AI Document Extraction: Streamlined Data Entry

TenancyDesk AI extracts data from documents automatically - cutting down manual data entry and speeding up deal creation.

Technology

OCR Engine: AWS Textract + custom UAE document training

Coverage: Tuned for common UAE documents — Emirates ID, UAE passports, Title Deeds (DLD), Trade Licenses, and DEWA bills. Every extracted field carries a confidence score so you can review and correct low-confidence values before saving.

Processing speed: a few seconds per document

---

Supported Documents

1. Emirates ID (Front & Back)

Extracted fields (12):

• Full name (Arabic + English)
• ID number (784-XXXX-XXXXXXX-X)
• Nationality
• Date of birth
• Gender
• Issue date
• Expiry date
• Card number
• Address (from back side)

Confidence scores: Field-level confidence (0-100%)

Example:

{
  "document_type": "emirates_id_front",
  "extracted_data": {
    "name": "Omar Mohammed Khalid",
    "id_number": "784-1988-1234567-1",
    "nationality": "United Arab Emirates",
    "date_of_birth": "1988-05-15",
    "expiry_date": "2028-12-31",
    "confidence_score": 99.2
  },
  "verification_status": "verified"
}

Auto-fill: Name, ID number, DOB, nationality populate deal form instantly

2. Passports

Extracted fields (10):

• Full name
• Passport number
• Nationality
• Date of birth
• Issue/expiry dates
• MRZ (Machine Readable Zone) data

Visa page detection: Automatically identifies UAE visa page, extracts:

• Visa number
• Visa type (residence, visit, tourist)
• Visa expiry date

3. Title Deeds (Dubai Land Department)

Extracted fields (8):

• Owner name (must match landlord exactly)
• Plot number
• Building number
• Unit number
• Property type
• DLD registration number
• Issue date

Verification: Cross-check owner name with landlord in deal

4. Trade Licenses

Extracted (6):

• Company name
• License number
• Issuing authority (DED, free zone)
• Activity type
• Expiry date

5. DEWA Bills

Extracted (5):

• Premise number (8 digits)
• Account holder name
• Building/unit number
• Billing month

---

How It Works

Automatic Extraction Workflow

1. Document Upload:

• Agent or tenant uploads via Deal Portal
• Supported formats: PDF, JPG, PNG, HEIC
• Max size: 10 MB

2. AI Processing (2-5 seconds):

• Image preprocessing (deskew, enhance, denoise)
• OCR text extraction
• Field identification (name, numbers, dates)
• Data structuring (JSON format)

3. Validation:

• Format checks (Emirates ID = 15 digits with dashes)
• Expiry date checks (not expired, >30 days remaining)
• Name matching (tenant name vs Emirates ID name)
• Confidence thresholds (reject if <85% confidence)

4. Auto-fill Deal Form:

• Extracted data populates corresponding fields
• Agent reviews and confirms
• Override if AI made error

5. Status Indicators:

• 🟢 Verified (99%+ confidence, all checks passed)
• 🟡 Review Required (85-99% confidence or minor issues)
• Rejected (< 85% confidence or major issues)

---

Manual Review Workflow

When AI confidence <99%:

1. Agent notification: "Emirates ID extraction: 92% confidence - please review"
2. Side-by-side view:
   • Left: Original document image
   • Right: Extracted data with confidence scores
3. Agent actions:
   • Confirm (if correct)
   • ️ Edit (fix errors)
   • Reject (re-upload needed)

Learning: Agent corrections train AI for future accuracy

---

Privacy & Security

PII Protection:

• Extracted data encrypted at rest (AES-256)
• Encrypted in transit (TLS 1.3)
• Access logged (audit trail)

Data retention:

• Extracted data: 7 years (UAE law requirement)
• Original images: 7 years
• Extraction logs: 90 days

GDPR/PDPL aligned: Right to access, deletion, rectification (per Article 15-17 + UAE PDPL principles)

---

AI extraction significantly reduces manual data entry and the errors it introduces. No setup required - it just works.