AI Connections: security, permissions and limits

How access is protectedCopy Anchor Link

• Secure sign-in. The AI app never sees your BaTimo password. You sign in to BaTimo yourself and approve the connection on a BaTimo screen.

• One company per connection. The connection is tied to exactly the company you choose. The AI cannot reach any other company.

• Two permission levels.

  • Read access: view settings, conversations, leads, bookings, products and automations, and read the Help Center.
  • Management access: view and change them, add images, delete items after your approval, and send WhatsApp messages.

• Team permissions still apply. The AI acts as the person who approved the connection. Company owners can do everything. A Staff member can only do what their permissions in Users > Manage Permissions allow (see Team members and permissions):

  What the AI works with	Permission needed
  Assistant, FAQs, knowledge documents	Assistants
  Booking services, staff, settings	Bookings
  Products and categories	Products
  Message flows, comment and story-reply automations, Instagram posts	Automations
  Leads	Leads
  Conversations and WhatsApp messaging	Conversations
  Channels, connection status and website chat	Channels
  Images	The permission of the feature the image is for (for example Products for a product image)
  Help Center search and setup check	None; the setup check only covers areas you have permission for

  The same permissions apply everywhere in BaTimo, so a staff member can never do through an AI app what they couldn't do themselves.

• Suspended companies are blocked. If BaTimo suspends a company, existing AI connections stop working.

• Checked every time. Each time the AI does something, BaTimo checks again that the connection is still valid, the person still belongs to the company, and still has the permission. If you remove someone from the company or revoke a permission, their AI connection stops working immediately, including campaigns scheduled earlier that have not been sent yet.

• Sign-ins expire. The AI app's access is renewed automatically every hour for up to 30 days; after that you sign in again. The approval screen itself expires after about 10 minutes.

• Revoke any time. Settings > AI Connections > Connected Apps > Revoke invalidates the connection right away.

• No passwords or logins shared. The AI never sees the passwords or login details of your channels and calendars.

• No secrets in setup files. The plugin files for ChatGPT and Claude and the Claude Code and Cursor files you download from Settings > AI Connections contain only the BaTimo server address and descriptions. You always sign in to BaTimo in your browser; the AI app never receives your password.

• Instagram and WhatsApp sign-ins stay in the Console. The AI can't connect Instagram or WhatsApp for you. It gives you the link to the right Console page, you sign in there, and the AI checks the result. Turning a channel off keeps its connection and conversations.

• Deletes need your approval. BaTimo refuses every delete until the AI confirms it. A good AI app asks you first.

• Other companies stay invisible. FAQs, flows, contacts or images of another company are reported as not found.

• Changes are recorded. BaTimo keeps a record of every change made through AI Connections (what was done, when, by which connection and person, to which item). The content itself is not stored in that record.

• Safe images. An image can come from a public web address that starts with https://, a file you attach in ChatGPT, or image data the AI app sends directly. Images are only downloaded from public https:// addresses: private and local network addresses are refused, even when a link forwards to one, and BaTimo never sends your BaTimo sign-in to the download address. Whichever way the image arrives, BaTimo checks the file content (not the file name or the type the AI states), accepts only JPEG, PNG, GIF and WebP (no SVG), refuses oversized files and images, scans the file, and re-saves the image, which removes hidden data such as location metadata.

What the AI seesCopy Anchor Link

When the AI reads conversations or leads, the content (customer names, phone numbers, messages) is sent to the AI service you use (ChatGPT or Claude) and handled under that provider's terms and privacy settings. Only connect AI apps you trust with your customers' data, and tell customers where appropriate. Reading does not mark messages or leads as read.

Good practiceCopy Anchor Link

1. Grant the minimum. Use read-only connections for analysis and reporting.
2. Review before sending. Customer messages can contain anything. An AI reading a conversation could be misled by instructions hidden in a customer's message. Ask the AI to show you any campaign, recipient list and message text and wait for your approval before it schedules anything.
3. Check before resending. If a WhatsApp message's status is unclear, ask the AI to check the batch before sending again (see WhatsApp follow-ups).
4. Revoke connections you no longer use, and review Connected Apps regularly.
5. Staff: give team members only the permissions they need.

Limits at a glanceCopy Anchor Link

TroubleshootingCopy Anchor Link

See Troubleshooting AI Connections.

LimitationsCopy Anchor Link

• BaTimo cannot control what an AI provider does with data it receives.
• An AI can make mistakes; review important actions.