How AI Automation Separates Tacit Knowledge into Data to Enhance Maintainability

The KT Cloud Azure Transition team shared their trial-and-error experiences and solutions while automating repetitive infrastructure request handling tasks with AI. Initially, they passed requirements directly to the AI to generate scripts, but faced 'automation debt' as business logic was deeply hardcoded into the code, causing maintenance costs to surge upon changes.

Data-izing Tacit Knowledge

The core of the problem lay in relying on AI without explicitly documenting the team's tacit knowledge. To resolve this, they applied the 'separation of concerns' principle, extracting mapping rules from JavaScript code into independent CSV data files.

Improved Maintainability

After restructuring, adding new environments or changing cost departments no longer required code modifications; updating only the CSV files sufficed. Consequently, operations staff and planners without coding knowledge could safely manage automation logic via Excel, evolving from individual scripts to a team-level system.