> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://poly-ai.ferndocs.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://poly-ai.ferndocs.com/_mcp/server.

# Level 1 complete

> PolyAcademy – You have a functioning agent with an inspectable baseline and can test safely across environments.

> **Tip**
>
> **Level 1 complete!** You've built a working agent from scratch. It answers questions, speaks with a voice you chose, and is safely versioned across environments.

You can now answer these questions without guessing:

* Why did the agent respond that way?
* Which topic or rule was responsible?
* How do I change this behavior safely?
* How do I verify the change worked?

## Level 1 recap quiz

#### Check your understanding

#### You created a topic called 'help' with good content, but it triggers for almost every user question. What went wrong?

* The content is too long
* The topic name is too generic – it matches too many intents
* The sample questions are too specific
* The voice settings are misconfigured

**Answer:** The topic name is too generic – it matches too many intents

The topic name is the strongest signal for retrieval. A name like 'help' is too broad and will match almost everything. Rename it to a specific, single-intent name like 'parking_policy' or 'check_in_time', then re-test to confirm it only triggers for the intended question.

#### You edited a topic, published the change, and promoted to Sandbox – but the agent still gives the old answer. What should you check next?

* Whether you're testing against the correct environment (Sandbox, not Live)
* Whether the voice model needs reloading
* Whether the agent's personality overrides topic content
* Whether you need to restart the project

**Answer:** Whether you're testing against the correct environment (Sandbox, not Live)

If the change is published and promoted, the most likely issue is testing the wrong environment. Confirm you're calling the Sandbox number or chatting in the Sandbox context – check the environment selector in Conversation Review to verify.

#### In Conversation Review, the matched topic is correct but the agent's answer is wrong. Where is the problem?

* The topic name needs changing
* The sample questions are wrong
* The topic content needs updating – the right topic triggered but the answer text is incorrect
* This is an ASR error

**Answer:** The topic content needs updating – the right topic triggered but the answer text is incorrect

If the right topic triggers but the answer is wrong, the issue is in the topic's content field – not its name or sample questions. Open the topic, update the Content text, publish, and re-test.

#### Before promoting a version from Sandbox to Pre-release, what is the most important safety step?

* Notify your team that a promotion is happening
* Compare versions to confirm all changes are intentional
* Update the publish note with a summary
* Check that the voice settings haven't changed

**Answer:** Compare versions to confirm all changes are intentional

Always compare versions before promoting. Open the version diff and confirm every change is intentional – if anything looks unexpected, investigate before promoting.

## What's next?

You have a solid foundation. Here's where you can go from here:

#### [Level 2: Advanced configuration](/academy/poly-academy/poly-academy-level-2/add-complex-topics)

Add multi-turn topics, functions, response controls, audio tuning, and advanced diagnostics

#### [Maintain your agent](/academy/maintain/home)

Best practices for keeping a live agent healthy and up to date

> **Tip**
>
> **Level 2 picks up right where you left off.** You'll use the same agent you built here and start adding real capabilities – functions that talk to APIs, multi-turn conversations, and fine-tuned speech controls.