From Hours to Minutes: How AI Transformed HR Document Management for a Major Port Operator
Indexing, Separation and Classification
Solution
The project had the following requirements:
- training and configuration within a week
- process a back-file of over 30 million pages
A major Caribbean port operator struggled with managing over 50 years' worth of HR records, around 300,000 files in total, leading to significant inefficiencies. Each HR file contained over 100 pages, divided into up to 24 sub-document types like hiring documentation, salary notifications, reprimand letters, contracts, and training certificates. The process of scanning and indexing these files was labor-intensive due to the following challenges:
- Manual Document Preparation: The existing software required pre-labeled separator sheets to classify the 24 sub-document types, adding an extra manual step before scanning.
- Time-Consuming Data Entry: Key fields like Document Date and Document Subject/Description still required manual input, further slowing down the process.
- Handwritten Text: More than 50% of the documents contained handwritten notes, making automatic recognition and classification difficult.
- Slow Processing: With manual data entry and classification steps, it took several hours to fully scan and index a single HR file, which averaged 150 pages.
Aluma introduced a smart solution using OCR/ICR and AI to simplify and speed up their HR file management process. Here’s how:
Data Collection and Training: Over four days, sample HR documents (covering 24 document types, including those with handwritten text) were collected to train the AI models. This allowed the system to recognize and classify documents without manual separators.
AI Document Classification: The AI was trained to automatically sort documents into 24 categories, handling handwritten and printed text, saving time and effort.
Automated Data Extraction: Key details, like document dates and descriptions, were extracted from both typed and handwritten records, simplifying the indexing process.
Testing and Deployment: After successful testing, the solution was fully deployed, marking a new, efficient way to manage decades of HR records.
Impact
Drastic Time Savings and Efficiency Gains
The automated system reduced the time to scan and index each HR file from hours to minutes, enabling swift processing of documents without intensive manual effort.
Enhanced Accuracy and Automated Handwritten Recognition
By utilizing OCR/ICR and AI, the solution accurately classified and extracted key information from documents, including those with handwritten text, which previously presented a challenge.
Improved Scalability and Reduced Manual Intervention
The automated process could handle the full archive of 300,000 HR files, reducing the need for manual data handling and ensuring consistent, scalable record-keeping for the growing organization.
