Requirements
Our client was manually keying data from writs into their case management system. The process was slow and introduced risk of human error, presenting an opportunity to implement an AI-powered document automation solution.
Our earlier solution automatically captured relevant data from letter-style writs.
It used Google Cloud Vision and optical character recognition (OCR) to extract text from a single image.
Although this worked well for straightforward text documents, form-style writs presented a different challenge: checkboxes, multi-column layouts, labels paired with values, and content spanning multiple pages made them significantly harder to process using the same approach.
The existing solution was tailored to each client’s writ format, so every format change required individual development work.
We, therefore, looked for a more specialised extraction tool to improve accuracy and address the client’s objectives: increasing efficiency, reducing manual intervention, and lowering the risk of human error.
Solution
We worked using the Scrum framework, delivering the AI-powered document automation solution across four sprints.
We chose Google Cloud Platform’s Document AI Form Parser, which uses Google’s Gemini AI to pass over each document, correcting issues like skewed or inverted pages, sharpening the image if needed, and normalising the input before running the main data extraction.
The off-the-shelf Google model achieved around 70-75% accuracy on the writs we used for initial testing.
The writs were inconsistent by nature, as writs are often completed by hand and data entered outside of expected fields.
We fine-tuned the model for this document type using 100 documents supplied by the client. In subsequent testing across a range of input conditions, the model achieved a reported accuracy of 99.8%.
Once live, the process works as follows:

Results
100% More Processing Capacity
The automation halves the time required to complete each case, enabling twice as many cases to be processed in the same time, a 100% increase in processing capacity, while improving accuracy.
Accuracy
The client is now automatically processing around one-third of the writs received.
One-Third Automated
By reducing the manual effort associated with case handling, team capacity has increased, enabling individuals to focus on higher value activities and develop additional skills.
More Time for Higher-Value Work
Building on these results, the business plans to embed AI further across its processes. The aim is to strengthen overall operational capability while maintaining a consistent client experience.
Need help with AI-Powered Document Automation?
Discover how Koderly can help your business automate document processing, reduce errors and free up valuable staff time by filling out the form below, or viewing our software development page.

