AI Document Classification

Atlas uses AI to automatically classify documents when they're uploaded. This saves time and ensures consistent categorisation across your practice.

How It Works

1. Upload — a document is uploaded to an entity
2. Analysis — Atlas AI reads the document content (text extraction via OCR for images/scans)
3. Classification — the document is assigned a type
4. Confidence — a confidence score determines if you should review
5. Accept or Correct — you confirm or change the classification

Classification Types

Type	Keywords / Patterns
Financial Statements	"Statement of Financial Position", "Income Statement"
Tax Return	"ITR12", "ITR14", "VAT201", "SARS"
Bank Statement	Bank name, account numbers, transaction lists
Contract	"Agreement", "Terms and Conditions", signatures
ID Document	"Identity Document", ID numbers, photo identification
Registration Certificate	"CIPC", "Certificate of Incorporation"
Correspondence	Letters, memos, email printouts
Invoice	"Invoice", line items, amounts

Confidence Levels

Level	Action
High (above 90%)	Auto-classified — no action needed
Medium (60–90%)	Classified with a ⚠️ review badge
Low (below 60%)	Marked as "Unclassified" for manual review

Improving Classification

AI classification improves as you use it:

• Correct misclassifications to train the model
• Consistent document naming helps accuracy
• Clear, high-quality scans produce better OCR results

Manual Classification

If AI gets it wrong:

1. Click on the document
2. Click Reclassify
3. Select the correct type from the dropdown
4. The correction is saved for future reference