Ajackus partnered with VTS—a leading commercial real estate platform managing billions in assets—to build Proposal AI, an intelligent feature that transforms PDF proposals into structured, editable data in seconds.
Engagement Model
Team Augmentation
Your team, supercharged — every Ajackus engineer comes AI-native, shipping 40–60% faster from day one
Managed Delivery
Tell us the problem. We handle everything — discovery, architecture, build, testing, deployment — at AI speed
BOT Model
We build your offshore engineering center with AI-trained teams, operate it to maturity, then transfer ownership to you
Expertise
AI Development
Custom AI agents, LLM integrations, RAG pipelines, and AI enablement — from strategy to production
Web Development
Enterprise web applications built with AI-augmented engineering for 40-60% faster delivery
Mobile Development
Native and cross-platform mobile apps built at AI pace with rigorous automated quality
DevOps
Infrastructure that scales automatically, deploys safely, and costs less
Low-Code/No-Code
Airtable, Glide, Shopify, and WordPress solutions with AI automation layers — built in days
Data Analytics
Turn raw data into decisions that drive revenue — 80% faster with AI-assisted analytics
UI/UX Design
User-centered interfaces designed with AI-powered speed — 3x faster iteration, sharper insights
How VTS Cut Proposal Creation Time by 50%+ with AI-Powered Document Automation
Services
Generative and Agentic AI Development
Web Development
UI/UX Design
Team Augmentation
Technologies
50% +
Reduction in Proposal Creation Time
100%
Automated PDF Data Extraction
2
Ajackus Engineers Embedded with VTS
Overview
Executive Summary
Client Overview
The Challenge
Project Goals
Our Approach
Results and Impact
Frequently Asked Questions
Executive Summary
The Problem
VTS, a leading commercial real estate platform, faced a critical bottleneck in proposal creation. Real estate professionals spent several minutes manually entering data from existing PDF proposals into multi-part forms—slowing deal velocity and increasing error rates.
The Solution
Ajackus embedded two specialised engineers with the VTS product team to design and ship Proposal AI, an LLM-powered feature that parses PDF documents, extracts structured data, and auto-populates proposal forms with full user visibility and control.
The Result
Proposal creation time dropped from several minutes to seconds (50%+ reduction), user adoption surged, and the UI patterns created now serve as a reusable blueprint for future AI features across VTS.
Client Overview
VTS is a market-leading commercial real estate platform that transforms how landlords and brokers manage properties and close deals. The platform centralises critical data and workflows, enabling real estate professionals to make faster, data-driven 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. VTS has maintained a long-term development partnership with Ajackus, collaborating across multiple initiatives to enhance platform capabilities and accelerate feature delivery.
| Industry | Commercial Real Estate (PropTech) |
| Platform Scale | Billions of sq. ft. under management |
| Headquarters | New York, NY |
| Partnership | Long-term collaboration with Ajackus |
The Challenge
THE BOTTOM LINE
VTS needed a solution that could take an existing PDF proposal and instantly transform it into a structured, editable proposal inside the platform—without sacrificing accuracy or user control.
Creating proposal documents in VTS was a significant pain point for commercial real estate professionals, directly impacting their ability to respond quickly to leasing opportunities and close deals ahead of competitors.
Core Pain Points
- Time-intensive data entry: Professionals spent several minutes manually completing extensive, multi-part proposal forms for every new deal.
- Repetitive copy-paste workflows: Brokers and landlords already had existing proposals in PDF format but had to manually recreate every data point inside VTS.
- Competitive disadvantage from slow response times: In a market where speed wins deals, minutes of delay per proposal meant lost tenants.
- High error rates from manual transcription: Complex form fields across multiple proposal sections increased the risk of costly data entry mistakes.
- Adoption friction: The tedious workflow discouraged users from fully leveraging VTS’s proposal creation capabilities.
Project Goals
The project was designed to transform the proposal creation experience through intelligent automation while maintaining user trust, data accuracy, and seamless integration with VTS’s existing infrastructure.
| Goal | Success Criteria |
|---|---|
| Eliminate manual data entry | 100% automated extraction from uploaded PDFs |
| Reduce proposal creation time | From several minutes to seconds (50%+ reduction) |
| Maintain data accuracy | Full user visibility into AI-populated fields with edit controls |
| Intuitive AI-assisted UX | Novel UI patterns that build trust through transparency |
| Seamless platform integration | Zero disruption to existing workflows and manual entry paths |
| Drive user adoption | Measurable uptake driven by demonstrable time savings |
Our Approach
Ajackus embedded two specialised engineers directly within VTS’s product team, combining deep AI integration expertise with VTS’s domain knowledge to move from proof of concept to production-ready feature at speed.
