Situation
A large pharmaceutical company was using an AI system for internal purposes. Concerned that AI-generated chat data may be required for future litigation, the client reached out to their TCDI project manager to determine if that type of data could be ingested into their eDiscovery platform in a usable format.
Because the export structure lacked official documentation at the time, the project manager engaged TCDI’s Systems Operations (SysOps) team to develop a solution for processing and ingestion.
TCDI’s Systems Operations team developed a custom solution to convert and ingest AI chat exports in
8 hours
Resolution
The AI exports were structured in JSON, a format designed for systems and applications to exchange data but not necessarily for people to read. This structure, combined with the lack of documentation, made it difficult to determine how individual messages related to one another, rendering the data effectively unusable in its raw form.
To resolve this challenge, TCDI’s SysOps team used the company’s internally hosted reasoning model within its on-site Tier III data center to analyze and document the export structure. Specifically, they were identifying and outlining how each message connected within a conversation. Once the structure was defined, the same model was used to develop and refine code that converted the JSON exports into a readable format (HTML and PDF). These files were designed to replicate the look and feel of a typical chat thread while preserving all underlying content and metadata.
Through multiple validation cycles, engineers confirmed that the outputs were accurate, readable, and defensible for eDiscovery review. Additionally, because all data was kept within TCDI’s infrastructure, client information never left the company’s secure, controlled environment.
Impact
The project was completed within eight hours, including research, code development, validation, and a successful run using a test dataset. By comparison, traditional code development and testing would have taken at least a week to achieve similar results.
The project highlighted how TCDI’s internally hosted AI capabilities, combined with the specialized expertise of its SysOps team, can accelerate the development of solutions to complex data challenges while maintaining the security and defensibility clients expect.
The client reviewed and approved the workflow for future use, confirming that TCDI had effectively addressed their request. The result: a repeatable process that will be compatible with eDiscovery platforms for defensible processing, review, and production of AI-generated data.
