TabX - Your Enterprise-Grade AI Financial Coach

Inspiration

Banking apps today are mostly transactional—customers log in, transfer money, and log out. For Shinhan Future's Lab, the challenge was clear: How do we transform the SOL app from a simple utility into a daily financial companion that increases DAU/MAU, grows CASA balances, and drives cross-selling?

We realized that raw transaction data is useless without interpretation. Users want to save money and build wealth, but they lack personalized guidance. We were inspired to build a proactive AI agent for financial health. By leveraging our enterprise-grade AI platform, TabX, we wanted to prove that a secure, highly intelligent financial coach could create a win-win scenario: users get smarter with their money, and Shinhan achieves its strategic KPIs.

What it does

Our solution is an embedded AI Personal Financial Coach for the Shinhan SOL app, powered by TabX. It moves beyond standard FAQ chatbots by acting as a proactive wealth advisor:

Proactive Insights & CASA Growth: Analyzes spending habits and income to suggest realistic, personalized savings plans, gently nudging users to keep more funds in their Shinhan accounts (growing CASA).
Contextual Cross-Selling: Detects life events or spending anomalies to naturally recommend relevant Shinhan products (e.g., suggesting a Shinhan credit card for travel when a user asks about budgeting for a trip, or offering insurance).
Omnichannel Readiness: While designed as an embedded widget for the SOL app, TabX's core architecture allows Shinhan to instantly deploy the exact same AI coach to Zalo or Facebook Messenger to capture leads outside the app.

How we built it

We didn't just build a prompt; we built an enterprise infrastructure using TabX and Alibaba Cloud.

LLM Engine: We utilized the Qwen model series deployed via Alibaba Cloud for its exceptional reasoning capabilities and natural language fluency. Qwen acts as the "brain," parsing complex financial queries and generating empathetic, accurate advice.
Enterprise Platform (TabX): We used TabX as the orchestration layer. We utilized TabX's Knowledge Base (RAG) feature to ingest Shinhan's product catalogs, interest rates, and financial guidelines.
Integration & Deployment: We used TabX's no-code deployment widget to embed the AI seamlessly into a mock SOL app frontend.
Analytics: We configured TabX's built-in Lead Capture and Analytics to track conversation metrics, giving the Strategy Division real-time insights into user intent and cross-sell conversion rates.

Challenges we ran into

Hallucination in Financial Advice: In fintech, an AI cannot give wrong interest rates or hallucinate loan terms. We solved this by strictly grounding the Qwen model using TabX's Vector-indexed RAG, ensuring the AI only pulls from approved Shinhan documentation.
Data Security: Handling transaction data requires extreme care. We leaned heavily on TabX's architecture, which includes tenant isolation and AES-256-GCM encryption, ensuring that our proof-of-concept respects enterprise-level security standards.
Balancing Empathy with Sales: It was challenging to make the AI sound like a helpful coach rather than a pushy salesperson. We spent significant time fine-tuning Qwen's system instructions to prioritize the user's financial health first, making the cross-sell recommendations feel like a natural next step.

Accomplishments that we're proud of

We are incredibly proud of the speed to value. By leveraging TabX's existing infrastructure, we bypassed weeks of backend development. We were able to build a secure, multi-channel, RAG-enabled AI agent in a matter of hours, allowing us to focus entirely on the business logic, the prompt engineering for Qwen, and solving Shinhan's specific problem statement.

What we learned

We learned the sheer power and adaptability of the Qwen models when paired with a robust RAG system. We also learned that the biggest bottleneck for enterprise AI adoption isn't the AI itself, but the surrounding infrastructure—security, analytics, and deployment pipelines. By combining Qwen with TabX, we validated that these barriers can be eliminated.

What's next for TabX

The next step is a deep technical integration with Shinhan's core banking APIs to pull real-time transaction histories dynamically, moving from synthetic data to live user insights.

Additionally, we plan to expand our deployment capabilities specifically for the Vietnamese market by deepening our Zalo integration, allowing Shinhan to use TabX not just for customer support, but as a primary customer acquisition channel on social media. We will also be integrating Alibaba Cloud's broader ecosystem to enhance our vector search speed and security compliance.