What’s your problem statement?

Theme:

| Finance

Track

| Banking

Statement

| AI is getting better and better at impersonation and fraud. How do we keep people safe?*

Objective:

| To build a proactive, AI-enhanced security framework that effectively protects users from impersonation and fraud threats, keeping the banking ecosystem secure as AI technologies continue to evolve.

The problem Oxfin : Agent Assist solves

Oxfin : Agent Assist for BANKS

| NOTE : This solution is completely live on Production at Oxfin
Oxfin is a telecaller solution designed for bank customers to help flag and stop fraudulent activities and impersonation attempts in real-time whenever a transaction takes place. Whenever a transaction that is suspicious according to user patterns happen , user receives a call for 2FA with voice and secret password.
Instead of IVR which any other can impersonate by accepting it, image someone which has reasoning and understanding capability and someone who can converse with you in realtime , understand your voice , your passwords , have access to your accounts , can take actions i.e ChatGPT with access to your data .
This call will come when then is a possible fraud suspected by bank , and on the call you can confirm the details and bot will itself take the decision to educate you as well if its. a fraud .
Youtube Demo : Youtbe Link

Key Features ( Live in Production )

• Authentication & Identification: Real-time verification to ensure customer transactions are secure.
• Customer Problem Identification: Instant detection of potential fraud and impersonation.
• Live Issue Resolution: Enables on-the-spot problem-solving for customer concerns.
• Call Summarization: Automated summary of issues discussed, allowing quick reference.
• Data Consolidation: Provides a single data dump from multiple channels, enhancing situational awareness.

Challenges we ran into

The technological implementation for this talk , while reducing latency was a task in telsef , it should be fast enough to makean api call , stream data on websocket , generate response and tell user back in low latency was a major challenge which was tackled .