Available for Agent+ plan only
If Replies Aren’t Accurate
Check the source: Look at the “Knowledge used to generate this reply” under the draft.
If it pulled from the wrong page → uncheck or remove that page from Knowledge.
If the right content wasn’t available → add it as a snippet.
Confirm knowledge sync: Make sure your help center or website source shows as Synced in the Knowledge tab.
If Replies Sound Generic
Prompts may be too vague: Refine tone, add Do’s and Don’ts, and fill in Additional Details.
Enable post analysis: Turn on “Use AI post analysis as context” in your prompt to ground replies in the actual post.
Turn on knowledge grounding: Connect your help center or product pages so replies include real information.
If Replies Use the Wrong Tone
Check which prompt was triggered: If advanced defaults are enabled, the wrong prompt may have matched the filter. Adjust your filters.
Update the prompt: Add explicit rules (e.g., “Don’t use emojis in complaint replies”).
If Workflows Aren’t Drafting Replies
Check workflow status: Go to Control Panel > Workflows and confirm it’s active.
Verify filters: If filters are too strict (e.g., sentiment + keyword + tag), the workflow may never trigger.
Confirm prompt selection: Each reply workflow needs a specific AI prompt selected.
If Volume Is Overwhelming
Batch approve: Use the Ready for Send section to approve up to 25 pre-drafted replies at a time.
Batch reply: Generate 25 AI replies in one action from the To Review tab.
📌 See:
Continuous Optimization
Iterate on prompts: Every time you see a weak reply, add the correction as a Do, Don’t, or snippet.
Prune knowledge: Remove irrelevant or generic pages (like privacy policies) so they don’t confuse the AI.
Expand workflows: Start with common FAQs, then add new workflows for campaigns, product launches, or regions.
💡 Pro tip: Treat this as a feedback loop. Every small adjustment compounds, making your AI faster, sharper, and more on-brand over time.