OUR IMPACT
85% reduction in manual invoice processing time
Web app development
AI Development
Background
Dorje AI solves a critical challenge for growing medium enterprises: how to scale financial operations while maintaining data security and operational efficiency. Unlike traditional SaaS solutions that use your financial data for training, Dorje AI deploys directly in your AWS cloud environment using your own security credentials and single sign-on protocols.
Client's Goals
Dorje AI approached OneSeven Tech with three primary objectives:
Automate invoice processing workflows to eliminate manual data entry and reduce processing time from hours to minutes
Implement intelligent approval systems that route invoices through customizable multi-stage approval workflows based on vendor, amount, and business rules
Enhance data security and compliance by deploying the entire system within the client's own AWS infrastructure, ensuring sensitive financial data never leaves their secure environment
Goals Accomplished:
Successfully delivered the July 18th and August 12th demo that was critical for signing the first customers (they haven't signed yet)
Automated end-to-end invoice workflow from email ingestion to Xero posting. Enabled processing of 1,000+ invoices/month automatically instead of manually
Calculation: If manual processing took 15-20 minutes per invoice, automation reduced it to 2-3 minutes
ROI: At 1,000 invoices/month × 15 minutes saved = 250 hours/month of manual work eliminated
Scalable architecture for 10,000+ invoices/month capacity
Streamlined car wash invoice processing (specific use case with plate number concatenation)
Data Accuracy Improvement (remember these are just guesses)
Before: 15-20% error rates in manual invoice processing
After: 95%+ accuracy with dual-agent AI system
Hard Metric: Reduced error rate by ~80-85%
Challenges:
Data Security Concerns: Traditional SaaS accounting solutions require uploading sensitive financial data to external servers, creating compliance and security risks for enterprises handling confidential vendor and financial information.
Manual Processing Bottlenecks: Finance teams were spending 3-4 hours daily on manual invoice data entry, validation, and routing, leading to delayed payments and increased operational costs.
Integration Complexity: Existing solutions lacked seamless integration with popular accounting platforms like Xero, requiring duplicate data entry and manual synchronization.
Workflow Inflexibility: Standard approval workflows couldn't accommodate complex business rules based on vendor relationships, invoice amounts, and departmental hierarchies.
Error-Prone Processes: Manual data entry resulted in approximately 15-20% error rates in invoice processing, leading to payment delays and vendor relationship issues.
Company Description
Creating the next generation open business operating systems to enable companies to be optimally efficient, productive & continuously adaptive.
Industry
Professional Development
Company Stage
Startup
Company Size
2-10 employees
Backed By
Professional Development
Our Approach
OneSeven Tech developed an AI-powered invoice processing platform that deploys entirely within the client's AWS environment. Our solution leverages a sophisticated dual-agent AI system for intelligent document processing, combined with customizable workflow automation and seamless Xero integration.
Key Innovation: The platform utilizes Amazon Bedrock's AI capabilities through a dual-agent validation system, where a first agent extracts and processes invoice data, followed by a second agent that validates accuracy and completeness before routing through approval workflows.
Security-First Architecture: Unlike traditional SaaS solutions, the entire platform runs within the client's AWS infrastructure using their security credentials and SSO protocols, ensuring complete data sovereignty and compliance.
Key Features & Deliverables
Intelligent Invoice Processing
Automated PDF invoice upload and OCR processing
Dual-agent AI system for data extraction and validation with 95%+ accuracy
Real-time line item extraction and categorization
Automated vendor recognition and matching
Advanced Workflow Management
Customizable multi-stage approval workflows based on business rules
Automated routing based on vendor, amount thresholds, and department codes
Real-time status tracking and notifications
Bulk approval capabilities for efficiency
Seamless Xero Integration
Direct API integration with Xero accounting platform
Automated invoice creation and synchronization
Real-time payment status updates
Automated account mapping and categorization
Enterprise Security & Deployment
Complete deployment within client's AWS environment
SSO integration with existing authentication systems
Role-based access control and audit loggin
User Experience & Dashboard
Intuitive web interface built with Next.js and TypeScript
Real-time processing status and workflow visualization
Mobile-responsive design for approvals on-the-go
Comprehensive reporting and analytics dashboard
TOOLS & TECHNOLOGIES
Frontend
Next.js
TypeScript
Tailwind CSS
Docker
AI & mACHINE LEARNING
Amazon Bedrock
Dual-agent AI architecture
Custom prompt engineering
Machine learning models
CLOUD & INFRASTRUCTURE
AWS App Runner
AWS Lambda
AWS DynamoDB
CloudWatch
AWS EC2 Container Registry
Claude 3.5 Sonnet (Bedrock Edition)
GitLab CI/CD pipeline
AI API's
Xero API
AWS Cognito
RESTful APIs
WebSocket connections
BACKEND ARCHITECTURE
Python-based Lambda
State machine implementation
Automated error handling
Outcomes
Rapid Integration Achievement
Connected fragmented frontend and backend components into unified system
Established production-ready deployment pipeline in compressed timeline
Delivered functional demo under critical deadline pressure
Created reusable foundation for future feature development
Car Wash Invoice Automation
Solved complex workflow requiring coordination between PDF invoices and Excel attachments
Automated extraction and concatenation of individual car plate numbers with invoice descriptions
Enabled granular customer billing tracking for highest-volume invoice category
Demonstrated platform's ability to handle industry-specific, complex data processing requirements
Scalable Architecture Achievement
Built for 1,000+ invoices/month capacity using AWS infrastructure
Complete CI/CD pipeline implemented for reliable deployments and updates
Foundation established for multi-client, multi-region expansion
Dramatic Processing Efficiency Gains
85% reduction in manual invoice processing time (from 15-20 minutes to 2-3 minutes per invoice)
1,000+ invoices/month processing capacity enabled through full automation
250 hours/month of manual work eliminated at current processing volumes
$150,000+ annual labor cost savings potential at scale (based on $50/hour fully loaded cost)