Available for Agent+ plan only
Do safeguards apply to all replies?
No. Safeguards only apply to replies generated through AI reply workflows (auto-send or pending approval).
They do not apply to manual replies, on-demand AI replies generated with Generate, or workflows that use saved replies.
Can I override a safeguard?
Yes. If a draft reply is flagged in a pending approval workflow, your team will see the safeguard warning. You can:
Dismiss the draft
Edit the draft
Send anyway
For auto-send workflows, flagged drafts never go live. They stay in review or archive, depending on your settings.
Why didn’t AI reply to a comment?
Most often, it’s because:
A safeguard blocked the reply (e.g., profanity, off-topic, competitor mention).
A workflow filter excluded that comment type (e.g., wrong sentiment or tag).
No prompt matched because advanced default settings weren’t configured.
Check your safeguards, workflow rules, and prompt filters.
What happens if my knowledge sync fails?
Failed sources are shown in the Knowledge tab with a red “Failed” status.
Fix the issue by:
Checking that the URL is correct and public
Re-syncing the source
Switching to a snippet if the page can’t be scanned
How often is knowledge updated?
Weekly. Connected websites are automatically re-scanned once a week. You’ll get a notification when the sync is complete.
Which prompt does AI use if I have multiple?
If Advanced Default Settings are off → AI uses your default prompt.
If they’re on → AI picks the prompt that matches the message filter (e.g., sentiment, tags, assets).
Can I batch approve or reject replies?
Yes. From the Ready for Send section, you can approve, reject, or archive up to 25 pre-drafted replies at a time.
Can I batch generate replies?
Yes. In any To Review tab, switch to Batch Mode and click AI Reply. The platform generates up to 25 replies at a time.
📌 See: How to Use Batch Replying
Which prompt is used when I click “AI Reply” on-demand?
When you generate a reply manually, the system applies the following default logic hierarchy (highest to lowest priority):
Advanced filters (if enabled)
If the comment matches the filter conditions, the filtered prompt overrides everything else.
Works like workflow filters (platform, asset, tag, sentiment, etc.).
User default
If no filters match, the user’s personal default applies.
Each user can only set one default.
Company default
If no user default is set and no filters match, the company default applies.
Only one company default can exist at a time.
What happens when a conversation triggers on 2 or more reply workflows?
We will only generate one reply, taking into account the the last updated workflow.
Do pending replies follow the same pre-checks as auto-send?
For pending reply workflows, we skip the rate limit checks, but there are still a few conditions that must always be validated before a draft reply is generated:
Comment visibility
The comment must be visible. Hidden or deleted comments are not eligible for replies.
Private reply checks
Conversation age: For non-feedback survey workflows, the comment must be less than 12 hours old at the time of reply generation.
Asset age: The conversation must be newer than the date the asset was added to your account.
Existing reply: Replies are allowed only if the last message is not from the brand, unless the setting “Allow responses if the last message in the thread is from the brand” is enabled.
Manual reply rules: Standard manual private reply rules also apply. For example, you can only respond if the last inbound DM from a user is less than 7 days old.
Mixed reply restriction: A public comment cannot already have a private reply if you’re attempting a mixed reply workflow.
Wall post restriction: Wall posts cannot be privately replied.
First message restriction: If the workflow is set to trigger only for the user’s first message in a thread, this also applies to pending replies.
Public reply checks
The comment must not already have a public reply (replied inside the platform or detected as replied outside the platform by a workflow)
💡 Tip: If something feels off — the reply is inaccurate, too generic, or in the wrong tone — use the Troubleshooting & Optimization Guide to quickly fix it.