How DiDi built intelligent contact center QA with Amazon Bedrock

This article describes how DiDi built an intelligent contact center quality assurance system on Amazon Bedrock, replacing an opaque third-party solution with a transparent, self-owned AI architecture.

• Intent verification pipeline improved accuracy from 38% to 86% through two-level information isolation design
• Compliance evaluation pipeline achieved 90%+ accuracy using dynamic prompt assembly for multilingual, multi-business-line coverage
• Voice of Customer (VOC) pipeline compressed hours of manual trend analysis into minutes using three-stage processing
• System covers Spanish and Portuguese across ride-hailing, food delivery, and financial services business lines
• Core principle: precise context management—controlling what information the model sees—drives reliable LLM application performance
• Amazon Bedrock Guardrails provides PII masking and sensitive information redaction for responsible AI practices

DiDi's system demonstrates that LLM application performance depends more on context management and data definitions than on prompt wording alone, delivering transparent, auditable QA decisions with full reasoning chains.