← Blog

How AI matching works

ClaimStation can suggest matches between your bank transactions and your claims. Sometimes the rule-based engine is confident; sometimes it isn’t. For those uncertain cases, we offer optional AI-assisted matching. Here’s how it works and what it means for your data.

Two-step matching

Matching runs in two steps:

  1. Rule-based — We compare transactions to claims using description keywords, amounts, dates, and reimbursing entity names. When confidence is high, we suggest a match without using AI.

  2. AI-assisted — When confidence is medium, we can send an anonymised version of the relevant data to an AI provider (we use OpenAI) to improve the suggestion. The AI result is combined with the rule-based score. You still review and confirm every match; nothing is linked automatically.

Your data is protected

Before any data leaves ClaimStation we run it through an anonymisation step. We remove or replace:

  • Account numbers, card details, policy numbers, member IDs
  • Names, emails, phone numbers, addresses
  • Reimbursing entity names (mapped to generic types)

Only anonymised descriptions, amounts, dates, and category types are sent. The AI cannot identify you or your accounts.

Opt-in and reversible

AI matching is off by default. You must enable it in Settings → AI Features and consent before any data is sent. You can disable it at any time. Matching Intelligence (learning from your confirm/reject actions) is separate and stored only in your account — it is not sent to external services.

For more detail, see our Security page and the in-app AI-assisted matching guide.