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

# Improved AR Calculation with Additional Grounding Factors

The accuracy component of Automated Resolutions (AR) is now measured by incorporating **Processes, Actions, Coaching, and Company Descriptions** as additional grounding factors.

### What's Changed

Previously, AR evaluation **only considered knowledge content** to ground the evaluation of accuracy. Now, additional grounding factors are considered.

These enhancements provide a more precise accuracy assessment, reducing cases where conversations were previously judged as inaccurate—particularly when the AI agent leveraged processes, actions, or coaching during a conversation.

### How This Impacts AR

AR should generally stay the same or improve:

* More **grounding factors** now contribute to verify accuracy.
* Conversations that previously lacked knowledge references may now be correctly classified as **accurate** based on new grounding factors.
* Contradictions between knowledge and other elements (processes, actions, coaching, or company descriptions) may result in a conversation being considered inaccurate and therefore unresolved.

### Why We're Confident AR Accuracy Has Improved

* False positives, where the LLM predicts a conversation was resolved despite the user explicitly saying in the survey that it wasn't, and false negatives, where the LLM predicts a conversation was not resolved despite the user explicitly saying in the survey that it was, have both decreased as a result of this calculation change.

[Learn more about how AR is calculated](/docs/optimization/conversations/resolution-rate-optimization#ar-classification).