Phase 1: Proof of Concept
The engagement began with a focused validation phase. The combined team tested AI extraction accuracy against real-world proposal PDFs across diverse formats, established baseline performance metrics, and gathered early user feedback to confirm the concept’s viability before committing to full development.
Phase 2: Collaborative MVP Development
With the concept validated, development moved into three parallel workstreams:
Backend and Data Processing
The team built a robust PDF parsing pipeline integrated with a large language model (LLM) for unstructured data extraction. A custom data transformation layer was developed to map AI-extracted output to VTS’s structured proposal form fields, complete with validation logic and edge-case handling for varied PDF formats.
Frontend and User Experience
Rather than hiding the AI behind a “magic” button, the team designed novel UI patterns that gave users full visibility into what the AI had done. This included dynamic field highlighting to distinguish AI-populated entries from manual ones, inline confidence indicators for uncertain extractions, one-click acceptance or editing of suggested values, and a bespoke loading animation that communicated real-time processing stages.
Data Transformation Architecture
The most significant technical challenge was bridging unstructured LLM output and VTS’s structured form schema. The team built a sophisticated transformation layer that maps free-form AI data to specific field requirements, handles type conversions and formatting, validates against existing VTS data schemas, manages confidence scoring, and provides fallback mechanisms for ambiguous extractions.
Phase 3: Iteration and Enhancement
Post-launch, the team iterated continuously based on user feedback and adoption metrics—refining AI accuracy, adding support for new proposal types and formats, and expanding feature capabilities to meet emerging user needs.
Results and Impact
The collaborative development of Proposal AI delivered measurable business impact across operational efficiency, technical achievement, and user satisfaction.
50%+
Faster Proposal Creation
100%
Automated PDF Extraction
0
Disruption to Existing Users
Operational Improvements
- Proposal creation time reduced from several minutes of manual data entry to seconds of automated extraction and review.
- Complete end-to-end automation of data extraction from uploaded PDF proposals.
- Seamless integration with existing VTS proposal workflows—zero disruption for users who prefer manual entry.
Technical Achievements
- Successfully processed diverse PDF formats with high extraction accuracy across varied document structures.
- Created novel, reusable UI component patterns for AI-assisted data entry—now serving as the design blueprint for future AI features across VTS.
- Built a robust data transformation layer capable of handling complex mapping between unstructured LLM output and structured form schemas.
User Impact
- Strong and growing adoption rate with consistently positive user feedback.
- Reduced friction in the proposal creation workflow, enabling real estate professionals to respond faster to tenant enquiries.
- Knowledge transfer to VTS’s internal team, enabling independent extension and maintenance of the feature.
Why It Worked
Transparency Builds Trust
Users adopt AI faster when they can see what was automated and edit with one click. Never hide the AI—show it.
The Transform Layer Is Key
Bridging unstructured LLM output to structured forms is where AI projects stall. Investing here made the feature reliable at scale.
Embedded Teams Ship Faster
Embedding Ajackus engineers with VTS eliminated handoff delays and enabled knowledge transfer that outlasts the project.
Frequently Asked Questions
How does AI improve proposal creation in commercial real estate?
AI-powered document automation can parse existing PDF proposals, extract structured data using large language models, and auto-populate platform forms in seconds—eliminating manual data entry and reducing errors while keeping users in full control of the final output.
What technologies does Ajackus use for AI document automation?
For this engagement, the team used React and TypeScript on the frontend, Ruby on Rails on the backend, LLM integration for intelligent data extraction, and a custom-built PDF parsing pipeline with a bespoke data transformation layer for mapping unstructured AI output to structured form fields.
How long did it take to build and ship Proposal AI?
Ajackus embedded two specialised engineers with the VTS team to move from proof of concept through MVP to production-ready feature in a rapid, phased timeline—validating early, shipping incrementally, and iterating based on real user feedback.
Can AI document automation work with different PDF formats?
Yes. A key technical challenge was handling diverse PDF structures. The team built a robust parsing pipeline with fallback mechanisms and confidence scoring to ensure reliable extraction across varied document layouts and formats.
Let’s Build your Project Together
