How AI financial coaching educates customers and generates leads
If you know anything about financial services, at a certain point, you need human mediation. From insurance and credit to risk-based products and investments, a financial advisor is a crucial guide because financial products are complex. Even if customers do their own research, most customers want to talk with an advisor who will validate their thinking, help them weigh their options, or advise them with knowledge of their financial standing.
However, many clients don’t ever get the guidance they need, either because an advisor isn’t available, direct-to-consumer processes seem easier in the short term, or because customers are reluctant to admit they need financial help. “The best way to address these needs is to create an AI financial coach. This coach would be available 24/7, with a breadth of specialties, not unlike an advisor who just happens to have expertise on everything,” Andre says.
The Sanlam Studios team began by launching an AI financial coach that could engage on anything in the credit specialization, with an open-ended structure designed to keep users engaged with follow-up questions and personalized insights. Meet their trusted digital finance coach.
The 4 features of this advanced AI coach
- Educating customers using LLMs and curated content
This AI financial coach answers questions about a customer’s credit standing, presents and assesses their credit report, advises them on interest repayment, and can even help customers produce a letter to creditors.
The team at Sanlam Studios enhanced the agent with curated content through their knowledge base to give it the foundational knowledge of financial advisors. They’ve even created a credit quiz, which the coach leads the customers through to test their financial knowledge. All this is for the purpose of offering customers trusted digital financial guidance.
- Using a hybrid UX/UI interface without leaving the coaching flow
They’ve also built a credit reporting flow that allows the agent to access the credit bureau, pull the customer’s credit report, and offer customized guidance—all without leaving the agent's hybrid conversational flow. The agent leads the customer through the credit verification process, gathering their ID number and answers to security questions. Then the team deployed a hybrid interface to get the best out of each interface, whether visual or conversational. “Just because it can be a conversation doesn't mean it should be. Oftentimes, the right interface is a hybrid one,” Andre says.
This approach has enabled them to create an agent that does several tasks at once—displaying a timer, ensuring the customer knows when the verification process is about to expire, offering a deterministic conversation flow to maintain security, and generating a visually engaging credit summary. As a result, customers have a more engaging experience without the cognitive load of a lengthy text-based conversation.
“Just because it can be a conversation, doesn't mean it should be. Oftentimes, the right interface is a hybrid one.” - Andre Fredericks, COO, Sanlam Studios
- Generating leads and making personalized offers based on user data
Within the flow of education, engagement, and credit expertise the agent offers, it also makes suggestions for financial services and transfers customers to advisors for further follow ups. The coach takes into account the context of the conversations up to this point, the customer’s financial position based on the credit report or financial check quiz, and based on those factors, makes relevant financial offers. These can either be direct-to-consumer offers or offers that require a conversation with a human advisor.
- Offering empathetic judgment-free financial support
Many customers struggle with shame about their perceived lack of financial understanding or financial positions. Andre explains, “The digital coach changes all that. It’s confidential and imbued with an empathetic persona.” By creating a judgment-free persona, the AI coach becomes the go-to product for customers who are afraid to make unwise financial choices but are reluctant to talk to an advisor. And because the agent is focused on engagement and education, users feel more open to sharing their struggles and receiving financial help.
These features are just the beginning. The Sanlam Studios team worked tirelessly for six months to build this advanced agent, and now they’re reaping the benefits.
More than just metrics: conversions, quality, and engagement
Since launch, the Sanlam Studios team has measured success across 37 different metrics. Why? Because in order to conduct a successful experiment, you need data. Instead of merely relying on CSAT scores and lead conversions, they measure the impact of their AI financial coach across three categories.
- Conversions: The AI financial coach has made a significant impact on leads and conversions. Of the offers that the coach makes to customers, 45% of them lead to conversations with human advisors on related financial services.
Impact: 45% offer-to-lead rate, from AI chat to financial advisor
- Quality: The AI coach has reduced the number of user follow-up questions to zero, showing it is extremely capable of addressing ambiguity and complex questions. Similarly, 100% of user questions are successfully answered from their knowledge base.
Impact: Reduced the number of user follow-up questions to 0
Impact: 100% of user questions can be answered by their knowledge base
- Engagement: The AI coach has seen double the conversation engagement as most agents. With no marketing, they have seen a steady increase in new and active clients—350+ clients daily. Based on AI summaries, most conversations end with positive indicators from the customer. Whereas, neutral or negative tone indicators have enabled the team to iterate and improve the agent over time.
Impact: From 0 to engaging with 350+ users daily, with no marketing spend
Technical overview
Here's how the Sanlam team approached the AI agent's technical framework and build:
- Orchestration & Models: Built and deployed with Voiceflow using multiple LLMs, depending on the specific task
- Hybrid Classification & Prompting: Extensive use of prompt chaining to accomplish most conversational interactions. Limited use of deterministic flows to discrete actions only, including obtaining a credit report from the credit bureau. The team relied on contextual summarization, relevance evaluation, ambiguity clarification, and zero- and few-shot prompting techniques.
- Memory: Built a conversational architecture that allows open-ended discussion that leverages “working memory” and a conversational context that spans multiple conversations.
