Jean-Pierre Dupont
Conseiller
Sophie Martin, CGP
Total Actifs
1,5m €
+4.2%
Total Passifs
153 000 €
-1.8%
Patrimoine net
1,3m €
+5.1%
Revenus annuels
141 000 €
+2.3%
Situation personnelle
Composition familiale
Actifs immobiliers
777 000 €
Épargne & Prévoyance
273 000 €
AI prototype · Wealth management · 2025
I designed and prototyped a conversational wealth advisor AI agent using ElevenLabs, combining voice and chat to help advisors collect context faster and prepare clearer client diagnostics.
The problem
Wealth advisors need a clear view of each client before giving guidance, but scattered information and time-consuming conversations make diagnosis slow and inconsistent.
Three shifts that defined how the agent supports advisors before, during and after a client conversation.
Combined voice and chat to make the diagnostic flow more natural and flexible.
Turned conversations into clear client needs, risk signals and next steps.
Designed the agent to support advisor judgment, not replace the human relationship.
The agent first identifies the decision to make, uses the structured client context, gives an initial view, then asks only the most useful follow-up question.
The interaction was designed to be short, calm and readable out loud, with responses that go directly to the point and avoid dense advisory-style paragraphs.
The diagnostic output follows a structured advisor-copilot format: initial reading, strengths, vulnerabilities, advice priorities and the best next question to ask the client.
The agent first identifies the decision to make, uses the structured client context, gives an initial view, then asks only the most useful follow-up question.
Decision to make
Identify the core question
Structured client context
Use known variables
Initial view
Clear first reading
Key risks
Surface vulnerabilities
One best follow-up question
Move the conversation forward
The interaction was designed to be short, calm and readable out loud, with responses that go directly to the point and avoid dense advisory-style paragraphs.
By voice
By text
The diagnostic output follows a structured advisor-copilot format: initial reading, strengths, vulnerabilities, advice priorities and the best next question to ask the client.
Chat started
do the diagnostic
Initial reading
Jean-Pierre Dupont, 47, widower with one child, salaried executive. Net worth of 1.3M€ with a globally solid but structurally unbalanced profile.
Strengths
Net worth is significant for age and income level. Life insurance at 216,000€ is a useful transmission tool. Liabilities remain moderate at 153,000€.
Key vulnerabilities
Real estate represents about 60% of total assets, creating a strong illiquid concentration. Retirement savings at 34,000€ is very low for age 47 — the most concerning point. Cash reserves at 23,000€ are tight with little safety margin. Widower status with one child makes protection and transmission particularly sensitive.
Advisory priorities
1. Urgently strengthen retirement savings.
2. Verify life insurance beneficiary clauses.
3. Build a more comfortable cash reserve.
4. Gradually reduce real estate concentration.
Best follow-up question
Have you set up any protection for your child in case of emergency, and are your life insurance beneficiaries up to date?
Measured improvements in advisor speed, structure and confidence after introducing the agent.
-30% diagnostic preparation time.Advisors moved faster from intake to a structured client view.
+25% faster need identification.Key signals surfaced earlier in the conversation.
-40% time structuring notes.Conversations turned into clear summaries automatically.
+40% clearer next-best action.Advisors knew what to do next with more confidence.