Website Crawling Enhancements

• The website scanner now uses the new crawl engine that powers Briink's batch screenings, additionally the crawl depth parameter has been re-added to URL crawl jobs, giving you more control over how deeply the system crawls websites during document collection.
• Enhanced web crawler with better detection of blocked requests for more reliable document fetching

Enhanced Multi-Choice Question Processing

• Multi-choice questions in the EQA API now use structured output prompts, which improves the consistency and reliability of responses. The system also includes automatic retry logic when processing multi-choice answers, reducing failed responses due to validation errors.
• Fix multi-choice questions annotation formatting issue

Table Question Processing

• Improved table processing to handle special characters that could cause data corruption
• Improved AI system reasoning for complex ESG table type questions

API Updates

Screening Exports

• You can now specify custom export_columns when creating screening exports, giving you control over which fields appear in your exported data.

Table Processing

• Improved accuracy for table-based questions with better handling of fixed-row structures and dynamic tables.
• Table responses now maintain proper row alignment with reasoning, reducing data mismatches in complex table extractions.
• Inline references are now automatically stripped from list responses for cleaner output.

Document Collection

• Single documents collected via crawler are now properly detected and labeled, improving tracking accuracy.

Screening Entities

• You can now define screening entities to scope questionnaire analysis to specific companies or organizational units, improving answer relevance and accuracy.

Entity Scope for Questions

• Questions now support entity scope, allowing context-aware retrieval and answer generation tailored to specific entities within your documents.

*Active for all API users

We made a major leap in questionnaire importer quality, extraction are cleaner, more accurate, and more consistent across real world documents

What’s to expect?

• Much higher extraction accuracy and output consistency
• Multi-sheet excel support
• Expanded file support: .xls, .pdf, .csv, .docx, .pptx, .txt, .md, and .html

File Handling Improvements

• XML file parsing is now supported for questionnaire imports.
• Direct file uploads for questionnaire templates with configurable size limits and improved MIME type validation.

Screening enhancements

• You can now view the credit cost of screening operations after running them
• You can now redo document collection for an entire screening or individual tasks
• Questionnaire responses now include overall reasoning for multi-choice questions

Company entity updates

• Company creation now supports passing a custom ID and location parameters

Performance improvements

• Fixed multiple N+1 query issues across screening, batch operations, and company lookups for faster response times
• Resolved race conditions in screening operations for improved reliability
• Fixed the question title refinement feature stalling issue

Table Question Processing

• Reduced response latency for complex table questions
• Improved handling of table-based questions to ensure cell outputs conform to structured question types
• Improved model to further fine-tuned language model for common ESG domain table questions

Enhanced Document Collection

• The web crawler now uses an improved domain specific anchor-aware strategy to better identify and prioritize relevant ESG documents on company websites. This improves the accuracy and completeness of automated document discovery during screening tasks.

We've radically improved how tables are processed—delivering smarter, more consistent results you can trust.

AI That Truly Understands Your Tables

Our new table processing engine doesn't just fill in cells—it reasons through your table questions holistically. Briink's system now analyzes your documents, understanding context, relationships, and logical dependencies across rows and columns to deliver coherent, accurate results.

What's New:

• Contextual Awareness – The AI now considers the full table structure when processing each cell, ensuring answers are logically consistent across related fields!
• Much faster! – Tables can now be processed much with much lower overall latency, meaning you get results faster!
• Structured Reasoning – Every cell now includes transparent reasoning, so you can see exactly how the AI arrived at each answer, as well as full auditability and confidence in the granular results
• Conditional Logic – The new table system is now also able to correctly interpret conditional logic and dependencies implicitly defined by the meanings of the columns to improve consistency further.
• Graceful Edge Cases – Complex tables with varying structures, missing data, or ambiguous inputs are handled intelligently rather than failing silently
• Dynamic Row Management – Need to add vendor records, update line items, or remove outdated entries? New API endpoints let you add and delete table rows on the fly—with full support for choice-based columns and validation to keep your data clean.

That's not all!

Questionnaires now support a native Date question type with intelligent parsing. Extract and validate dates from documents automatically—whether they're full dates or year-only values.

Team Collaboration on Screenings

Work together more effectively with new collaboration features that help you manage, collaborate and share when working in teams on complex ESG questionnaires:

• Assignments – Assign team members to specific responses for clear ownership
• Comments – Discuss and collaborate directly on screening answers
• Due dates – Set deadlines and track completion timelines