TMF Week 2024 - Day 3 - Session 2: Using AI to Transform Unstructured Data in TMF Migrations

Introduction

Introduction

Good afternoon, good morning, and good evening to everyone, depending on where you are located. Welcome to Day 3 of TMF Week and the second session of the day. We are excited to have you here to discuss the transformative role of AI in managing unstructured data during TMF migrations.

Housekeeping Rules

Before we dive in, let’s go over a few housekeeping rules:

• All sessions are recorded, so there’s no need to inquire about recordings in the chat.
• Attendees will receive a certificate of attendance for the entire TMF Week conference.
• A Q&A session will follow the presentation, where attendees can ask questions, and network breaks will be provided after each presentation to facilitate discussions on best practices.

Session Overview

Today's session will feature a 30-minute presentation followed by a 15-minute Q&A and a networking break. The topic at hand involves the use of AI to transform unstructured data in TMF migrations, particularly focusing on classification processes.

Introductions

We are pleased to welcome our speakers, Morton Neelsen and Nick Larsson from Aista.

Morton Neelsen

As a principal consultant at Aista Life Science, Morton has spent 20 years in the pharmaceutical industry, working in the eTMF space within clinical organizations.

Nick Larsson

Nick is the director of technical services at Aista and has nine years of diverse IT experience. He is enthusiastic about sharing insights on applying AI in document management.

Challenges with Unstructured Data

Morton and Nick highlighted three main challenges faced by clients dealing with unstructured data:

1. Data quality, essential for successful AI applications.
2. Mitigating unrealistic expectations on AI's capabilities.
3. Knowing how to initiate the process of working with these advanced technologies.

Key Concepts

• Dark Data: Refers to data with unknown formats and no extensions, often originating from legacy systems.
• Unstructured Data: Readable data that lacks proper metadata for easy classification.

Typical Use Cases

• Migrations during mergers and acquisitions.
• Improving in-house systems, facilitating the transfer of poorly categorized documents.
• Data analysis and mining for valuable business insights.

Implementing AI

For leveraging AI technologies, an organization needs to:

• Decide on cloud or on-premise infrastructure based on data sensitivity.
• Utilize Microsoft Azure and OpenAI services effectively.
• Implement AI with a focus on security and compliance.

Real-Life Example

Morton shared an example involving the processing of over 23,000 files, where 14% were unknown types and 73% were readily usable but often required additional processing due to quality issues.

Addressing Limitations

Ignoring the limits of AI can lead to unrealistic results. Innovations will provide significant results only up to 70-80% in practice.

Infrastructure Setup

Using Microsoft Azure and OpenAI services facilitates the establishment of an efficient infrastructure for processing unstructured data.

Conclusion

The speakers emphasized the need for clearly defined objectives while embracing AI, starting small, and scaling gradually. The integration of AI tools should foster improved processing without compromising best practices.

Q&A Session

A lively discussion followed the presentation, addressing various audience inquiries regarding implementation timeframes, validation processes, data privacy concerns, and more.

Keyword

• AI
• TMF
• Unstructured Data
• Migration
• Classification
• Dark Data
• Microsoft Azure
• OpenAI
• Document Management

FAQ

How long does it take to implement an AI solution for TMF migrations?

It typically takes around 30 hours to set up the service in a Microsoft Azure environment.

Can ChatGPT be used for end-to-end data mapping for ETMF migration?

Yes, if the source and target systems are structured with good data quality.

What is the process to ensure document integrity during migration?

The migration process is designed to extract and classify without altering the original documents.

How can companies mitigate dependency on generative AI during TMF migration?

By maintaining a verification process that ensures data quality, AI can be used as a drafting tool, minimizing over-dependence on its outputs.

What privacy concerns are associated with using AI systems like Microsoft Chat GPT?

Data should never include sensitive information. When using Microsoft platforms, the data remains within the organization’s premises and is protected by Microsoft’s security measures.

Thank you for attending the session, and we hope you find this information helpful as you navigate the challenges of TMF migrations using AI technologies!