Summary
Key results
84% precision in automated document classification, compared to 50% with manual processing
96% precision in recognising equipment tags from documents
Manual data entry as a source of errors and financial loss
As a global Enterprise Asset Management provider, Prometheus Group had to process huge volumes of documents – from photos of nameplates and specifications to acceptance and commissioning reports. Data was manually transcribed into systems, which resulted in:
- frequent record errors leading to inaccurate stock levels
- delays in updates causing parts unavailability or excessive ordering
- higher operational costs because teams had to redo work and correct mistakes
With the company’s growing scale, the manual process became a major bottleneck. It limited efficiency and made asset maintenance planning harder in critical sectors such as energy, oil & gas, and pharmaceuticals. To reduce costs and improve operational precision, the client decided to entrust the project to Sii Poland.
Automating document processing with AI and AWS cloud
The goal of the project was to fully automate the extraction and classification of data from technical documents, reducing errors and accelerating operational processes. Sii’s experts in artificial intelligence, data science, and cloud built a multidisciplinary team of 34 that worked in Agile and closely collaborated with the client.
Scope of work included:
- converting scans and paper documents into digital data using Optical Character Recognition (OCR), making the content ready for analysis and integration with the client’s systems
- automatically extracting key information with Natural Language Processing (NLP) – the system identifies equipment names, quantities, and technical parameters, streamlining inventory updates
- recognizing visual markings with Computer Vision – the system detects tags, codes, and graphical elements and assigns them to the correct records
- training and evolving AI models in AWS SageMaker, providing a scalable platform that can handle new document types without increasing effort
- enriching the client’s knowledge base using publicly available online sources (web sourcing) to retrieve equipment characteristics
- translating equipment taxonomy and characteristics into 17 languages
This enabled Sii to scale the process without increasing headcount.
Higher data accuracy and lower operating costs
The new AI solution improved data quality and consistency, resulting in more efficient asset tracking and maintenance planning. The client gained the ability to process higher document volumes without increasing staffing, reduced the risk of errors, and lowered operational costs.
The system achieved 84% accuracy (compared to 50% with manual work) and an F1 score of 96%, combining high precision and recall. In practice, this means the system recognizes and assigns equipment tags faster and more reliably, eliminating the risk of mistakes typical of manual data entry. The implementation also created a strong foundation for further global expansion of inventory management services.
“With Sii’s support, we eliminated the biggest bottleneck in our inventory process. Automation not only accelerated the work, but also improved the quality of the data that underpins our decisions,” said a client representative.