> For the complete documentation index, see [llms.txt](https://docs.usepylon.com/pylon-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.usepylon.com/pylon-docs/agents/support-agent/monitor-and-iterate/improve.md). # Improve Improvements finds where your support agent fell short on real issues and proposes specific fixes, so you don't have to hunt for problems yourself. Open it from **Improvements** on your agent's page to see that agent's proposals, or from **Settings > AI controls > Improvements** to see proposals for all your agents. #### How it works As your team resolves escalations, Pylon compares the answer a teammate actually gave with what your support agent said. It also reads the thumbs up and thumbs down customers left on the agent's replies. Wherever there's a gap that a real resolution can back up, Pylon proposes a change. Every agent with recent escalations is covered. Each proposal links to the issues behind it, so you can see exactly why it was suggested. #### Types of improvements * **Guidance**: A change to the agent's guidance, such as a new rule or a clarified policy. * **Skill**: A new skill, or an update to an existing one, for a scenario the agent didn't know how to handle. * **Knowledge gap**: A new or updated knowledge base article for a question your resources didn't answer. #### Review a proposal Select a proposal to see what would change, the issues behind it, and Pylon's reasoning. Sort by **Most evidence** to start with the proposals backed by the most issues, or by **Most recent**. Before you accept a change, click **Preview** to see how the agent would reply to the original issues with the change in place. Previews don't change your agent or your knowledge base. Then choose what to do: * **Accept** or **Publish** to apply the change. For articles, you can also **Apply as Draft** to save the change as an unpublished draft. * Edit the proposal first if it's close but not quite right, or click **Regenerate** for a new draft. * **Dismiss** if it isn't useful, and pick a reason, such as **Already documented**, **Not a real gap**, **Duplicate**, or **Against policy**. Dismissal reasons help Pylon make better proposals over time. #### Review past issues By default, Improvements learns from escalations as they're resolved. To find gaps in older issues, click **Run Improvements** and choose a time range, from the last 24 hours to the last 30 days. Pylon reviews every issue resolved in that range, including issues your teammates handled without the agent. You can follow progress in the **Run logs**. #### Apply internal articles automatically To close knowledge gaps faster, turn on **Apply internal articles without review** in the Improvements settings. Pylon then writes proposed internal articles as soon as they're drafted, and applies any internal article proposals already waiting for review. Customer-facing articles, skill changes, and guidance changes always wait for your review. So does anything in a knowledge base that requires approval. {% hint style="warning" %} Internal articles aren't shown to customers, but your support agent uses them to answer customers. If an article written without review contains something wrong or sensitive, it can reach a customer through a reply. {% endhint %} #### Build a habit Review Improvements weekly. Start with the proposals that have the most evidence, preview them, and apply the ones that hold up. For bigger changes, run a backtest to confirm the change helps across more issues before you roll it out. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.usepylon.com/pylon-docs/agents/support-agent/monitor-and-iterate/improve.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.