AI-powered contract extraction for a travel company
A mid-size B2B travel company applied AI to automate hotel contract extraction and loading. The solution slashed manual work, boosted data accuracy, and cut contract onboarding from weeks to days — giving the company faster, leaner, and more scalable operations.
Challenge
Our customer, a mid-size B2B travel company, manages numerous direct hotel contracts. These contracts were manually uploaded into the Extranet system used for managing inventory, rates, and availability.
The manual process was slow, resource-intensive, and prone to errors. Accuracy issues in contract data created further inefficiencies downstream.
The company needed a faster and more reliable approach, and the automation project had to be delivered under a very tight deadline.
Business goals
The project aimed to:
- Reduce reliance on manual data entry for contract extraction
- Speed up the loading of frequently used contracts
- Improve data accuracy
- Lower operational risk associated with errors in contract handling
Solution
The high-level process
To optimize contract handling, we built a system combining a web interface, AI-powered extraction service, and integration with the Extranet.
Web interface and backend solution
Allow users to:
- Upload contract files
- View and select previously uploaded contracts
- Trigger AI-based data extraction for a selected contract
- Review and manually edit extracted data after processing
- Choose from available extraction templates and re-run extraction when needed
- Save and reuse extraction templates for recurring contract formats
AI data extraction service
Handles the full cycle of contract recognition and data capture:
- OCR conversion: Transforms PDF contracts into text for further processing
- Contract clustering and template storage: Groups similar contracts and stores reusable templates
- Template matching: Identifies the correct template for each contract using text analysis and matching algorithms
- Field extraction with LLMs: Uses ChatGPT guided by predefined templates to extract required fields, accounting for variations across documents
- Post-processing: Reviews and corrects ChatGPT outputs to handle errors and inconsistencies, ensuring clean data
Extranet integration
Validated contract data is sent directly to the Extranet database, reducing manual effort.
Technology stack
- Backend: .NET C#
- Frontend: Angular + PrimeNG
- Database: PostgreSQL
- Data extraction: Python, ChatGPT API, Google OCR API
Results
Within the first rollout phase:
- Contract loading time became 4-5 times faster compared to manual processing
- Data accuracy improved by 80%, reducing the need for corrections
- Workload for staff dropped by 80%, freeing them for complex tasks
- Onboarding time for new contracts cut from weeks to days, improving time-to-market for new hotel offers
Conclusion
For a travel company dealing with large volumes of hotel contracts, AI-powered automation proved to be a strong efficiency driver. It reduced repetitive work, improved accuracy, and accelerated contract onboarding — delivering both short-term ROI and long-term scalability.
