Skip to content
Guides

Shape how your assistant helps

Set the assistant’s profile and permissions through Setup, then configure how it communicates. Forms and templates write the same settings; guidance adds reviewed coaching.

Setup and Capabilities

Setup collects your organization type, intended work, information sources and sensitive-data needs. Review the proposed profile, its capabilities and its checklist before activation. Some configurations require MFA or platform review. Capabilities shows the effective permissions and readiness of the assistant.

Profile activation is separate from writing a persona or uploading content. Goals and guidance cannot enable a disabled tool, provide missing data, expand a plan or override platform policy. See the quickstart for initial activation.

Behavior and Goals

BehaviorSet brand voice, escalation sensitivity, handoff wording, language policy, business facts, locations and hours. Incoming media controls and assistant audio replies have separate sections. See voice configuration.
GoalsDescribe your organization, desired outcomes, work your team handles outside the chat, required details, tone, prohibited phrases and escalation preferences. This page also explains the organization defaults behind inherited values.
AI usage limitChoose the response when the workspace runs out of AI allowance. Handoff at the limit is a separate choice and requires an available confirmed escalation route; it is not inferred from other escalation settings.

Keep factual claims in Knowledge, the catalog or the dedicated business-facts fields. A style instruction that mentions a price or a service is not a verified source for that claim.

Fill one template

  1. From Assistant, select Fill one template and download the supported document template.
  2. Fill in your business information and upload it. Import creates a preview; it does not immediately change the assistant.
  3. Review each extracted field against the current value and choose what to apply. Replacing a value is an explicit field decision.
  4. Apply the reviewed changes and inspect the result. Any guidance draft created by the import still needs its separate review and activation.

The template fills the same canonical settings as the forms. If configuration changes after a preview was generated, a stale apply can be refused; refresh the import against current values. Import history supports rollback when safe and a reverse-change review when later edits prevent a direct restoration.

Add Agent Guidance

Use Add a guidance document for supplemental coaching in Goals, Behavior & tone or Example conversations. Write text or upload a supported PDF, Word, text or Markdown document. Inspect the extracted instructions, including any removed instructions, before approving it.

Approved is not live. Activate the reviewed version for it to influence conversations. The page distinguishes drafts, documents awaiting review, approved versions, live versions and versions that were replaced, revoked or expired. Only eligible live guidance applies.

Guidance supplements your settings; it does not replace them, publish business knowledge, grant action permissions or bypass security controls. Use Knowledge for the material customers need answers from.

Organization baseline

The Goals page can show a versioned organization baseline. Fields indicate whether they inherit defaults or contain your own changes, including changes applied by a template. Your explicit choices take precedence over inherited defaults.

When a newer baseline is available, inspect the adoption preview. Adopt pins that version for fields that still inherit organization defaults and preserves fields you customized. Publishing a newer baseline does not silently replace the pinned version. Reset a field only when you intend to return it to its inherited value.

Sector playbook

Where supplied, the Goals page also shows a sector playbook, its version and available assistant roles. Review an offered version and the enabled or default roles before adopting changes. These roles describe the assistant’s approach to a conversation; they are separate from the human roles on Team and remain subject to the active capability policy.