How AI Categorization Works in WIMD

One of WIMD's most powerful features is automatic transaction categorization. Instead of manually tagging every purchase, our AI does it for you. But how does it actually work?

The Basics

When a new transaction comes in from your bank, WIMD's AI looks at several pieces of information:

• Merchant name – The description your bank provides (e.g., "STARBUCKS #12345")
• Transaction amount – Helps distinguish between similar merchants
• Your existing categories – The AI knows what options are available
• Your correction history – What you've recategorized in the past

How the AI Decides

The AI uses pattern matching and context to make decisions. Here's a simplified example:

For more ambiguous transactions, the AI considers multiple factors. A $50 charge at "AMAZON.CA" could be groceries, electronics, or household items. The AI makes its best guess based on patterns, but it may need your help.

Training the AI

Every time you correct a categorization, you're training the AI to work better for you. Here's how it learns:

1. Direct corrections – When you change a category, WIMD remembers that "MERCHANT X should be Category Y."
2. Pattern building – After a few corrections, similar transactions are automatically categorized correctly.
3. Rules override – If you create an explicit rule, it always takes priority over AI suggestions.

The Learning Curve

In your first week, you might correct 10-20% of transactions. By month two, that number drops significantly. By month three, the AI handles almost everything correctly—because it's learned your specific patterns.

When AI Isn't Enough: Rules

Sometimes the AI can't reliably categorize certain transactions. That's where rules come in. Rules are explicit instructions that always override the AI.

For example, if your side hustle income comes through as "E-TRANSFER FROM VARIOUS," the AI might not know it's income. Create a rule: "When description contains 'E-TRANSFER FROM' AND amount is greater than $500, categorize as Side Hustle Income."

When to Use Rules vs. Training

• Use AI training for standard merchants (coffee shops, restaurants, stores)
• Use rules for personal or ambiguous transactions (rent, transfers, side income)
• Use rules when you need 100% reliability for specific transactions

Confidence Levels

WIMD shows you how confident the AI is in its categorization. High confidence means the AI is very sure. Medium confidence means it made an educated guess. This helps you know which transactions to review.

Continuous Improvement

The AI isn't static. As you use WIMD, it gets better at understanding your spending patterns. New merchants, recurring payments, and seasonal purchases—over time, the system learns them all.

Your corrections also help improve categorization for other users (anonymously, of course). When hundreds of people correct the same merchant, the AI learns that pattern for everyone.

Summary

WIMD's AI categorization saves you hours of manual work. It's not perfect out of the box, but it learns quickly. Spend a few minutes correcting categories in your first week, set up rules for edge cases, and let the system do the rest.