Sanlam Studios built an AI financial coach with Voiceflow that educates users and drives leads by offering personalized financial guidance and seamless advisor handoffs.
Read the Sanlam storyCommon use cases
What teams in financial services & banking automate first
Every one of these is a real agent conversation, running in the channel the customer started in.
The agent authenticates the customer, reads live balances and transactions from core banking, and explains what moved — without a queue and without asking them to recite their own account.
See how servicing agents are built
Money actually moves in the conversation: the agent matches the saved payee, checks the balance, schedules the transfer, and turns it into a standing order — with an audit trail on every step.
See how transactional agents are built
Freeze the card, open both disputes, and bring in a human the moment the customer asks — with the case already assembled, then hand back to the agent to confirm.
See how handoff works
Document capture and identity checks in the channel the applicant already uses, with the agent stating plainly where a human reviewer takes over rather than implying instant approval.
See how onboarding agents are built
45%
Offer-to-lead conversion
Also building on Voiceflow
Risk and compliance
Certified for security. Auditable for regulators.
Production volume, and the certifications to run it on regulated data.
99.95
%
Agent uptime
50
ms
Voiceflow latency
300
K
Messages per minute scale
Works with the banking CX stack you already run.
Voiceflow hands off to your human team with the full conversation, so nobody asks the customer to start again.
There is no migration project. Voiceflow sits on top of the CX stack you already run, and speaks to anything else through the API.
Most banks are live on their existing helpdesk the same day, with no change to core systems.
Frequently asked questions
Can agents access core banking systems securely?
Yes. Agents call your systems through the same authenticated APIs your apps use — Voiceflow stores no core banking credentials in the conversation layer, and every tool call is logged for audit.
How do we keep answers compliant?
Answers are grounded in sources you control, and guardrails constrain what an agent may say about rates, products, and advice. Anything outside scope routes to a human rather than improvising.
What about regulated record keeping?
Every conversation, tool call, and handoff is retained as a transcript you can export, so servicing interactions sit alongside your existing records.