Document Intelligence: what's new in Q1 2026

The first quarter of 2026 has been our most productive release cycle for Document Intelligence. Every improvement we shipped this quarter came directly from teams managing large lease portfolios who told us what was slowing them down.

Here is a summary of everything that went live.

Faster extraction with parallel processing

The most requested improvement was speed. Teams uploading batches of 100 or more lease documents during initial onboarding were waiting too long for extraction to complete. So we rebuilt the extraction pipeline from the ground up to support parallel document processing.

The practical impact:

• A batch of 100 standard commercial leases that previously took 45 minutes now completes in under 12 minutes
• Individual document extraction is 30% faster due to optimized OCR preprocessing
• Progress indicators now show per-document status, so teams can begin reviewing completed extractions while the rest of the batch continues processing

These improvements are available to all customers automatically. No configuration changes required.

Improved table parsing for complex payment schedules

Lease agreements frequently contain payment schedules formatted as tables, and these tables vary enormously in structure. Some use merged cells. Others span multiple pages. Many include footnotes or conditional payment tiers that are critical to getting the data right.

Our previous parser handled standard formats well but struggled with these edge cases. The updated parser uses a layout-aware model that understands table structure at a visual level - similar to how a human reader interprets rows, columns, and groupings.

Key improvements:

• Accurate extraction of multi-page payment schedules that span page breaks
• Correct handling of merged header cells and nested sub-tables
• Recognition of escalation patterns within payment tables - annual increases, CPI adjustments, and step-rent structures identified automatically
• Support for tables embedded within narrative text, not just standalone table elements

In our internal testing across 2,000 lease documents, table extraction accuracy improved from 87% to 96%. The largest gains were on scanned documents and older PDF formats, which are exactly the documents that cause the most manual rework.

Twelve new supported document types

Document Intelligence now supports extraction from 12 additional document types beyond standard lease agreements. These cover the ancillary documents that teams frequently need to reference when building their lease portfolios:

• Lease amendments and modification letters
• Estoppel certificates
• Subordination, non-disturbance, and attornment agreements
• Commencement date confirmation letters
• Tenant improvement allowance agreements
• Right of first refusal and expansion option notices
• Guaranty agreements
• Lease termination agreements
• Rent deferral and abatement agreements
• Assignment and assumption agreements
• Sublease agreements
• Ground lease agreements

For each document type, the system extracts the relevant fields and links them to the associated parent lease in your portfolio. Modifications, amendments, and related documents are automatically connected to the correct lease record - reducing the manual filing and cross-referencing work that used to consume hours of every onboarding cycle.