P241: AI Verification of Key Information in the Events Application and Approval Process

Poster Presenter

Xiaofeng Xiang

Sr. Medical Compliance Manager
Bayer Healthcare Company Limited
China

Objectives

This project aims to automate the verification process for key information in pharmaceutical company events application and approval using artificial intelligence (AI) Technology, thereby reducing human errors and mitigating business risks.

Method

We initially developed a website on the AWS Cloud, harnessing the language comprehension and generation capabilities of GPT-4, which empowered us to perform file comparison and analysis efficiently. The website can generate a comparison result checklist to identify any inconsistencies.

Results

After logging into our website, users can upload the files they want to compare on the left part of the webpage, the files that need to address inconsistencies are supporting documents for the activity application such as investment invitation, conference agenda, contract and attachments, Fair Market Value table, and application form. The file formats can be Word or PDF. Even for documents containing scanned images, the program can recognize the text information through OCR technology. After a brief processing time of 15-60 seconds, depending on the file's complexity, the webpage will display the key information of each document on the right side, and provide a judgment conclusion on whether the content is consistent. The comparison results are highly satisfactory, ensuring consistency between the predetermined key information in contracts and the actual content.

Conclusion

In the process of applying for events activities, inconsistencies often arise in documents such as investment letters, conference agenda, contracts and attachments, FMV tables, and application forms. It is a daily business pain point faced by pharmaceutical enterprises. In order to address this issue, and strengthen the accuracy of event application document management, we leverage AI technology automate to extract predetermined key information from events documents, such as association names, activity names, conference dates and locations, amounts, and returns. By generating a checklist and comparing the extracted information, we easily identify any discrepancies and minimize human-induced errors in the verification process. This innovative approach has proven to be a valuable solution that increased efficiency by automating the extraction and comparison tasks, highlighted the significance of accurate document management, mitigated business risks, and enhanced GDP (Good Documentation Practice) to all aspects of daily document management.