Prosper learns your business. Forever.
Every correction teaches Prosper a pattern. Every pattern makes future categorization faster and more accurate. The longer you use Prosper, the more it knows about your specific business.
You correct a transaction
"AMZN*2K7B3X" is Office Supplies, not inventory.
Prosper remembers the pattern
Learns that AMZN* transactions for your business are office supplies.
Future transactions auto-categorize
Next time AMZN appears, Prosper handles it automatically.
150+
Avg patterns learned
~3 mo
Time to 90% accuracy
87%
Correction recall
Compounding accuracy
Month 1: 70% auto-categorized. Month 6: 90%+. Your patterns compound into a custom model.
Time savings that grow
Every correction saves you time on all future similar transactions. Early investment pays dividends.
Your knowledge, retained
Vendor mappings, category preferences, business context. All stored and used to improve suggestions.
Explainable, editable, deletable
See exactly what Prosper has learned. Edit patterns. Delete what does not work. You are in control.
AI Memory vs. Static Rules
| Feature | Prosper AI Memory | QBO / Xero Rules |
|---|---|---|
| Learning method | Semantic AI with embeddings | Exact-text rules |
| Pattern recognition | Fuzzy matching (AMZN* → Amazon) | Exact match only |
| Corrections | Learned automatically | Manual rule creation |
| Cross-vendor patterns | AI infers similar merchants | One rule per merchant |
| Business context | Uses your memories and history | No context awareness |
| Improvement over time | Compounding accuracy | Static rules |
Your learned patterns are an asset
After 6 months, Prosper knows your vendors, your category preferences, your edge cases. That knowledge took time to build. If you switch to another tool, you start from zero.
This is not lock-in. It is value. We show you exactly what Prosper has learned, and you can export your patterns anytime. But the intelligence is real, and it is yours.
How it works under the hood
- Corrections are stored with semantic embeddings
- Similar transactions recall past corrections (87% match rate)
- LLM uses corrections as few-shot examples in prompts
- High-confidence matches auto-confirm (configurable threshold)
What you can do
- View all learned patterns in Settings → AI
- Edit category assignments for any pattern
- Delete patterns that are not working
- Reset and re-learn from existing transactions
Ready to build your AI memory?
Start categorizing. Every correction makes Prosper smarter for your business.