OSL extraction
When you upload an Ontario Standard Lease, our AI extracts key information automatically. This page explains what gets extracted and how.
Extracted fields
Landlord information
| Field | Source in OSL |
|---|---|
| Name | Section 1 - Landlord(s) |
| Address | Section 3 - Address for Giving Notices |
| Section 3 - Email | |
| Phone | Section 3 - Phone |
Tenant information
| Field | Source in OSL |
|---|---|
| Name(s) | Section 2 - Tenant(s) |
| Section 2 - Email | |
| Phone | Section 2 - Phone |
Property information
| Field | Source in OSL |
|---|---|
| Address | Section 2 - Rental Unit |
| Unit Number | Section 2 - Unit (if applicable) |
| Parking | Section 2 - Parking Space |
Lease terms
| Field | Source in OSL |
|---|---|
| Start Date | Section 4 - Tenancy Start Date |
| End Date | Section 4 - End Date (fixed-term) |
| Lease Type | Section 4 - Fixed-term or Month-to-month |
| Base Rent | Section 5 - Lawful Rent |
| Rent Due Date | Section 5 - Due Date |
| Payment Method | Section 5 - Payment Method |
Services and utilities
| Field | Source in OSL |
|---|---|
| Electricity | Section 6 |
| Heat | Section 6 |
| Water | Section 6 |
| Internet | Section 6 |
| Parking | Section 6 |
| Air Conditioning | Section 6 |
Deposits
| Field | Source in OSL |
|---|---|
| Rent Deposit | Section 7 |
| Key Deposit | Section 7 |
Extraction technology
The extraction pipeline uses a two-provider AI chain for reliability:
- Anthropic (primary) — First-choice provider for extraction
- Google Gemini (fallback) — Used if Anthropic fails or times out
There is no OpenAI provider in the extraction chain.
The system also performs:
- PDF text parsing — For text-based PDFs
- OCR — For scanned documents
- Clause legality scanning — Automatically flags illegal lease clauses per the RTA
Progress tracking
A real-time progress UI shows the extraction pipeline's stage (Preparing, Uploading, Analyzing, Extracting, Validating) rather than per-field or per-section progress.
Confidence scores
Each extracted field has a confidence score:
High confidence (80-100%)
- Field clearly readable
- Standard format recognized
- Likely correct
Medium confidence (60-79%)
- Field readable but unusual format
- Some ambiguity
- Review recommended
Low confidence (below 60%)
- Field difficult to read
- Non-standard format
- Manual verification required
Handling edge cases
Handwritten text
- Printed text extracts best
- Handwritten may need OCR
- Very poor handwriting may fail
Multiple tenants
- All tenant names are extracted
- Each becomes a separate tenant record
- All linked to the same lease
Non-standard formats
If your OSL has been modified:
- Core fields usually still extract
- Custom additions may be missed
- Review carefully and add manually
Edit and regenerate
After extraction, you can edit any field and regenerate the OSL:
- Review extracted data on the OSL preview
- Click Edit Details to modify any field
- Click Regenerate to produce an updated OSL with your changes
- The regenerated OSL reflects your corrections
Comparing to original
You can compare extracted data to the original PDF:
- View extracted data
- Click "View Original"
- Side-by-side comparison
- Correct any discrepancies
If you later edit the lease's rent, dates, or other key fields, the generated PDF may become out of date. See Stale OSL detection below.
Stale OSL detection
When you edit lease terms (rent amount, start date, due day, or other OSL-relevant fields) after a PDF has already been generated, the system automatically marks the cached document as out of date.
An amber "Standard Lease is out of date — regenerate" banner appears on the lease Overview tab whenever the stored PDF no longer reflects the current lease terms.
Downloading a stale OSL will serve the old PDF. Always regenerate before sending the document to tenants.
How to regenerate
- Open the lease and go to the Overview tab
- Click Regenerate Standard Lease (available on the Overview tab for unsigned and declined leases)
- The new PDF is generated and the out-of-date banner clears
Which leases can be regenerated?
| esig status | Can regenerate? |
|---|---|
| None (unsigned) | Yes |
| Declined | Yes |
| Sent / Signed | No — editing terms on a sent/signed lease does not mark it stale |
Regenerating resets the stale flag and stores a fresh PDF. The prior document is replaced.
Improving extraction
Before uploading
- Use clean, high-quality scans
- Ensure all pages included
- 300 DPI or higher recommended
After uploading
- Review all fields
- Check medium/low confidence items
- Correct errors before confirming
Data validation
The system validates extracted data:
- Dates are valid
- Rent amount is reasonable
- Postal codes are valid format
- Email addresses are valid format
Invalid data is flagged for review.