Real-time speech translation technology is reshaping the operational model of global call centers. In the past, multinational enterprises needed to set up separate agent teams for each language. Now, a unified communications system combined with neural machine translation (NMT) allows agents to serve global customers in their native language.
In July 2024, Microsoft announced that its Azure Speech Services achieved a BLEU score of 95% (a metric for translation quality) in real-time translation scenarios, a 12-percentage-point improvement over 2022. This means that during customer service calls, the translations agents hear are nearly indistinguishable from those provided by professional human translators.
SAP, a Germany-based B2B software company, has deployed this technology for its global support hotline. When a Japanese customer describes a complex ERP issue in Japanese, the system not only translates it into German in real time for the agent to understand, but also automatically retains technical terms (e.g., "SAP HANA database") in their original form to avoid ambiguity. The entire process has a latency of less than 500 milliseconds, with the customer barely noticing any translation delay.
GlobalConnect's real-time translation module further integrates contextual understanding: if a customer says "I'm in a hurry," the system automatically condenses the agent's response into brief instructions instead of performing a word-for-word translation. Industry analysts believe that by 2025, real-time speech translation will become a standard feature of call center cloud platforms, especially benefiting travel, finance, and cross-border e-commerce industries.