Permissions & Access
Decide exactly what your AI Assistant can see and change, area by area, plus which abilities it's allowed to use.
Permissions & Access#
Your AI Assistant can take real actions on your behalf — but only the ones you allow. Under AI Assistant > Settings > Access, you control its permissions area by area and switch its higher-level abilities on or off. Think of it as onboarding an employee: start narrow, expand as you build trust.
Permission Areas#
Each area has independent View, Add, Edit, and Remove toggles, so you can let the assistant look without letting it change.
| Area | Covers |
|---|---|
| Knowledge & Q&A | Documents, files, and Q&A pairs |
| Promos & Announcements | Time-sensitive content like prices, stock, events, and closures |
| People | Contacts, tags, and notes |
| Conversations | Chat history, messages, and feedback |
| Sales Pipeline | Deals, stages, and revenue |
| Staff & Services | Team members, services, and branches |
| Calendar & Bookings | Appointments and schedule exceptions |
| Core Agent & Security | Model behavior, channels, domains, and limits |
| Escalation & Handovers | Handover sessions and staff escalation |
| Comments & Reviews | Social comments, reply drafts, and rules |
| Analytics & Reports | Performance metrics and conversation stats (view only) |
Abilities#
On top of the per-area access, you toggle what the assistant is allowed to do:
| Ability | What it allows |
|---|---|
| Compose replies | Draft customer responses for you to review |
| Auto-send composed replies | Send those drafts automatically (requires Compose replies) |
| Proactive alerts | Notify you about important changes and issues |
| Action chaining | Carry out multi-step requests in sequence |
| Remember memories | Learn and store useful context from your conversations |
| Manage comments & reviews | Update social comment and review rules and statuses |
Recommended Starting Point#
- View-only first — let the assistant read across areas so it can answer questions and pull numbers, without changing anything.
- Add Compose replies — let it draft, but keep auto-send off so you approve everything.
- Expand gradually — grant Add/Edit on the areas you trust it with (e.g., Promos, Q&A) as you see it perform well.
- Keep sensitive areas tight — be deliberate about Core Agent & Security and anything that affects customers directly.
Sensitive actions ask you to confirm before they run, and the assistant can never exceed the access you've granted — if it lacks permission, it simply tells you instead of acting.
Related Pages#
- AI Assistant Overview
- Teams & Members — human roles and permissions
- Settings > Security — agent-level security controls