> For the complete documentation index, see [llms.txt](https://riskguard-1.gitbook.io/riskguard-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://riskguard-1.gitbook.io/riskguard-docs/whitepaper/05-ai-intelligence.md).

# 5. AI Risk Intelligence

> A score tells an investor that something is wrong. Intelligence tells them what to do about it.

The final layer of RiskGuard translates the quantitative output of the risk engine into clear, actionable language. This is powered by a large language model integrated directly into the analysis pipeline. Rather than presenting the user with a dashboard of numbers and leaving interpretation as an exercise, RiskGuard produces a concise, structured analysis for every portfolio.

## The structure of an analysis

Each AI analysis follows a deliberate three-part structure designed to be read in under thirty seconds:

* **What is going well** — the strengths of the portfolio, grounding the analysis and avoiding pure alarmism.
* **The single biggest risk, with specific figures** — not a generic warning, but the one concrete exposure that matters most right now, named explicitly and quantified.
* **A concrete recommended action** — a specific, executable next step, such as reducing a particular allocation by a defined amount and redeploying it.

The analyses name protocols explicitly, cite real figures, and flag specific risks such as impermanent-loss exposure in a particular liquidity pair. Each carries a clear financial disclaimer: the analysis is informational and is not financial advice.

## Efficiency and design

The AI layer is engineered for cost efficiency and speed. Analyses are generated using a fast, economical model and cached per session, keeping the marginal cost of each analysis to a fraction of a cent. This makes high-quality risk intelligence economically viable to offer broadly — including to free-tier users — rather than reserving it for premium customers.

## Beyond static scoring — alerts and anomaly detection

Risk is not static. RiskGuard runs continuous monitoring that watches for material changes in a user's positions: sharp drops in protocol TVL, abnormal yield spikes, changes in admin-key configuration, and signs of liquidity exit. When a material risk event is detected, the user is alerted in the dashboard and, for high-severity events, in real time by email and optionally by Telegram, with a daily digest summarizing lower-priority signals. Users control which severities they receive and how often.

## Performance history and exportable reports

RiskGuard periodically snapshots each user's portfolio value and renders a performance history chart that breaks down value over time by chain. Users can export their portfolio history as a CSV file and generate a complete portfolio risk report as a downloadable PDF — combining the consolidated risk score, the per-dimension breakdown, the full position table with range status, performance over the period, and the AI analysis into a single shareable document.

{% hint style="info" %}
**A note on yield reporting.** For concentrated-liquidity positions, RiskGuard deliberately reports *range status* — whether a position is in-range, out-of-range, or full-range — rather than an estimated yield figure. An out-of-range position earns no fees and silently accumulates impermanent loss; surfacing that fact is more actionable than an approximate percentage. Fee-based yield is reported where it can be computed accurately.
{% endhint %}

{% hint style="success" %}
**The product thesis.** Detection finds the position. Scoring measures the risk. Intelligence tells the user what to do. Monitoring watches it over time. Together these four layers turn a portfolio of opaque positions into a continuously-assessed, clearly-explained risk picture.
{% endhint %}


---

# 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 dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://riskguard-1.gitbook.io/riskguard-docs/whitepaper/05-ai-intelligence.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

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.