Go back

AI Upsell Agent

+ $10M of client and firm revenue

About

Boosted an investment bank's revenue by tens of millions with an upsell AI (Artificial Intelligence) agent embedded in a chatbot that analyzed user's browsing history to upgrade them from the self managed investment account to one with an assigned financial advisor. This enables the client to work with the experience of the firm leveraging its technologies to provide the best investment return for the client without worrying about how the economy is affecting their money. Problem came from the firm wanting more ways to grow revenue from self-directed clients, the typical human sales cycle already existed where someone would reach out to the client by phone or email, but one thing that commonly came up 'it's not the right'. So what if we can 'find the right time' with AI?

Impact

  • Increased firms and clients returns by tens of millions
  • Migrated more assets from the client from external firms into the firm
  • Increased clients trust with the firm

How we started?

We met with all stakeholders including Head of AI initiatives, developers, few financial advisors and business owners to better understand this initiative and its potential impact and risks on the firm. The chatbot design system and infrastructure was already built however lacked the integration of artificial intelligence.

Initial Proposed ideas

  • Always have the chatbot icon there and if the user seeks help a ‘we notice you don’t have a financial advisor’ initial message comes up in the chatbot
  • Chatbot to show after X seconds of user being on the site
  • When user visit X page chatbot would show up

Final MVP solution

We had some user data of what worked from financial advisors and didn’t quite do client user research with these concepts, instead we strategically thought it would be best to launch a pilot to a sample of self-directed clients and learn from the data while training it. Here are the factors we incorporated to the AI

  • The chatbot would automatically appear after X seconds on different pages to see if there are any pages that convert more
  • It would get smarter to recognize if a combination of client’s demographic and asset value correlates with page that resonates the most for the selling opportunity
  • Then we would find the time factor if a user spends X seconds on the site what’s the best time to bring up the chatbot
  • Would also factor in the sequence of pages users viewed to find correlation with conversions
HR CRM

Outcome

The data although unpredictable got smarter overtime down to the particular user meaning it took account the full user’s browsing history on the site to figure out the best upsell time personalize for the client increasing both the firms revenue and asset growth by tens of millions. It knew what sequence of pages and time for the chatbot to appear that the client would be most engaged and likely to convert. The possibilities for the future gets more interesting when considering the content on the page or articles posted on the firms website which are dynamic which may be used by the AI agent for upselling opportunities.

Other Projects

Aim Execute Solve