A leading Indian diagnostics and stem cell company, serving over 5,000 partner labs nationwide, was stuck manually processing handwritten Test Requisition Forms (TRFs) a bottleneck that capped how fast it could grow. DCM Infotech built an Agentic AI solution using Google Vertex AI that reads and interprets handwritten TRFs with over 90% accuracy, letting the client onboard more partners without hiring more staff.
The results at a glance
- 90%+ accuracy in handwritten TRF recognition using Agentic AI
- 8–9 staff redeployed from manual data entry to higher-value roles
- 100% scalability — onboarding new partner labs without adding headcount
- Improved profitability through faster, error-free workflows
About the customer
The client is a well-established diagnostics and stem cell company in India, serving more than 5,000 labs and collection centers across the country. Every partner sends physical Test Requisition Forms the documents that specify which tests to run and carry the clinical context behind them. These forms are handwritten, dense, and often full of tick marks and shorthand notes that make them genuinely difficult to digitize.
Before automation, each TRF passed through two people: one for data entry, one for verification. That two-person model is slow by design, and it doesn’t scale, growing the partner network meant growing the data entry team at roughly the same rate.
The business challenge
TRF forms are a harder automation problem than they look. Handwriting styles vary across thousands of referring doctors and technicians, layouts are dense, and tick marks are frequently ambiguous. The client had already tried automating this process before working with DCM previous vendors couldn’t get past roughly 60% accuracy, which left the team skeptical that automation could work here at all.
The client came to DCM Infotech after seeing DCM’s automation work on invoice processing for another enterprise client, and decided to test whether the same approach could hold up against a much harder document type.
DCM’s approach
DCM implemented an Intelligent Document Processing (IDP) for healthcare solution combining OCR, AI, NLP, and Agentic AI to automate the extraction and interpretation of handwritten TRFs not just transcribing them, but reasoning about what the content actually means.
Key technologies used
- Google Vertex AI — for advanced handwriting recognition, using document text detection to extract handwritten content directly from scanned TRFs
- Agentic AI layer — for reasoning through the extracted content: interpreting tick marks, cross-referencing data fields, and flagging genuine ambiguities for human review instead of guessing
DCM’s team also worked with the client to redesign the physical TRF layout, clearer tick boxes, more consistent spacing — which further improved the AI’s read accuracy. The technology did the heavy lifting, but the form redesign made its job measurably easier.
Proof of concept and results
The proof of concept, run on a sample set of real TRFs, achieved more than 90% accuracy a substantial jump from the roughly 60% ceiling the client had hit with prior automation vendors. That accuracy level was high enough for the client to trust the system with live volume rather than treating it as a pre-processing step that still needed full manual verification behind it.
Business impact
- Scalability unlocked — the client can now onboard additional partner labs without a proportional increase in data entry headcount
- Productivity gains — 8 to 9 staff previously tied up in manual TRF entry and verification were redeployed to revenue-generating roles
- Higher accuracy and reliability — fewer human transcription errors, faster turnaround from form receipt to lab processing
- Profitability boost — operational efficiency gains translated directly into better margins and room for continued growth
Why this matters for Indian diagnostics more broadly
India’s diagnostics sector relies heavily on manual document processing, making scalability challenging as laboratory networks expand. Agentic AI solutions for diagnostics & healthcare are helping organizations overcome these constraints through intelligent document processing, handwriting recognition, workflow orchestration, and autonomous decision-making. By reducing dependence on manual data entry and verification, healthcare organizations can improve operational efficiency while maintaining accuracy at scale.
Lessons learned
- Form design matters. Even small, collaborative changes to a form’s layout can materially improve AI reading accuracy.
- Domain expertise is not optional. Understanding how diagnostic workflows actually function — not just the document format — was central to getting this right.
- Agentic AI goes beyond automation. The difference between this result and prior 60%-accuracy attempts wasn’t better OCR — it was adding a reasoning layer capable of handling ambiguity, not just clean, expected input.
Conclusion
DCM Infotech’s Agentic AI solution changed how one of India’s largest diagnostics companies processes Test Requisition Forms moving from a two-person manual bottleneck to an autonomous workflow with industry-leading accuracy. By combining Google Vertex AI and a purpose-built agentic reasoning layer with real domain expertise, DCM delivered a result the client’s previous automation vendors couldn’t reach: scalability, profitability, and operational resilience, without adding headcount.
FAQ
What accuracy did DCM Infotech’s Agentic AI achieve on handwritten TRFs?
Over 90% accuracy in the proof of concept, compared to roughly 60% accuracy from the client’s previous automation attempts.
What technologies power this Agentic AI solution?
The solution combines Google Vertex AI for handwriting recognition, the Langchain framework for document workflow orchestration, and an Agentic AI reasoning layer for interpreting ambiguous content and making decisions.
Did automation replace the diagnostics company’s data entry staff?
No. Eight to nine staff previously handling manual TRF entry and verification were redeployed to higher-value, revenue-generating roles rather than being let go.
Can this approach work for other handwritten or unstructured document types?
Yes. The same combination of handwriting recognition, document segmentation, and agentic reasoning applies to other unstructured document workflows: invoices, compliance forms, and clinical notes among them.
See more Agentic AI use cases for diagnostics and healthcare, including automated test requisition processing, LIMS integration, and HIPAA-compliant form completion, on DCM Infotech’s Agentic AI Solutions page.