Client Onboarding
An LLM-driven conversational interview that replaces PDF intake forms with a 10-minute adaptive conversation — producing a structured financial profile, risk assessment, and draft IPS.
A conversational AI system that conducts adaptive client onboarding interviews for wealth management firms. The client talks for ten minutes instead of filling out forms. The system extracts a structured financial profile — risk scoring with rationale, prioritized goals with contribution modeling, account-linking recommendations, and a draft Investment Policy Statement — ready for the advisor client and advisor to review and refine together before the first real conversation.
The interview adapts in real time. A straightforward profile moves fast. A client with a business sale, an inheritance, and three kids approaching college triggers deeper follow-up on estate planning, tax-loss harvesting, and education funding. Extraction runs through a two-tier validation pipeline: a fast model handles most transcripts, with automatic escalation to a more capable model when validation fails. Every extracted field carries a confidence score — the system flags what it doesn’t know.
Built in a few hours with Claude Code. SvelteKit frontend, Hono API, both deployed as Cloudflare Workers. The companion blog post makes the business case: onboarding is a revenue problem, and the technology to fix it is ready.