Summary
SSD processes over 20,000 building permits annually, requiring manual reading and categorization of each document. We developed an AI-driven solution using the Kognitos AI Tool to automate the extraction of data and categorize documents into 26 overlapping categories. Despite the unstructured nature of the documents, our system achieved 90% accuracy in data extraction and 75% in categorization, significantly reducing manual effort and allowing staff to focus on other tasks.
Challenge
Motivation to change
SSD receives more than 20 000 building permits per year which are manually processed and forwarded for further processing or archivation. Every document must be opened and read to determine what are the next steps. Project focus was on extraction of data from the document (date), contextual reading and automation of the categorisation into 26 categories.
Solution
Change delivery
Major challenge of the project was unstructured format of the documents received by SSD and its categorisation. Documents are categorised into 26 categories which often overlap but are not specifically named so its important to understand the context of the document.
Tools & means
- Kognitos AI Tool
Outcomes
Change outcomes
In their team they have one person focused only on reading and categorising building permits which is more than 20 000 documents every year. We have managed to extract needed information with accuracy of 90% and categorise documents with an accuracy of 75%, resulting in more time for their colleague to focus on different tasks.
Used services
-
20 000documents per year
-
2categories
-
26subcategories