AI Strategy for ComfortDelGro
Integrated AI to CDG Zig projects, microservices for Back-End.
Business Challenge
Too many microservices leading to sometimes the Agent AI mislead structure from this service to the other services. Hallucination stills occur, when too much data repeated in many scopes and new scopes definition without evidence, AI stills based on the old scope with out verify. Find a way to connect real time monitoring data (CloudWatch, Grafana, TrueWatch, Logging) to current autonomous flow.
Solution Delivered
Implement skills for AI in microservices. Deliver guideline to using MCP connecting Confluent, Jira and Bitbucket. Input resources knowledge file for each services for Agent AI. Deliver the best practices for each solution from tickets.
Outcomes Achieved
Productivity improvements Quality improvements Accuracy improvements Knowledge management improvements Operational efficiencies Onboarding improvements Risk reduction Compliance improvements AI performance improvements
Measurable Business Outcome
Productivity improvements clearly, code was easier to maintain. Usually it takes around 2 hours to review, now it just ~30mins to 1 hour. Scope are defined correctly (~90%). Hallucination occurs less frequently. Now AI can compute the impact better based on monitoring data for every solutions made from ticket for humans. More CDG members started AI transition workflow.