How Converla AI stays grounded in your knowledge
Converla AI answers from your approved knowledge — not invented policies — for WhatsApp and web chat.
This Converla Help article explains how Converla pulls approved knowledge into AI replies so policies are not invented. It is written for MENA and global operators who need practical steps—not a one-line blurb.
Only enable what your paid plan and brand settings allow. Enabled channels currently include WhatsApp, Messenger, Instagram, Telegram, Web Chat, and SMS.
What it is
How Converla AI stays grounded in your knowledge covers how Converla pulls approved knowledge into AI replies so policies are not invented. Use it as an operating guide inside your brand workspace.
Everything below assumes you already have (or are creating) a Converla account on a paid plan and will verify changes with a real test message.
When to use it
- You are setting up or fixing: how Converla pulls approved knowledge into AI replies so policies are not invented
- Before go-live on WhatsApp or other enabled channels (WhatsApp, Messenger, Instagram, Telegram, Web Chat, and SMS)
- When onboarding new agents who must follow the same process
- When symptoms appear (missing messages, weak AI answers, wrong permissions, or sync gaps)
Step-by-step
- Sign in to the correct brand workspace — each brand keeps inbox and knowledge isolated.
- Open the relevant Admin area (Integrations, Inbox, Knowledge, or Billing) for this topic.
- Apply the configuration for: how Converla pulls approved knowledge into AI replies so policies are not invented.
- Send a test from an enabled channel in WhatsApp, Messenger, Instagram, Telegram, Web Chat, and SMS or from the website widget.
- Confirm the outcome in the Inbox (or the relevant status screen) and note any Meta or sync warnings.
- Document ownership for your team (who connects channels, who replies, when to hand off to a human).
Operator tips
- Keep names, roles, and knowledge sources tidy so the setup stays repeatable across shifts.
- Always test with a known phone or account before announcing changes to customers.
- Connect this topic to inbox assignment and knowledge — how Converla pulls approved knowledge into AI replies so policies are not invented rarely works in isolation.
Common mistakes
- Changing the wrong brand in a multi-brand account
- Ignoring plan limits and assuming a feature is “broken”
- Testing on a disconnected channel or a widget domain that is not allowed
- Changing many settings at once without an intermediate test message
Related settings
- Integrations — WhatsApp, Meta, and other channel connections
- Inbox — assignment, search, and agent presence
- Knowledge — Manual/FAQ sources for Help and AI grounding
- Billing — plan, seats, and usage limits
FAQ
Is this on every plan?
Availability follows your paid plan (Starter, Growth, or Scale) and platform flags for your brand. Check live pricing, then settings.
Which channels are involved?
Enabled channels currently include WhatsApp, Messenger, Instagram, Telegram, Web Chat, and SMS. If something expected is missing, check plan limits and platform channel toggles.
Nothing changed after I saved — what next?
Hard-refresh, retest the channel, and confirm you are on the correct brand. For WhatsApp delivery, start at Integrations → WhatsApp.
How does this relate to AI handoff?
If automation is involved, keep knowledge factual and keep a clear path to a human in the unified inbox.