> ## Documentation Index > Fetch the complete documentation index at: https://help.ciarem.ai/llms.txt > Use this file to discover all available pages before exploring further. # Is my agent working correctly? > How to read Monitoring: readiness, automated tests, and live health. Most bot platforms show you how much the bot does. Ciarem's **AI configuration > Monitoring** answers a different question: is it doing its job **well**? It has three tabs, one for each stage of trust. ## Readiness: before you launch The **Readiness** tab shows a clear verdict ("Ready" or "Not ready yet") over two groups of checks: **Required to launch** (business profile complete, an active AI agent, an indexed knowledge base, a connected channel) and **Recommended** (transfer rules enabled, team availability, tests passing, and more). Each failing check links to the screen that fixes it. See the [launch checklist](/getting-started/launch-checklist). ## Tests: proof before and after every change The **Tests** tab runs scenarios against your actual configuration: * **Generated tests** are created automatically from your setup: your escalation rules, safety topics, language, and each knowledge document. You never have to write them, and they refresh as your configuration changes. * **Your custom tests** are scenarios you add yourself with **Add scenario**, for behavior specific to your business you want checked every time. Results live in the **Results** subtab. After you change a setting or update knowledge, rerun and compare; that is how you edit with confidence instead of hoping. The Tests tab: generated tests for escalation, safety, language, and knowledge, plus your custom tests ## Live health: once you are live The **Live health** tab summarizes real conversations over the window you pick (24 hours, 7 days, 30 days), with an overall status badge such as **Healthy**, and shows how many AI turns the numbers are based on: * **Handled by AI**: the share of turns the agent handled on its own * **Transferred**: the share handed to a human * **Couldn't answer**: where the agent had no grounded answer * **Errors**: technical failures * **Response time**: how fast the agent answers (p50 / p95) Below the numbers, **Why conversations were transferred** breaks transfers down by rule (for example "Upset customer"), which turns a percentage into something you can act on: a knowledge gap to fill, a rule to tune, or a genuinely human moment to leave alone. Live health with the overall status badge, the five metrics, and why conversations were transferred ## A note on the first weeks With low conversation volume, percentages move a lot. Live health tells you how many AI turns it is based on precisely so you can judge for yourself; do not chase a percentage in week one. Focus on transfer reasons and on running the playground with real customer questions.