SoFi's AI-powered financial coach now handles roughly half of its customer conversations about investing, a sign that conversational AI is moving beyond simple Q&A into sustained financial guidance. The company built Coach on eight years of human planner interactions, and it treats context and communication as core product features rather than afterthoughts.
SoFi Head of Advice and Planning Brian Walsh said customers expect different levels of interaction depending on the situation. "People want to interact and receive guidance in different ways," he said. "Sometimes it's going to be a convenient chat-based experience. Sometimes it's going to be a little bit more complicated, more high-touch with humans."
Context beats answers
The gap between helpful and harmful financial advice often comes down to what the adviser knows about the client. Telling someone with a fully funded emergency reserve to invest excess cash is reasonable. Saying the same thing to someone carrying a high-interest credit card balance could be a mistake.
"It's one thing to say, 'Hey, you have a bunch of extra cash in this account. This cash should be invested, or it should be used to pay down debt,'" Walsh said. "But if you don't have any context in how it fits into their overall picture, that could be correct, it could be incorrect."
Most consumers hold accounts across five to fifteen institutions, which makes assembling a complete picture difficult. SoFi's Coach pulls that data together, then uses visual formats to get the point across. A written statement about discretionary spending falling by 30% into one category rarely changes behavior; a pie chart where that category dwarfs everything else can trigger a different response.
"There are a lot of financial concepts that are much easier to express visually than they are in writing," Walsh said. "It opens up this whole new dynamic of how we can communicate these topics in as simple and understandable a manner as possible."
Coach also retains memory about family situations, priorities, and previous discussions, so customers do not have to rebuild their financial story with every conversation. That matters for behavior as much as convenience - this kind of contextual software is one of the practical ways AI for Finance is changing product experiences.
Built-in boundaries for AI guidance
SoFi designed Coach around rules drawn from a decade of human planner interactions. Rule sets flag when a conversation is moving beyond safe automated territory into matters that require judgment. At that point, the system escalates to a person. This approach matters for any company building AI features inside a regulated industry, who may find the human-in-the-loop model relevant when applying AI for Product Development to their own risk areas.
"There are going to be clear rule sets that are built in," Walsh said. "This is going to expand beyond the scope of the advice or the guidance that we want to provide. At that point, we'll escalate this to a human interaction."
The setup creates a two-sided gain, Walsh said. People who previously received no advice at all get useful direction, and human planners receive relevant client data already gathered by the Coach system.
"You get this snowball effect where you're exposing expertise to more people," he said. "You're making the human side more efficient."
Why this matters for finance and product development professionals
For product teams building AI assistants, the lesson from SoFi is that escalation logic is a feature, not a liability. The people who designed the product list exactly where the software stops and sorted out how data gathered during AI conversations flows back to human specialists to make them more effective. That division of labor - not the quality of the chatbot's answers - looks like the actual sustainability of the model.
Your membership also unlocks:
