Handling customer balance inquiry questions through tawk.to without exposing sensitive data

Hi everyone,

I’m currently using tawk.to live chat and AI Assist to handle customer queries for a financial services website, and I’m trying to improve how we respond to balance inquiry related questions. A common example is users asking how to check their FAB bank balance or whether their latest balance is updated correctly.

Since we obviously cannot fetch or display actual banking data inside chat, the challenge is guiding users clearly while keeping responses accurate, secure, and helpful. Sometimes AI Assist gives answers that are either too generic or too detailed for a sensitive topic like balance inquiry.

I wanted to ask how others are handling similar situations. Do you rely on predefined responses, AI Assist prompts, or a mix of both when dealing with bank balance inquiries? Also, how do you phrase replies so users feel supported without expecting the chat to act like online banking itself?

Any tips on training AI Assist or structuring chat flows for financial queries would be really helpful. I want to keep the experience smooth while staying compliant and avoiding confusion.

Thanks in advance to anyone who can share their experience.

Hi @TOM2121 ,

Thanks for sharing this. A best-practice approach is to combine predefined responses with well‑trained AI using reliable data sources, so you can control exactly how the AI responds in each scenario while setting clear expectations upfront. Since AI Assist cannot access sensitive or real‑time banking data, it should be trained to clearly state this and direct users to secure channels such as the official banking app, online portal, or customer support.

Also, this article will help you tweak the responses so the AI gives more accurate replies: Help Center | tawk.to | How to fix incorrect responses delivered by AI Assist