I’m exploring building an AI agent using Vertex AI that can help users understand why their account balances sometimes appear inconsistent. For example, a user might check their FAB bank account and see an outdated balance, even though recent transactions have already posted. I want the agent to take structured transaction data and produce concise explanations like “This payment is pending” or “The latest deposit hasn’t cleared yet.”
My goal is to integrate this into a chatbot so customers can ask, for instance, “What’s my current FAB bank balance?” and get a quick, accurate answer without needing to contact support.
I’m especially interested in advice on:
Grounding the agent with financial data to prevent hallucinations
Handling real-time transaction feeds efficiently
Whether logic like pending transaction calculations should live inside the agent or upstream in the database
Has anyone tried building a Vertex AI agent for a similar banking or fintech scenario? Any architectural tips, sample approaches, or best practices would be very helpful.
Hi Barackboru, your idea for a Vertex AI agent is solid. Delayed balances often come from pending transactions or uncleared deposits, so handling calculations upstream and letting the agent explain results usually works best. You can also read this balance check guide
to see how FAB bank handles pending transactions and delays, which can help design clearer explanations for users.
Thanks a lot for your reply. It really helped me think more clearly about handling the calculations upstream and using the agent mainly for explanations.
The balance check guide you shared was useful too. Appreciate you taking the time to help.
This is a solid use case, but Vertex AI should act as an explanation layer, not the system calculating balances. Keep all balance logic and pending transaction handling in your backend, and pass only structured, verified data to the model so it generates clear, concise explanations.
By computing everything upstream and letting the AI focus purely on interpretation, you reduce hallucinations, maintain auditability, and ensure compliance while improving the customer experience.
