info>

### 4. Assemble an account's status from boards + Brain

**Trigger:** the client update is due — a weekly meeting, a "how's my campaign
going?" ask, the end of the month.

The agent reads the account's [board](/en/features/boards) — what shipped,
what's in progress, what's blocked — cross-references what's logged in the
[Company Brain](/en/features/brain), and assembles a **status summary** in a
[document](/en/features/documentos): this week's deliverables, next steps,
items waiting on the client. All grounded in what's actually on the cards, not
in memory.

| The status pulls from        | What it becomes in the summary |
| ---------------------------- | ------------------------------ |
| Tasks completed on the board | "Delivered this week"          |
| Tasks in progress            | "In production / next steps"   |
| Blocked cards                | "Waiting on the client"        |
| Account notes in the Brain   | Recent context and decisions   |

The summary is ready for the manager to review and send. **What goes to the
client still passes through the account lead** — the agent prepares, the person
signs off.

### 5. A routine that flags accounts with no movement

**Trigger:** a [routine](/en/features/rotinas) the agent schedules for itself —
a weekly sweep across every active account.

On fire, the agent walks the [boards](/en/features/boards) and the
[CRM](/en/features/crm) looking for accounts with **no recent activity** — no
card moved, no interaction logged for too long. For each one, it raises a
[proactive notification](/en/features/rotinas) in the account owner's bell and
opens a follow-up task, with the context of what's stalled.

Account D — the one nobody remembered talking to — stops slipping through the
cracks by accident. The silent cool-off becomes a heads-up on a day you can
still turn it around.

## What stays human

<Warning>
  The agent **proposes**, never decides or sends on its own, whenever the
  subject is:

  * **A final client deliverable** — a piece, a campaign, a report that goes
    out under the agency's name. It drafts and prepares; the lead reviews,
    adjusts, and approves.
  * **A commercial proposal** — scope, price, terms of a job or retainer.
  * **Money** — payment, refund, invoice adjustment, media budget.
  * **Contract and signature** — any commitment that becomes an obligation.
  * **Heavy external, legal, or regulatory communication**, governance
    decisions, and destructive actions.

  In those cases the agent prepares everything — the draft, the context, the
  recommendation — and **waits for the lead to approve.** Every action lands in
  the auditable trail, with author and justification.
  [Stars](/en/billing/stars) cap spend per account and per agent before any
  expensive call.
</Warning>

## Where to start

<Steps>
  <Step title="Upload one account's context to the Brain">
    Put one account's tone of voice, past pieces, and standard brief into the
    [Company Brain](/en/features/brain). That's where the agent pulls context
    from — without it, every draft starts generic.
  </Step>

  <Step title="Let the agent draft your next deliverable">
    On a [board](/en/features/boards) task, ask for the first pass of the
    content or report. Review it, adjust in the [chat](/en/features/chats), and
    see where the draft already lands — and where it still needs your hand.
  </Step>

  <Step title="Schedule the sweep for quiet accounts">
    Create a weekly [routine](/en/features/rotinas) that flags accounts with no
    movement. One agent, one cadence — and autonomy grows like a ratchet, not a
    leap.
  </Step>
</Steps>

## Next steps

<CardGroup cols={2}>
  <Card title="Scout — the researcher" icon="magnifying-glass" href="/en/agents/scout">
    Who pulls market, competitor, and trend with cited sources — the raw input
    for every brief.
  </Card>

  <Card title="Documents" icon="file-lines" href="/en/features/documentos">
    Where briefs, content, and reports start as a draft and take shape with
    your review.
  </Card>

  <Card title="Boards & tasks" icon="table-columns" href="/en/features/boards">
    Each account's board — the source of truth that feeds the client status.
  </Card>

  <Card title="Commercial" icon="handshake" href="/en/use-cases/commercial">
    How the same pattern serves prospecting and closing new accounts.
  </Card>
</CardGroup>