ClearMint — AI Personal Finance & Budget App
Your money, explained in plain English every morning
Project Overview
A fintech startup wanted to build a budgeting app that went beyond manual category tags and static pie charts. ClearMint connects to bank accounts via Open Banking APIs, classifies every transaction automatically, builds a dynamic budget that learns from actual behaviour, and sends a daily 'money briefing' — a three-sentence plain-English summary of yesterday's spending and what it means for the week ahead.
The Challenges
- 1
Transaction descriptions from banks are notoriously cryptic ('SQ *COFFEE HOUSE 4892') — a generic classifier needed fine-tuning on real bank data.
- 2
Static budgets break immediately — if rent goes up, the whole budget is wrong. The system needed to auto-update budget lines from detected recurring changes.
- 3
Users open finance apps once then abandon them; the daily briefing had to be genuinely insightful, not generic, to drive retention.
- 4
Open Banking API coverage varies by country and institution — the sync layer needed graceful fallback for unsupported banks.
Our Approach
The transaction classification model is a fine-tuned DistilBERT trained on 4.2 M labelled bank transactions across UK, US, and EU institutions. It achieves 94% accuracy on unseen descriptions and is updated monthly with user-corrected labels (corrections that disagree with predictions are the highest-signal training examples). A rule-based recurring-payment detector identifies subscriptions and regular billers, auto-populates budget fixed costs, and alerts on price changes. The daily briefing is generated by GPT-4o Mini given a structured context object: yesterday's spend by category, budget vs. actual, any anomalies, and the user's stated financial goals. It is intentionally short — three sentences that are genuinely personalised.
Key Features & Metrics
94% transaction classification accuracy across UK, US, and EU bank descriptions
Automatic recurring-payment detection and budget pre-population
Daily AI money briefing: 3 personalised sentences generated every morning
Savings goal tracker with AI-calculated contribution targets based on spending gaps
Spending anomaly alerts: 'This is 40% higher than your usual grocery spend this week'
Open Banking integration for 3,400+ institutions with CSV import fallback
Results & Business Outcome
Users saved an average of 23% more in the 90 days post-adoption compared to the 90 days prior. Day-30 retention hit 58% — 3× the personal finance app category average. The app achieved a 4.9★ App Store rating from its first 1,000 reviews.
Financial stress comes from uncertainty, not from spending. When people understand exactly where their money goes — in plain English, every morning — they make better decisions naturally.
Let's Build Something Intelligent Together
Tell us about your project. We'll respond within 24 hours with a custom plan and transparent pricing.