Dashboard

Quick Start

Best Practices

Use these practices to improve answer quality, reduce failed handoffs, and keep your chatbot useful over time.

Documentation Index

Fetch the complete documentation index at: /llms.txt

Use this file to discover all available QwryAI pages before exploring further.

Train with Clear Content

The chatbot performs best when your source material already explains the answer in customer-friendly language. Treat the chatbot as a layer on top of your help content, not a replacement for missing documentation.

  • FAQ pages
  • Shipping and return policies
  • Product guides
  • Warranty and refund rules
  • Troubleshooting articles
  • Store-specific support instructions
QwryAI conversation details with customer context
Use customer conversations to identify missing or confusing source content.
  • Review the exact words customers use.
  • Turn repeated questions into FAQ or policy content.

Set Clear Answer Boundaries

Tell the chatbot how to behave when information is missing. A useful support chatbot should answer from known content, ask a clarifying question when needed, and escalate instead of inventing details.

  • Do not guess order status, pricing, or policy details.
  • Use the brand tone your support team already uses.
  • Escalate sensitive account, refund, payment, or angry-customer situations.
  • Ask for missing context before giving a final answer.

Test Like a Customer

Customers rarely ask questions in perfect language. Test short questions, unclear questions, misspellings, frustrated messages, and questions with missing details.

  • Where is my order?
  • Can I return this after opening it?
  • Do you ship to my country?
  • I need a person.
  • This does not work.
  • Which product is better for me?

Review and Improve

After launch, use Knowledge Gaps, Analytics, Conversations, escalations, and tickets to decide what should be improved next. A small weekly review is usually enough to keep the chatbot accurate.

QwryAI knowledge gaps analysis
Knowledge Gaps highlights questions that need better source content or clearer chatbot instructions.
  • Review weak answers weekly.
  • Add missing content, then test the same question again in Playground.
Review areaWhat to improve
InstructionsTone, escalation rules, missing-information behavior, and safety boundaries.
Data sourcesUnreadable files, outdated policies, vague FAQs, and missing product details.
Suggested linksPages the chatbot should recommend when a written answer is not enough.
Suggested imagesVisual references that help customers understand products or setup steps.
Model choiceBalance speed, cost, reasoning depth, and answer style for the support use case.