- Security & Stack: Developed proprietary middleware that sits between agent orchestration and customer channels for security and translation formatting.
- Reporting & Analytics: The Agent extracts all data required to orchestrate proactive engagements and deliver performance insights.
How do you do it too? Throw out your first 4 agents and experiment with safe bets
Most agents do customer support of some kind, a reactive approach that works well for many companies. But not for Sanlam Studios. “The goal was always proactive engagement—building an AI agent that educates and elevates digital advice,” Andre explains. “It’s also a unique opportunity to understand our customers' needs. We’re conducting research, experimenting, and problem-solving as we go.”
Even as they trek a new path for themselves, they have a few stars pointing them in the right direction. Here’s how they’re building (and iterating) this advanced AI financial coach.
Experiment with safe bets
Andre confessed something to me—they didn’t start this agent with Voiceflow. They only realized after spending some time working directly with LLMs that they needed orchestration processes around AI, especially those that enabled them to move quickly and experiment.
“Once we launched our AI agent in Voiceflow, we could begin in earnest. We built and threw away four or five agents. That’s how quickly we moved,” Andre says. Their choice to move so fast was informed by their chosen vertical—credit. It was a high-engagement vertical with curated knowledge that an agent could quickly be trained upon. By focusing on that vertical, they could more quickly experiment and perfect the agent’s understanding, responses, and working memory over the course of several conversations.
They started with a safe bet, built a foundation, and experimented with it until their agent was performing at high levels of success before launch.
“I told legal and compliance, we have to be first”
This small team, as part of a hundred and six-year-old financial institution, was able to do something most teams would only dream of. How? They got legal and compliance on board from the start. “From the first meeting, I told them exactly what our vision was, how we’d do it, and how they could partner with us in the process,” Andre explains. “Naturally, we have to walk the tightrope of our industry—financial services have necessary security regulations. We just ensure the choices we make are defensible to regulators.”
Just as they would manage risk or handle errors in human advising, they applied the same approach to the AI coach. “Human advisors, AI, and the people who build agents make mistakes. We’ll hold everyone to the same standard. But not perfection.” Andre explains. The team set the benchmark for success at the same level as a human advisor’s abilities and shortcomings. So when they inevitably make mistakes—as humans and AI are prone to do—they can reasonably defend the choices they’ve made to innovate.
By embracing experimentation, working with legal and compliance teams early in the process, and positioning the project with defensible choices, the Sanlam team is pushing the boundaries of what we expect agents to do.
Thrive on execution, not just ideas
The Sanlam Studios team isn’t overly concerned with just ideas or proof of concepts. They strive to execute. “We’re a small team—just a few of us and a few engineers. And none of us are afraid to get our hands dirty,” he says. They have big conceptual ideas—like making digital advice easier to understand and accessible—but are able to break down those ideas into a technical strategy.
It takes a visionary team that can take all the pieces—conversation design, LLMs, UX/UI—and find the areas of connection. Executing on those ideas is even more challenging. But their team is committed to moving through execution and iteration, and that makes all the difference.
What’s next for Sanlam Studios? Expand, expand, expand
Moving forward, the team has no desire to slow down. “We’re excited by the unknown because, in a short amount of time, we’ve amassed IP and knowledge through experimentation. But we’re not precious about any of it—we can achieve our outcomes because we’re chasing them relentlessly,” Andre says.
In fact, the next phase is expansion. In most other settings, coaches have an enduring relationship with their clients over a long period of time. In the future, the AI financial coach will have a vast working memory of customer profiles, preferences, and past financial conversations. That way, the conversations become more personalized with each engagement. Eventually, it will even proactively start conversations and offer ongoing reminders. “Like a coach, it will be there to keep you honest on your financial journey,” Andre says.
“We can achieve our outcomes because we’re chasing them relentlessly.” - Andre Fredericks, COO, Sanlam Studios
And if all that wasn’t enough, the team is making plans to expand the agent into new verticals and channels. They’ve built the foundational knowledge of their agent with the credit vertical, and they’ll soon offer financial coaching across specialties like insurance, health, estate planning, savings, and investments. As for channels, they have their sights set on WhatsApp—the chosen messaging platform for over 90% of their customers. The AI coach will educate their customers across web, mobile, and WhatsApp, proactively following up and maintaining an ongoing coaching relationship.
The little team that could (at a 100+ year old institution)
At the end of the day, the drive of this team is what sets them apart. “There's always someone doing better—obviously we want to stay ahead. But most importantly, we’re building something valuable for our clients—a service that makes a meaningful difference in their lives,” Andre says. Their team has always been committed to prioritizing engagement and education, and as a result, they’ve created an impressively adept AI financial coach, a hybrid interface that enables better customer experiences, and made a positive impact on revenue. All while innovating with AI at a hundred-year-old, multinational financial institution.
Needless to say, when we met the Sanlam Studios team, their drive and skill shocked us. Truthfully, folks in our office watched their demo and were amazed. At the end of our conversation, Andre reflects that moment back to me. “We were so glad to hear our ideas weren’t half-baked [laughs]. It’s a great validation of our hard work. We still have a lot we want to achieve. And we know we can get there.”
.svg)