Client

VTS is a leading commercial real estate (CRE) platform that transforms how landlords and brokers manage properties and deals. The platform centralizes critical data and workflows, enabling real estate professionals to make faster, more informed decisions about their portfolios.

Trusted by major real estate firms globally, VTS manages billions of square feet of commercial space and facilitates deal-making for some of the world’s most valuable properties. As the company continues to innovate, VTS focuses on eliminating friction in the deal-making process through intelligent automation.

VTS has maintained a long-term partnership with Ajackus, collaborating on various initiatives to enhance platform capabilities and accelerate feature development.

Challenge

Creating proposal documents in VTS was a significant pain point for commercial real estate professionals, directly impacting their ability to respond quickly to opportunities.

The critical obstacles included:

• Time-intensive data entry requiring several minutes to manually fill extensive multi-part forms
• Repetitive manual work copying information from existing PDF proposals into the platform
• Slow response times to potential tenants, impacting competitive advantage
• High error risk from manual data transcription across complex form fields
• User friction discouraging adoption of the proposal creation workflow

Real estate professionals often had existing proposal documents in PDF format but faced significant manual work to recreate them in VTS. This bottleneck needed an innovative solution that could maintain data accuracy while dramatically accelerating the process.

Goals

The project focused on transforming the proposal creation experience through intelligent automation while maintaining user trust and control.

The main goals included:

• Eliminating manual data entry by automating extraction from existing PDFs
• Reducing creation time from several minutes to seconds
• Maintaining data accuracy and giving users full visibility into AI-populated fields
• Creating intuitive UX patterns for AI-assisted workflows
• Seamlessly integrating with existing VTS data and proposal infrastructure
• Driving user adoption through demonstrable time savings and ease of use

Journey

Ajackus collaborated closely with VTS developers to rapidly build a reliable, user-friendly AI feature that transformed a complex manual process into a seamless automated experience.

Phase 1: Proof of Concept

• Validating AI accuracy with real-world proposal PDFs
• Testing data extraction reliability across various formats
• Establishing baseline performance metrics
• Gathering initial user feedback on the concept

Phase 2: MVP Development

Collaborative Development Approach

Working alongside VTS’s internal development team, Ajackus engineers brought specialized expertise in AI integration and novel UI patterns, accelerating development velocity and bringing the feature to market quickly.

Backend & Data Processing

• Building robust PDF parsing pipeline
• Implementing LLM integration for unstructured data extraction
• Creating data transformation layer to map AI output to structured form fields
• Developing validation logic to ensure data accuracy
• Handling edge cases and various PDF formats

Frontend & User Experience

• Designing novel UI patterns for AI-assisted data entry
• Implementing visual indicators showing which fields were AI-populated
• Creating bespoke loading animation to provide engaging feedback during processing
• Building review interface for users to validate and edit AI suggestions
• Handling complex form interactions with both manual and AI-populated data

Phase 3: Technical Challenges Solved

Data Transformation Architecture

The most significant technical challenge was bridging the gap between unstructured LLM output and VTS’s structured proposal form. The combined team developed a sophisticated transformation layer that:

• Maps free-form AI-extracted data to specific form field requirements
• Handles data type conversions and formatting
• Validates against existing VTS data schemas
• Manages confidence scoring for extracted information
• Provides fallback mechanisms for ambiguous data

Displaying Unstructured Data

Presenting AI-extracted information required innovative UI solutions:

• Dynamic field highlighting to distinguish AI-populated vs. manual entries
• Inline confidence indicators for uncertain extractions
• One-click acceptance or editing of suggested values
• Contextual hints showing the source location in the original PDF
• Graceful degradation when AI cannot confidently extract certain fields

User Experience Innovation

The team created a seamless experience that balanced automation with user control:

• Real-time progress feedback during PDF processing
• Custom loading states that communicate AI processing stages
• Intuitive review workflow that focuses attention on fields needing validation
• Preservation of user edits during subsequent AI-assisted updates
• Clear differentiation between high-confidence and uncertain extractions

Phase 4: Continuous Enhancement

• Iterating based on user feedback and adoption metrics
• Implementing visual indicators showing which fields were AI-populated
• Adding support for new proposal types and formats
• Refining AI accuracy through ongoing training
• Expanding feature capabilities based on user requests

Results

The collaborative development of Proposal AI has transformed how VTS users create proposals, delivering dramatic time savings and driving strong user adoption.

What went well:

Operational Improvements

• 50%+ time reduction in proposal creation (several minutes → seconds)
• 100% automation of data extraction from PDF documents
• Seamless integration with existing VTS proposal workflows
• Zero disruption to users who prefer traditional manual entry

Technical Achievements

• Successfully processed diverse PDF formats with high accuracy
• Built novel UI patterns for AI-assisted data entry now serving as model for future features
• Created robust transformation layer handling complex data mapping scenarios
• Implemented intelligent validation ensuring data quality and accuracy

User Impact

• Strong and increasing adoption rate with high user enthusiasm
• Exceptional user satisfaction with consistent positive feedback
• Reduced friction in proposal creation workflow
• Competitive advantage for VTS users through faster response times

Development Success

• 2 specialized Ajackus engineers collaborated with VTS team to deliver full feature
• Rapid development from proof of concept through MVP to full feature
• Accelerated time-to-market through combined expertise
• Ongoing enhancement with continuous feature additions and improvements
• Knowledge transfer enabling VTS to extend the feature independently

Key Technologies

Development Stack

• Frontend: React, TypeScript
• Backend: Ruby on Rails
• AI/ML: Large Language Model (LLM) integration
• Data Processing: Custom PDF parsing and extraction pipeline

Specialized Solutions

• Custom data transformation layer for LLM output
• Novel UI component library for AI-assisted workflows
• Bespoke loading and progress visualization system
• Intelligent field mapping and validation framework

Key Takeaways

Why This Collaborative Development Succeeded

1. Combined Expertise
   1. Ajackus brought specialized AI integration and novel UI pattern expertise
   2. VTS provided deep domain knowledge and platform understanding
   3. Collaboration accelerated development velocity and quality
2. User-Centric AI Design
   1. Prioritized user control and transparency over full automation
   2. Visual indicators built trust in AI-generated suggestions
   3. Maintained flexibility for both AI-assisted and manual workflows
3. Phased Approach
   1. POC validated concept before major investment
   2. MVP gathered real-world feedback early
   3. Iterative enhancements based on actual usage patterns
4. Technical Innovation
   1. Solved complex data transformation challenges
   2. Created reusable UI patterns for future AI features
   3. Balanced accuracy with performance
5. Integration Excellence
   1. Seamlessly fit into existing VTS workflows
   2. Leveraged existing data and infrastructure
   3. No disruption to current users while adding powerful new capability

Lessons Learned

UX Insights

• Users trust AI more when they can see what was automated
• Loading states are critical for AI processing experiences
• Providing escape hatches to manual editing increases adoption

Technical Insights

• Data transformation layer is crucial for LLM-to-form mapping
• Handling diverse PDF formats requires robust parsing strategy
• Complex form interactions need careful state management with AI

Product Insights

• Saving minutes per proposal drives immediate adoption
• Users enthusiastically adopt features that eliminate tedious work
• Continuous enhancement based on feedback maintains momentum

Collaboration Insights

• Specialized expertise accelerates feature development
• Close collaboration with internal teams ensures platform alignment
• Knowledge transfer creates lasting value beyond the initial feature