Knowledge base — FAQs.
FAQs are the heart of your chatbot. The bot only answers from what’s in your KB.
Adding a single FAQ
Knowledge Base tab → Add FAQ form. Two fields:
- Question — what a customer might type (this is what gets matched)
- Answer — what you want the bot to say (in your voice, in your words)
Click Add to knowledge base. The FAQ is searchable instantly — no waiting, no batch processing.
Editing an FAQ
Each FAQ card has an Edit button. Click → inline form opens with the current question and answer. Make your changes → Save. The change is searchable immediately.
Editing the FAQ in the dashboard re-syncs the search index automatically. If you’ve ever heard “the bot is still saying the old answer” it’s because someone edited the database directly without re-syncing — which the dashboard never does.
Deleting an FAQ
Each FAQ card has a Delete button. Click → it’s removed from the search index immediately. No undo — but the rest of your KB is untouched.
Bulk import (CSV/XLSX)
Knowledge Base tab → Import CSV / XLSX. Upload a spreadsheet with two columns:
| question | answer |
|---|---|
| What are your hours? | We’re open Mon–Fri 9am–5pm EST. |
| Do you ship to Canada? | Yes — Canada-wide flat-rate shipping is $9.95. |
Constraints:
- Max 200 rows per upload
- Headers required:
questionandanswer(case-insensitive) - Both columns must have content (rows with empty cells are silently skipped)
- File types:
.csv,.xlsx,.xls
The dashboard provides a template CSV you can download as a starting point.
How many FAQs should you write?
| FAQ count | Behaviour |
|---|---|
| 0–5 | Bot will fall back to handoff frequently — KB is too thin |
| 10–20 | Reasonable starting point; expect 50–60% deflection on common questions |
| 30–60 | Sweet spot for most SMBs; 70–80% deflection |
| 100+ | Mature KB; 80%+ deflection on routine questions |
Better to have 30 great FAQs covering your top 20 issues than 200 thin entries. Every entry should be a real answer to a real question — not generic marketing copy.
Writing FAQs that work
The bot does well when:
- Questions match how customers actually phrase things — “how much does it cost?” not “pricing structure inquiry”
- Answers are concrete and specific — “$49/month, billed annually” not “affordable, flexible pricing”
- One question per FAQ — don’t bundle three things into one entry
- Variations are encouraged — separate FAQs for “what’s your refund policy” and “can I get my money back” both work, even though they’re similar
The bot does poorly when:
- Answers reference things not in the KB (“see our pricing page” — the bot can’t follow links)
- Questions are too generic (“help”, “info”)
- Answers contradict each other across multiple FAQs