SOL AI Personal Financial Coach

SOL’s AI coach that grows engagement, CASA, and cross-sell using daily behavioral insights, safe-to-save nudges, and explainable readiness.

Comment

Inspiration

Vietnam’s payments ecosystem is rapidly digitizing, but “financial behavior” is still invisible for many customers: spending is unmanaged, savings are inconsistent, and customers often don’t know which product is right (or when). At the same time, banks want to increase digital engagement and cross-sell without pushing generic offers that reduce trust. The Shinhan INNOBOOST theme around an “AI Personal Financial Coach” inspired us to build an in-app coach that drives measurable bank KPIs while staying consent-first.

Team Member

  1. ** Zack Loo** - I build and scale AI-driven solutions for both startups and enterprises, turning complex problems into practical, production-ready systems. With a track record of delivering AI agents, automation platforms, and data-driven products, I focus on outcomes—efficiency, speed, and real business value, not just experimentation.

The problem (what we observed)

1) Engagement problem: Most banking apps are transactional. Customers open the app only when they must (transfer/pay), not daily. 2) Decision problem: Customers don’t have simple, actionable guidance to improve spending, bill discipline, and savings habits. 3) Product relevance problem: Cross-sell is often generic (broadcast campaigns). Low relevance reduces conversion and trust. 4) Retention problem: Without proactive insights, customers churn or become inactive after onboarding. 5) CASA growth problem: Customers don’t get “safe-to-save” guidance that helps them keep balances higher and more stable.

Our solution (what we built)

We built an AI Personal Financial Coach embedded in SOL with a daily habit loop: Qwen 3.5 was use as the LLM AI

1) Daily coaching loop (DAU/MAU driver)

  • Insight feed on SOL home: one personalized insight/day
  • AM/PM nudges (configurable): “Today’s safe-to-spend” and “Log + reflect”
  • Streaks and progress feedback to reinforce behavior

2) Personalized guidance (CSAT/NPS driver)

  • Qwen-powered conversational guidance in plain language
  • “One next action” design: the coach always ends with a small, doable step
  • Safety guardrails: no promises of approval, no harmful/illegal advice

3) Explainable readiness / financial health (trust + retention)

  • Readiness band (Low/Medium/High) based on user behavior signals
  • Top drivers shown clearly (e.g., “Bills paid on time 2 months in a row”)
  • Improvement plan: 1–2 recommended actions to move up a band

4) Contextual cross-sell engine (conversion driver)

Instead of generic campaigns, SOL shows offers only when a user’s context indicates a need:

  • Cards: bill autopay, rewards aligned to spend categories
  • Loans: short-term cashflow smoothing (when shortfall risk is detected)
  • Insurance: protection offers triggered by life-stage signals (when available) All offers include a transparent “Why this offer?” explanation and deep-link into existing SOL flows.

5) Savings nudges (CASA driver)

  • Micro-goals and scheduled deposit nudges
  • “Safe-to-save” recommendations based on recent spending stability
  • Deposit streaks and progress visualization

Why it matters (measurable bank outcomes)

This MVP is designed to deliver:

  • DAU/MAU uplift: daily insight feed + nudges increase app-open frequency
  • Cross-sell conversion uplift: context-based offers increase CTR and application starts
  • Retention uplift / churn reduction: proactive reminders + weekly summaries reduce drop-off
  • CSAT/NPS improvement: personalized guidance improves perceived value and trust
  • CASA growth: savings nudges and behavior coaching stabilize balances

What we learned

  • Trust is the product: consent-first, explainable insights beat opaque scoring
  • Low-literacy UX wins with simplicity: fewer actions, clearer language, repeated habits
  • Cross-sell must be contextual: relevance + explanation drives conversion without harming trust
  • Metrics must be designed into the product: every insight and offer is instrumented end-to-end

Built With

  • amplitude/firebase)
  • analytics-event-tracking-(e.g.
  • backend-apis-(node.js/typescript)
  • openapi
  • postgresql
  • qwen-llm
  • redis
  • sol-mobile-app-(android-first)
Share this project:

Updates