Optimizing Handwritten Contract Processing with AI Handwriting Recognition

For Funeral Directors Life (FDL), scaling operations required a strategic investment in automation. The core challenge involved the labor-intensive processing of handwritten contracts, a task that consumed significant time and limited data utilization.

Following a recent interview with FDL, this case study summarizes the process, metrics, and technical strategy FDL used to implement an AI-driven solution.

The Initial Challenge and Core Metrics

The initial automation project focused on converting the processing of handwritten paper contracts into a fully automated digital workflow.

The shift was achieved by using Grooper for intelligent extraction, followed by a Robotic Process Automation (RPA) bot to enter the data into the line-of-business system. This allowed the company to handle high volume without scaling personnel.

Advanced AI Handwriting Recognition for Non-Standard Documents

FDL expanded their use of AI to solve a traditionally “impossible task”: extracting data from non-standard funeral home price lists. Since every price list has a different format, layout, and structure, rule-based extraction was not feasible.

• Solution: FDL used Grooper’s AI Extract feature to leverage Large Language Models (LLMs) like OpenAI, providing simple instructions to structure the unstructured data.
• Optimization via Fine-Tuning: By creating and hand-keying only ten batches of documents, FDL fine-tuned a less expensive model (a 400 mini version). This process resulted in better accuracy than the initial, more expensive model, while reducing processing cost by a factor of ten.
• Accuracy Boosts: FDL reported an initial 10% bump in accuracy from using AI Extract, followed by an additional 15% bump in a state-specific model after fine-tuning.

Strategic Deployment and Data Security

FDL’s strategy emphasized both reliability and data privacy:

1. Hybrid Extraction: For maximum reliability, FDL uses traditional Grooper extractors first. The AI Extract feature is then deployed as a secondary fill-in to capture any fields missed by the traditional rules, significantly boosting overall accuracy.
2. Mobile Integration: The solution is integrated with their mobile app, Passare. Funeral directors can scan handwritten vital statistics forms directly in the app. Grooper converts handwritten text and returns the structured data, eliminating the need for manual re-typing and enabling smooth data flow into their case management system.
3. Security Posture: As an Enterprise customer using the OpenAI API, FDL confirmed that their data is not used to train the LLM models, addressing major security and privacy concerns common with free-tier or default public models.

Rationale for Platform Selection

FDL selected Grooper over three other evaluated platforms based on the following criteria:

• Control and Configuration: The platform offered the highest degree of control, scalability, and flexibility, allowing the development team to configure the system to meet their precise needs rather than being forced to adopt a vendor’s pre-defined workflow.
• Scalability: The architecture was lightweight and capable of scaling to handle future document processing needs.
• Comprehensive Recognition: The ability to combine multiple recognition technologies (including the latest in handwriting recognition) provides higher, more reliable accuracy than relying on single, un-combined cloud services.

This level of detail, process optimization, and quantifiable ROI demonstrates the transformative potential of strategic automation and AI-powered handwriting recognition in document-heavy industries.

For a full technical discussion, including the live demonstration of the mobile application integration, the complete webinar recording is available for registration.

Get the Video and Learn More: