AI Financial Coach

Managing money shouldn’t require a finance degree. During HackTX, we kept hearing the same story: people feel overwhelmed by fragmented bank apps, spreadsheets, and advice that isn’t tailored to their reality. We set out to build a simple, compassionate AI coach that speaks in plain language, highlights what matters this month, and nudges users toward better decisions without judgment.

What it does: Guides users through onboarding to set goals and a monthly budget. Shows a clean dashboard with current-month spending, over/under status, and category breakdowns. Generates AI recommendations that reflect the latest transactions and budget in real time. Surfaces potential stocks to buy/sell as an educational add-on, not financial advice.

How we built it: Frontend: React with a simple view state machine (Onboarding → Goal Setter → Dashboard). Data flow: Lightweight state and derived selectors to keep “top chips” and AI recommendations in sync. AI layer: Deterministic, explainable logic for monthly insights, with room to plug in an LLM later. UX details: Snappy updates on every relevant change; consistent, accessible components.

What we learned: Users trust insights that are consistent across the UI. If the “over” chip says $173, the AI should say $173—no surprises. Small UX decisions (e.g., removing refresh buttons and using live recompute) reduce friction and confusion. React hooks must be ordered deterministically; avoiding conditional hooks prevents hard-to-debug lint/runtime issues.

Challenges: Keeping derived metrics perfectly aligned across components without duplicating logic. Ensuring recommendations recompute on every relevant change without performance regressions. Balancing clarity with brevity in AI explanations so they’re helpful, not overwhelming. Handling edge cases (no transactions yet, partial months, category gaps).

What’s next: Plug-in real financial data sources with user consent and robust privacy controls. Add goal-driven simulations (e.g., “What if I reduce dining by 10%?”). Introduce an explainability panel showing how each recommendation was computed. Optional LLM integration for conversational “why” and “how” follow-ups, grounded in our transparent math. In short, this project turns messy financial snapshots into timely, trustworthy guidance that helps users make one good decision at a time.