AI Data Extraction
Let AI automatically pull data from your documents.
How It Works
BuildVision uses AI to read documents and extract structured data:
- Upload - Add invoice, receipt, or quote
- OCR - AI reads the text
- Extraction - AI identifies key fields
- Review - You verify and save
Works on PDFs, images, and scanned documents.
What Gets Extracted
From Invoices
| Field | Example |
|---|---|
| Vendor Name | ABC Lumber Supply |
| Invoice Number | INV-2024-0542 |
| Invoice Date | March 15, 2024 |
| Due Date | April 15, 2024 |
| Total Amount | $3,247.50 |
| Line Items | Materials, quantities, prices |
| Tax | $267.50 |
From Quotes
| Field | Example |
|---|---|
| Supplier | XYZ Electrical |
| Quote Number | Q-12345 |
| Valid Until | April 1, 2024 |
| Items | Products and pricing |
| Total | $8,500.00 |
From Receipts
- Store/vendor name
- Date
- Items purchased
- Total amount
- Payment method
Running Extraction
Select Document
Click on any invoice, quote, or receipt in the Document Center.
Click Extract Data
Click Extract Data button or the AI icon.
Wait for Processing
AI processes the document (usually 10-30 seconds):
- OCR reads all text
- AI identifies fields
- Confidence scores calculated
Review Results
Extracted data appears in a panel:
- Green = high confidence
- Yellow = needs review
- Red = could not extract
Edit if Needed
Click any field to correct it:
- Type the correct value
- AI learns from corrections
- Better accuracy over time
Save
Click Save to store the extracted data:
- Creates/updates invoice record
- Links to source document
- Ready for payment tracking
Accuracy & Confidence
AI shows confidence for each field:
| Confidence | Meaning | Action |
|---|---|---|
| High (90%+) | Very sure | Usually correct |
| Medium (70-90%) | Probably right | Quick verify |
| Low (under 70%) | Uncertain | Manual check needed |
AI accuracy improves over time as it learns your vendors' invoice formats.
Line Item Extraction
For invoices with multiple items:
- Each line item extracted separately
- Quantity, description, unit price, total
- Auto-categorized when possible
- Link to inventory items
Example Output
Line Items:
1. 2x4x8 Studs (50) @ $4.25 = $212.50
2. OSB Sheathing (10) @ $28.00 = $280.00
3. Roofing Nails (2 boxes) @ $15.00 = $30.00
Subtotal: $522.50
Tax: $43.11
Total: $565.61Batch Processing
Process multiple documents at once:
- Select multiple invoices (checkbox)
- Click Extract All
- AI processes in background
- Review each when complete
- Bulk approve accurate ones
Batch processing runs in the background - you can work on other things while it runs.
Creating Records
After extraction, data can become:
Invoice Records
Extracted invoices create entries:
- In your Incoming Invoices list
- With payment tracking
- Linked to projects
Expense Tracking
Line items can:
- Update project budgets
- Track material costs
- Compare to estimates
Improving Accuracy
Tips for Better Results
- Clear scans - Higher resolution = better OCR
- Standard formats - Typed invoices work best
- Good lighting - For photos of receipts
- Flat documents - No wrinkles or folds
Training the AI
The more you use it:
- AI learns vendor formats
- Familiar invoices extract faster
- Correction patterns improve future accuracy
Supported Documents
| Document Type | Extraction Quality |
|---|---|
| PDF invoices | Excellent |
| Scanned invoices | Good |
| Photo receipts | Good |
| Handwritten | Limited |
| Foreign languages | Varies |
Privacy & Security
Your document data is:
- Processed securely
- Not used to train external models
- Stored encrypted
- Access controlled by permissions
Troubleshooting
Q: Nothing was extracted
Document may be:
- Image quality too low
- Format not recognized
- Heavily stylized/decorated
Try rescanning at higher resolution.
Q: Wrong vendor detected
AI uses text recognition. If logo-based:
- Edit the vendor name
- AI remembers for next time
Q: Missing line items
Some invoices have unusual formats:
- Add line items manually
- Or edit extracted data
Next Steps
- Incoming Invoices - Track payments
- Document Types - Organize by category
- Search - Find extracted data