Where AI is used #
AI runs only when you explicitly ask for it:
- AI review buttons. The optional AI reviews on the trip match, document scan, interview, and bank statement pages send your entered data to an AI model and return scored feedback, shown in a clearly labeled "AI-assisted review" block alongside the rule-based result.
- The assistant's conversational replies. The BROVi assistant's answers are AI-generated.
- Catalog research, behind the scenes. AI helps find and triage candidate official links for the document library, but nothing it finds reaches you until a human reviewer approves it.
Higher paid tiers route AI reviews to stronger models; when a cap affects your request, you are told rather than silently downgraded.
Where AI is not used #
- Baseline scores are deterministic. The standard trip match, document, financial, and interview scores come from transparent fixed rules. The same inputs always produce the same score, and no model is involved.
- Document links are never AI-generated. The assistant does not compose or guess URLs; it only serves human-approved entries from the verified library.
- Nothing runs in the background. No AI processing happens on your data unless you pressed the button that says so.
Honest limits of AI output #
AI reviews and assistant replies can be wrong: confidently phrased, specific, and still incorrect. Treat them as a well-read second opinion, not a ruling. Where an AI output disagrees with an official source, the official source is right. AI-assisted content is labeled in the product, and the site-wide disclosure in the footer covers AI-generated material generally.
Honest limit
Charging follows reality: a credit is spent only when the AI work actually ran, and failed calls are refunded. A feature whose AI is not yet real is free rather than billed as if it were, which is why the rule-based scans cost nothing.