An AI assistant for a financial advisory firm
"Our AI assistant handles 80% of client questions before they ever reach the team. Response time dropped from hours to seconds."
— Founder, Financial Advisory Firm
The challenge
A [PLACEHOLDER: boutique] financial advisory firm with [PLACEHOLDER: ~$400M] under management was being eaten alive by inbound client questions. Most weren't complex — statement explanations, distribution timing, document requests, repeat questions about the firm's process. But each one pulled an advisor away from real planning work, and clients were waiting hours for replies that should have taken seconds.
The founder did not want a chatbot. He wanted something that knew his firm — its philosophy, its language, its compliance posture — and that he would be comfortable putting in front of clients.
Our approach
We started by reading. The firm had a decade of client emails, a written playbook, FAQ documents, and a compliance manual. We pulled them together with the founder and identified the categories of questions where an assistant could safely respond directly, where it should draft for advisor review, and where it should hand off to a human immediately.
We then built the assistant against that taxonomy — and importantly, gave it a published refusal pattern: clear language for the cases where it would not answer and would route to a person. The compliance team reviewed every response category before launch.
What we built
- A client-facing AI assistant grounded in the firm's playbook, FAQ, and process documentation — not a generic LLM with a logo on it.
- A three-tier response policy: answer directly, draft for advisor review, or route to a human — with the rules visible and editable by the firm.
- Source citations on every direct answer, so clients can see which document the assistant pulled from.
- An advisor inbox that surfaces every conversation flagged as sensitive, with the assistant's draft already prepared.
- An admin view that shows weekly question volume, top categories, refusal rate, and any drift in tone — so the firm can keep the assistant aligned over time.
Results
- ~80% of inbound client questions answered without advisor involvement.
- Response time dropped from hours to seconds for the questions handled directly.
- [PLACEHOLDER: ~10 hours/advisor/week] recovered for planning work.
- Zero compliance incidents in the first [PLACEHOLDER: 90 days] post-launch — the refusal pattern routed every sensitive question to a human.
Timeline
- Weeks 1–2: Document review, taxonomy of question types, compliance sign-off on response categories.
- Weeks 3–4: Assistant build against the firm's grounded knowledge base, with citations and refusal pattern.
- Week 5: Internal pilot — assistant answered for advisors only, every response reviewed.
- Weeks 6–7: Soft launch to a subset of clients, weekly review with the founder and compliance.
- Week 8: Full client launch, admin dashboard live.
Total elapsed time from kickoff to full client launch: [PLACEHOLDER: about eight weeks].
— The IguanAI Team
If your team is fielding the same client questions over and over, there is usually a clean place to put an assistant — and a clean place not to. Fifteen minutes is enough to tell.
Book a Free Call