Automating inbound reorder cards with Azure AI

Client background

The organization is a leading supplier of Durable Medical Equipment (DME), specializing in delivering critical health devices and supplies to individuals across the country. With a focus on personalized, efficient and high-quality service, the company has built a strong reputation and competitive advantage within the healthcare industry. 

The business challenge

The organization faced an operational challenge in processing reorder cards—physical forms completed by patients to continue or change their product subscriptions. The process was entirely manual, requiring staff to collect, scan and interpret each form. This approach was labor-intensive, introduced processing delays and increased the risk of data entry errors. 

To improve efficiency and reduce the burden on staff, the organization partnered with Baker Tilly to develop a proof of concept (POC) for automating the intake and processing of these forms. The goal was to extract key information—including handwritten text, checkboxes, selection fields, signatures—and update patient records within existing systems. 

Strategy and solution

Baker Tilly worked with business stakeholders to define requirements and assess the variability of form types. Using a standard reorder card template as a foundation, the team designed a solution leveraging Azure Document Intelligence, Microsoft’s AI-powered form recognition service. 

Azure Document Intelligence was trained to classify and extract relevant information from slightly varied versions of the form. The solution identified and pulled data from signatures, handwritten notes in margins, checkboxes and marked selections. To support data quality, a Human-in-the-Loop (HIL) process was included for reviewing low-confidence extractions. 

In parallel, a separate workstream focused on building a centralized data platform. Data extracted via Azure Document Intelligence was incorporated into this platform, enabling reporting, analytics and integration into downstream workflows. 

Results and impact

• Demonstrated automation feasibility: The proof of concept showed that Azure Document Intelligence could effectively extract handwritten notes, checkboxes and other structured data from reorder cards with a high degree of accuracy. 
• Informed future automation plans: The effort illustrated how similar document-driven processes such as patient intake or service requests could be streamlined using a similar approach. 
• Enhanced understanding of data integration needs: Integrating extracted data into a centralized data platform during the POC highlighted key considerations for enabling reporting, analytics and downstream system integration, helping the organization better plan for a full-scale implementation. 

This engagement successfully demonstrated how Azure-based AI solutions can address real-world operational challenges while supporting a broader data modernization strategy.