Training Tab
How to review AI responses, approve or reject them, and save the best answers to your FAQ library
The Training tab is where you teach your chatbot to get better over time. Every conversation your chatbot handles gets logged here so you can review what the AI said, approve good answers, reject bad ones, and save the best responses into your FAQ library for instant reuse.
Who Can See This Tab
The Training tab is visible to Admin, Office Admin, Spiritual Leader, and Care Leader roles. Other team members do not see it.
How It Works
When a visitor asks your chatbot a question, the AI generates a response on the fly. That response is automatically logged in the Training tab as a review item with a status of pending. You can then:
- Approve — The AI's answer was good. This confirms the response was appropriate.
- Reject — The AI's answer was wrong or unhelpful. Rejected items are flagged so the AI can learn from mistakes.
- Save as Answer — The AI's answer was so good you want to save it as a saved answer. Saved answers go straight into your FAQ library so the chatbot can reuse them word-for-word in future conversations.
The Review Interface
Each review card shows:
- The visitor's question — What the person actually asked
- The AI's response — What the chatbot answered (click "Show more" for long responses)
- Status badge — Pending (blue), Approved (green), Rejected (red), or Saved (gold)
- Agent type — Which care agent handled the conversation (Welcome, Prayer, Giving, etc.)
- Model — Which AI model was used (Haiku, GPT, etc.)
- Timestamp — When the conversation happened
Filtering and Searching
Use the filter bar at the top to narrow down reviews:
- Status filters — Click All, Pending, Approved, Rejected, or Saved to filter by review status
- Agent type dropdown — Filter by which care agent handled the conversation
This is especially helpful when you have many reviews and want to focus on just the pending ones that need attention.
Saving an Answer for Reuse
When you click Save as Answer on a review item, a modal appears where you can:
- Edit the answer — Tweak the AI's response before saving it (fix typos, add details, adjust tone)
- Choose an agent type — Assign the saved answer to a specific care agent (Welcome, Prayer, etc.) or leave it for all agents
- Pick a category — Organize the response under a category like Visitor Logistics, Pastoral Care, or Getting Involved
- Toggle Exact Response — When enabled, the chatbot returns this answer word-for-word instead of paraphrasing it. Use this for answers where precision matters (service times, addresses, policies).
After promoting, the response appears in your FAQ Management library and the chatbot will use it automatically for similar questions.
Stats at a Glance
Three cards at the top of the Training tab show:
- Saved Answers — How many trained answers are in your knowledge base
- Total Reviews — How many AI responses have been logged for review
- Auto-Answer Coverage — How many canned responses are available for auto-answering
Tips for Getting the Most Out of Training
- Check the Training tab weekly. Even a few minutes of reviewing pending items keeps your chatbot sharp.
- Save your best answers. The more saved answers you have, the more consistent and accurate your chatbot becomes.
- Use Exact Response for facts. Service times, addresses, and parking instructions should be exact. Faith-based answers can stay flexible.
- Reject confidently. If the AI said something wrong, rejecting it helps the system learn. You can always promote a corrected version.
Related Docs
- FAQ Management — Manage your full FAQ library
- Customizing Responses — Configure how your chatbot responds
- Knowledge Base — How the chatbot draws from your church's content