Intelligence

AI Risk Analyst

5 min read

What it is

Conversational AI grounded in YOUR trading data — not generic market advice. It sees your bot configurations, last 7 days of daily counters, last 3 days of trade records (max 50), and aggregate volume + gas spend. It does NOT see current market prices, news, sentiment, or anything outside that window. Use it for risk diagnosis, anomaly explanation, and concentration analysis.

How to think about it

Risk-management framed, not market-prediction framed. The sample questions are deliberately about YOUR portfolio behavior ("Which exchange has highest concentration risk?"), not market prediction ("Should I buy BTC?"). Trying to ask it "should I buy X" returns a refusal — that's by design.

Step-by-step

  1. Welcome screen with 6 sample questions

    Click any sample to start: anomaly analysis, exchange concentration, riskiest pair exposures, time-of-day pattern, CEX vs DEX balance, or 3 specific risk-reduction improvements. Or type your own question.

  2. What the AI sees (data transparency disclosure)

    Violet card on the welcome screen lists exact data access: bot configs + last 7 days counters + last 3 days trade records (cap 50) + aggregate volume/gas. NOT market prices, news, or anything outside this scope. Designed for privacy + scoped reasoning.

  3. Ask follow-up questions

    Conversation is persistent within a session. "What about my CEX/DEX split?" → "Now compare that to last month" works as a thread. The AI uses your data, not its training data, to answer.

  4. Full Analysis button

    One-click generates a comprehensive risk report covering all positions and strategies in the 7-day window. Useful as a weekly review starter.

  5. Conversation history

    Sidebar shows past conversations. Click any to resume. New conversation = fresh context with the same data scope.

Tips & pitfalls

  • NOT investment advice. The AI analyzes data; it does not predict prices or recommend specific trades. Investment-advice questions get redirected to risk analysis.
  • Be specific. "What's my exposure if ETH drops 30%?" gets better answers than "Am I at risk?" The narrower the question, the better the reasoning.
  • Use it before big decisions. "Walk through what could go wrong with opening a 30% portfolio position on SOL" gets a structured risk breakdown.
  • Conversations are private. Not shared with other users; not used to train models. Standard provider isolation.
  • If you have <10 trades in the window, the AI will say so honestly and suggest creating bots or running trades first. It won't fabricate analysis from empty data.
  • Combine with Risk Intelligence (the deterministic math page) — AI Risk Analyst is good at explaining the WHY behind the numbers Risk Intelligence computes.