> For the complete documentation index, see [llms.txt](https://yahya-qanie.gitbook.io/yps-ai-agent/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://yahya-qanie.gitbook.io/yps-ai-agent/how-answers-are-developed.md).

# How answers are developed

A plain-language explanation of evidence-grounded answers.

### From question to evidence

This diagram summarizes the intended user-facing process. It is not a guarantee that every answer will be complete or correct.

Explore the [interactive architecture](https://yahyaqanie.com/ypsagent) for a more detailed view.

<figure><img src="https://1435617392-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYA3KdXcYR2vQjCi6GLlA%2Fuploads%2F4Xl7YKHdfNue71kecUi6%2FScreenshot%202026-09-20%20at%202.19.33%E2%80%AFAM.png?alt=media&amp;token=7973d557-2689-4bcb-be67-be6d8fe27122" alt="Pastel card diagram: a YPS question is clarified, with a follow-up if context is missing. Relevant sources are retrieved and assessed. Sufficient evidence leads to a cited answer; limited evidence leads to an explanation of uncertainty. Both require the user to check sources and local context."><figcaption></figcaption></figure>

### Grounded in a curated collection

YPS AI Agent uses retrieval-augmented generation, often called RAG. In plain language, it draws on relevant material from a selected document collection to help produce an answer. This connects the response to identifiable YPS evidence.

The agent is not an unrestricted internet search engine. Its coverage depends on the sources available to it.

### Three kinds of content

* **Source-based information:** statements supported directly by the available documents.
* **Synthesis and interpretation:** conclusions drawn by combining or interpreting evidence across sources.
* **Proposed ideas:** activities, options, or recommendations developed for the question and grounded in YPS principles.

These categories should be distinguishable. A generated suggestion is not automatically a recommendation endorsed by the source organization.

### Clarification and uncertainty

When missing context would materially change the answer, the agent may ask a follow-up question. If available evidence is insufficient, it should explain what can be established and what remains uncertain.

### What grounding can and cannot do

Using a curated collection is intended to improve relevance and make evidence easier to inspect. It does not eliminate factual errors, incomplete coverage, misinterpretation, or translation problems. Users should still check important claims against the original documents.