How Queppelin deployed an ML-powered data extraction engine to automate high-volume lab report processing for a leading diagnostics organisation — reducing manual effort by 15 FTEs with zero backlog.
Our client is a diagnostics and pathology services organisation managing thousands of patient lab reports across multiple facility locations. With a growing patient load and increasingly complex multi-page report formats, their operations team faced a critical bottleneck: all data from incoming lab reports had to be manually keyed into their core system — with no API bridge available to automate the read/write cycle.
The organisation was processing 400–500 lab reports per day, each ranging from 7 to 48 pages in length. With no programmatic access to the lab system, every field had to be entered by hand. A separate checker layer added further overhead, and the growing backlog was approaching a week.
The system was designed to learn — not just extract. Rather than relying on rigid template-matching, Queppelin built a document understanding layer that continuously improved from real operational feedback, making it resilient to format variation across report types.
Incoming lab reports sampled and grouped by unique structural format to build representative training sets.
Document Understanding package trained in UiPath Data Manager to recognise fields across heterogeneous report layouts.
Live documents processed by the ML model, extracting structured fields with confidence scores per extraction.
Low-confidence items routed to Action Centre for human review — corrections captured as metadata for retraining.
ML packages retrained on a 20-day cycle using accumulated human feedback, progressively reducing exception rates.
Validated output written back directly to the lab application — no manual intervention required in steady state.
The impact was immediate and structural — not incremental. The organisation was able to redeploy its data entry staff, dismantle the checker function entirely, and eliminate all processing backlogs from day one of steady-state operation.
Queppelin used UiPath's intelligent automation suite, combining RPA with ML-powered document understanding and a human oversight layer — all integrated through a single orchestrated workflow.
