DPD in Russia deploys AI virtual assistant for customer support
DPD
DPD in Russia has launched an LLM-based virtual assistant for its customer service department. The system reduces average handling time by 15% and post-processing time by 17%, while improving consultation quality by 7%.
DPD in Russia introduced a virtual assistant powered by a large language model (LLM) for its customer service department. The system uses a "retrieve then answer" architecture: corporate documents are preprocessed, structured, and segmented, and when a query comes in, the most relevant fragments are retrieved to generate an accurate answer. Unlike traditional keyword-based search, the LLM understands semantic similarity. Early results show average handling time (AHT) decreased by 15%, after-call work (ACW) by 17%, internal consultation quality score rose by 7%, and confirmed errors dropped by 15%. Employee attrition decreased by 2%, and new specialist onboarding time was reduced to one day. DPD emphasizes that such projects require thorough data preparation to avoid errors from unstructured or contradictory documents. The company plans to expand AI use in customer service and adapt the technology for other departments.
- Abbreviations
- LLM = Large Language Model — большая языковая модель
- AHT = Average Handling Time — среднее время обработки обращения
- ACW = After Call Work — время постобработки
Source: CNews —
original