> For the complete documentation index, see [llms.txt](https://docs.11x.ai/llms.txt).
> Use it to discover all available pages before exploring further.

# Brand protection

What protects your brand when a digital worker speaks for you: knowledge base boundaries, approval modes, escalation rules, disclosure, and reputation assets.


When a digital worker speaks for you, your brand is exposed in two directions: **what it says**, and **how it arrives**. The first is governed by knowledge base boundaries, approval mode, and escalation rules. The second by sending domain and caller ID reputation. Both are configurable, and neither is safe by default.

<Callout kind="alert">
The single highest-leverage control is the **boundaries** section of your knowledge base - the claims a digital worker must never make. Absence of a claim is not a prohibition on making it. Write the prohibitions down as explicitly as you write the positioning.
</Callout>

## What it says

### Boundaries in the knowledge base

Record explicitly:

| Boundary | Example |
| --- | --- |
| Unsupported performance claims | Numbers you cannot substantiate |
| Compliance and security assertions | Certifications, data residency specifics |
| Roadmap or delivery commitments | Anything not shipped |
| Pricing beyond your published posture | Discounts, terms |
| Competitor characterisations | Claims about another vendor's product |
| Legal or contractual language | Anything a contract governs |

See [knowledge base](/alice/knowledge-base).

### Approval mode

Three levels, in increasing autonomy. Choose per segment and per motion, not once for the whole workspace.

<Columns cols={3}>
  <Card title="Review every message" icon="eye">
    Nothing sends unseen. Correct while calibrating a new segment or a new knowledge base.
  </Card>
  <Card title="Approval workflow" icon="check-check">
    A named person signs off. Appropriate for regulated industries, named strategic accounts, or a first rollout where internal trust matters.
  </Card>
  <Card title="Autopilot" icon="rocket">
    No per-message approval. Only after you have validated output on that segment.
  </Card>
</Columns>

<Callout kind="tip">
Escalate through these in order. Teams that start on autopilot and later discover a boundary problem have usually already sent the message that causes it.
</Callout>

### Escalation

Make these unconditional handoffs to a human regardless of confidence:

- Pricing beyond your published posture
- Contract, legal, or procurement terms
- Security and compliance specifics
- Roadmap or delivery commitments
- Anything from a **named strategic account** or an **existing customer**
- Any explicit request to speak to a person

See [smart replies](/alice/smart-replies) and [call flows](/julian/call-flows).

### Tone

Feed real **call transcripts** into the knowledge base. Transcripts teach the language your buyers actually use far better than a tone instruction does. Messaging also adapts by channel and by persona - see [AI personalization](/alice/ai-personalization).

### Disclosure

Whether and how a digital worker identifies itself, and any call recording notice, is configured by you per your policy and jurisdiction. See [consent and calling compliance](/security/consent-and-calling-compliance).

## How it arrives

Your sending domain and phone numbers are brand assets. Damage to either is slow to repair.

| Asset | Protection |
| --- | --- |
| **Corporate email domain** | Send outbound from a dedicated subdomain or separate domain, so outbound reputation never touches transactional and internal mail |
| **Sending reputation** | Automatic authentication, warm-up, placement testing before volume, and rotation of degraded domains |
| **Message construction** | Plain text by default, no tracking pixels, and every message unique - identical bulk mail is itself a spam signal |
| **Caller ID** | Branded caller ID with rotating numbers, plus automatic pause and remediation when a number is flagged |
| **Logo in the inbox** | BIMI displays your verified logo once DMARC is at enforcement - check with the [BIMI checker](https://www.11x.ai/tools/bimi-checker) |

<Callout kind="alert">
Domain isolation is the decision people regret skipping. If outbound reputation is damaged on your primary corporate domain, it affects invoices, password resets, and internal mail - not just campaigns. Decide this before you send, because moving later means warming a new domain from scratch.
</Callout>

## Your data is not training a model

Per the [11x terms](https://www.11x.ai/legal/autodial-terms), you retain ownership of the data you submit. 11x receives a non-exclusive, royalty-free licence to fulfil the service, uses data only in aggregate and anonymised form, and **does not train AI models on confidential information**.

## Frequently Asked Questions

<Expandable title="How do I stop a digital worker making a claim we cannot support?">
Add it to knowledge base boundaries explicitly, and move to an approval workflow until you have confirmed the boundary holds in practice.
</Expandable>

<Expandable title="What is the safest way to start with a new segment?">
Review mode, a narrow segment, and volume below target. Read 15 to 20 drafted messages before anything sends.
</Expandable>

<Expandable title="Should we send outbound from our main company domain?">
No. Use a dedicated sending subdomain or separate domain so outbound reputation is isolated from transactional and internal mail.
</Expandable>

<Expandable title="What happens if one of our numbers gets spam-labelled?">
It is paused immediately, remediation starts, and a clean number takes its place. If several are flagged at once, the cause is targeting rather than the numbers. See [connection boosts](/julian/connection-boosts).
</Expandable>

<Expandable title="Does 11x train models on our data?">
No. Per the terms, data is used only in aggregate and anonymised form, and AI models are not trained on confidential information. You retain ownership of what you submit.
</Expandable>

<Expandable title="How do we keep outbound and inbound saying the same thing?">
Both workers read the same knowledge base. Divergence is a scoping problem rather than a content one - check campaign-level scoping and call flow scoping.
</Expandable>

<Expandable title="Can we prove what was said on our behalf?">
Yes. Every call is recorded, transcribed, and summarized, and knowledge base changes are version-tracked. See [audit and oversight](/security/audit-and-oversight).
</Expandable>

## Next steps

<Columns cols={2}>
  <Card title="Audit and oversight" icon="file-search" href="/security/audit-and-oversight" horizontal>
    Proving what happened.
  </Card>
  <Card title="Knowledge base" icon="book-open" href="/alice/knowledge-base" horizontal>
    Where boundaries live.
  </Card>
  <Card title="Opt-outs and suppression" icon="hand" href="/security/opt-outs-and-suppression" horizontal>
    Honouring who not to contact.
  </Card>
  <Card title="Mailbox health" icon="mail" href="/analytics/mailbox-health" horizontal>
    Protecting sending reputation.
  </Card>
</Columns>

## Need help?

Email [support@11x.ai](mailto:support@11x.ai), or [book time with the team](https://www.11x.ai/demo).