Project
About project
About project
About project
The AI Contractor Assessor is a service that automatically reviews the documentation submitted by contractors participating in procurement processes across key industrial sectors, including extractive industries, metalworking, mechanical engineering, and related fields.
The primary users are companies that regularly engage contractors for a wide range of tasks, from employee medical services to equipment supply or catering. Contractors submit large documentation packages that include information about personnel, equipment, work experience, licenses, certificates, and other supporting materials.
The AI Contractor Assessor analyzes these materials automatically, enabling clients to quickly generate a shortlist of suitable contractors.
Changes After Implementation
- Document review time was significantly reduced: a 50-page package with a complex structure is analyzed in 5–7 minutes.
- The selection process became transparent, and decisions more objective.
- Conditions were equalized for all participants, reducing human bias.
The business gains direct benefits: procurement cycles are completed faster, projects start sooner, and downtime and costs caused by lengthy procurement procedures are reduced.
In terms of data security, the service meets the requirements of the financial sector, the public sector, and healthcare. It can be deployed on-premise on the company’s own servers and functions fully without internet access.
Business Challenge
Business Challenge
Business Challenge
Large industrial companies regularly conduct procurement cycles and face a recurring issue: high volumes of heterogeneous documentation significantly slow verification and increase the risk of errors. This directly affects project launch timelines and the execution of production plans.
At Doubletapp, we have many years of experience building machine learning and computer vision solutions and have seen firsthand how AI accelerates complex processes. We identified a clear demand for shorter timelines and increased procurement transparency and began developing a tool that automates procurement document analysis, making contractor selection faster, more accurate, and objective.
To validate the approach, we interviewed procurement specialists at major enterprises, studied their existing workflows, and tested early prototypes on real data. This allowed us to design a clear, user-friendly service that closely matches real operational needs.
Solution
Solution
Solution
To build a solution that genuinely transforms procurement workflows, we conducted extensive research. We spoke with ten industrial clients and engaged representatives from the country’s largest manufacturing companies—teams that work daily with large procurement documentation and understand process bottlenecks in depth.
We presented the product concept, ran live demonstrations, and invited clients to test prototypes using their own documents. Their feedback helped us refine core usage scenarios, adjust document analysis logic, and validate our hypotheses about which procurement stages could be automated without sacrificing control or accuracy. Today, we continue developing the product and running pilot implementations with large enterprises to fine-tune performance in real-world conditions.
These insights confirmed that accelerating procurement is not a single-company challenge. Fast and objective contractor selection in key industrial sectors delivers measurable macroeconomic benefits: earlier project launches, greater operational stability, job creation, and growth across related industries.
Engineering Excellence: Implementation & Technical Details
Engineering Excellence: Implementation & Technical Details
Engineering Excellence: Implementation & Technical Details
The system operates in three main stages:
- Procurement teams upload requirement documents and contractor submissions into the system. File formats are unrestricted. The platform converts all inputs into a format suitable for LLM processing. OCR is applied to scanned documents, along with preprocessing for other file types.
- The system analyzes all requirement documents using AI and generates a single, structured, numbered list of criteria to be checked against each contractor’s documentation.
- The system evaluates each contractor’s documentation for compliance. The output includes:
- A per-contractor breakdown showing which requirements are met, unmet, or require clarification
- Precise document references where each requirement is confirmed or contradicted, allowing procurement specialists to quickly verify conclusions
- A final summary table ranking contractors from most to least suitable
For the second and third stages, open-source LLMs are used and can be deployed entirely within the client’s infrastructure. Models such as DeepSeek V3.1 and Qwen3-235B-A22B demonstrated strong performance.
To assess system quality, we assembled a dataset of real procurement cases with contractor documentation and manually defined requirement lists. Each contractor’s compliance status was annotated. Based on this dataset, we can automatically measure system performance. At present, requirement extraction recall stands at 95%, and the accuracy of requirement compliance determination is 93%.
