Skip to successes (case studies)
Book Time Phone (347) 668-5684 Mail ygeffrard@gmail.com
Go back Go back

AI Upsell Agent

I designed an AI agent inside an investment bank's chatbot that reads browsing behavior to find the right moment to offer self-directed clients a financial advisor, adding tens of millions in client and firm revenue.

Overview

The firm wanted to grow revenue from self-directed clients. Its sales team already reached out by phone and email, but one answer kept coming back: "It's not the right time." So I asked: what if AI could find the right time? I designed an upsell agent inside the existing chatbot that learns from each client's browsing to suggest moving from a self-managed account to one with a dedicated financial advisor, so clients benefit from the firm's experience and technology without worrying about how the economy affects their money.

What the sales team kept hearing

“It’s not the right time.”

Key outcomes

Tens of millions

Tens of millions

Added to firm and client returns

More assets in-house

More assets in-house

Client assets brought over from external firms

Deeper trust

Deeper trust

Clients' trust in the firm, strengthened

How I started

I brought together every stakeholder, including the head of AI initiatives, developers, financial advisors and business owners, to understand the initiative's potential impact and risks. The chatbot's design system and infrastructure already existed; what it lacked was intelligence.

Initial ideas

  • Keep the chatbot always available, and open with "We notice you don't have a financial advisor" when a user asks for help
  • Show the chatbot after a set time on the site
  • Show the chatbot on specific pages

Final MVP

I had data on what worked for financial advisors. Instead of more client research up front, I recommended a pilot with a sample of self-directed clients, letting the data train the model. The agent learned from four signals:

  • Time on page — The chatbot appears after a set time on different pages, revealing which pages convert best
  • Who the client is — It learns which pages resonate for each combination of demographics and asset value
  • The right moment — It finds how long into a visit is the best time to start the conversation
  • The path — It weighs the sequence of pages a client viewed to find what leads to conversion
HR CRM

Reflection

Outcome

The agent got smarter with every visit, down to the individual client: it uses each client's full browsing history to time a personalized offer, and knows which sequence of pages and which moment make a client most likely to engage. The result was tens of millions in revenue and asset growth for the firm. Next, it can draw on the firm's dynamic content, such as articles and market pages, to find even more upsell opportunities.

Let’s build something people love to use

Hiring for a UX role or planning a product? Send me a note and I’ll get back to you as soon as possible.

Book Time Email Me Call Me
By submitting, you agree to receive SMS updates related to your inquiry

Based in:

New York and New Jersey

Quick links

Contact

Connect

© Copyright 2026 Yarley Geffrard. All rights reserved.