The Marlabs team used RPA in the form of CODIbot (Content for Disclosure bot). CODIbot extracts the data and transfers it to PharmaCM. More than 80% of the data fields were automated. Once the data is entered into PharmaCM, CODIbot sends a summary report to the research team.

To do this successfully, the Marlabs automation experts collaborated with the client’s medical writing team to update the clinical study protocol template. This enabled further automation so the data could be reused. In addition, the confidentiality of protocol documents, an essential compliance requirement, was assured using appropriate storage and encryption processes.

The automation team classified PharmaCM data fields into three groups:

• Not applicable (18): Field not filled out in PharmaCM
• Not in scope (31): Data fields that require SME expertise to be completed, for example, information that cannot be extracted unambiguously from the clinical study protocol
• In scope (61): Data fields that the bot can fill as the information is available in the protocol

Now, when the client’s clinical trial protocol team creates a new study and PharmaCM uploads the data to a file share, an email is sent to start CODIbot. First, CODIbot checks to see if the file has been processed before. If not, it extracts pre-specified protocol content such as study title, secondary ID, brief summary, ARMS and intervention section, study endpoints, eligibility criteria, and some administrative details. To do this, CODIbot uses a set of validation rules. Then, the extracted data is entered into PharmaCM, and an email is sent with a summary of the outcome and the execution report. If the process is unsuccessful, the report shows the reason for the failure. In addition, errors are collected by the team to improve the process further.

The technologies used to create the automation included process mining, RPA, OCR, AI-based intelligent automation, and conversational AI.