Sarah - AI Voice Banking Assistant

A voice assistant for a bank: balance, transfers, card blocking - spoken, not tapped.

−20%: load on support
20+: scenarios without an agent

ROLE Product, design, engineeringSTACK Speech-to-text · LLM · TelephonyTIMELINE 2025

I want something like this AI implementation

THE PROBLEM

The contact centre was buried in routine requests; the IVR menu annoyed customers and solved nothing.

WHAT WE BUILT

Sarah understands natural speech, authenticates the customer and closes 20+ scenarios without an agent. Anything harder goes to a human - with the context attached.

Live screenshots are on the way

Screens · 01/04

Home screen

What we shipped, in detail

Sarah classifies the call within the first seconds. "I lost my credit card" routes straight to the fraud desk with context attached, bypassing the queue entirely, while routine questions get resolved without a human at all.

Unlike a passive IVR menu, Sarah interviews the caller - name, verified phone number, reason for the call - and validates each field mid-conversation, so a human who picks up an escalated case already has verified context.

Within seconds of hangup the backend transcribes the call, extracts the key fields and pushes a structured summary to the bank's CRM, so staff triage by reading rather than replaying audio.

Voice recognition and synthesis were tuned for Norwegian dialects, with slower, clearer cadence for elderly callers, so one voice persona can serve the bank's full customer base.

A lost card at 2am no longer waits for branch hours - Sarah captures the request, blocks the card and routes to fraud support around the clock, closing a real gap in the bank's prior coverage.

Previous case

Amanda - AI Sales SDR

Next case

BOMMatcher

I want something like this.