AI System for Identifying Data Quality Issues in Donation Records

The problem

When donation forms, bank transactions, campaigns and receipts are held in separate sources, their fields must be matched by hand before a matching and data quality list can be prepared.

The donation operations specialist must compare every suggestion in the matching and data quality list with the source record; missing or conflicting evidence must not replace the final decision.

If an anonymous donation or partial refund is not reviewed separately, a donation could be linked to the wrong person; the record must not progress without a human decision.

What we set out to improve

• Time taken for the draft matching and data quality list to reach the donation operations specialist with its supporting evidence
• Outcomes of records referred for human review because an anonymous donation or partial refund was found
• Number of suggestions corrected after identifying the risk of linking a donation to the wrong person

How the system works

1. 01 Retrieve the donation form, bank transaction, campaign and receipt from authorised sources
2. 02 Verify field and document versions
3. 03 Prepare the matching and data quality list with its supporting evidence
4. 04 Exception check: an anonymous donation or partial refund is present
5. 05 Have the donation operations specialist accept, correct or reject the suggestion
6. 06 Decision gate: donation allocation, refund and receipt decisions
7. 07 Record the decision and source separately

Methods we used

• A source-identified data schema for the donation form, bank transaction, campaign and receipt
• A structure that separates deterministic rules from AI suggestions
• A screen showing the matching and data quality list beside its supporting evidence
• A human review queue when an anonymous donation or partial refund is present
• A record identifier that prevents duplicate processing

Where people stay involved

Data and security

Only the donation form, bank transaction, campaign and receipt fields required for the task should be processed. Role-based access should be applied, and unnecessary identity, health, financial and internal organisational data must not be sent to the model. Source retrieval, AI suggestions and the donation operations specialist's decisions should be recorded separately.

Who this suits

Frequently asked questions