Business Problem: When agents handle calls with customers who speak a different language, the auto-generated call transcript is only available in the language spoken during the call. Teams that operate across multiple regions — or where supervisors, and managers speak a different language than the customer — have no way to read or review those transcripts in a language they understand. This gap is further compounded by two additional challenges. First, regional language variations: a team in Spain reviewing calls in Latin American Spanish, or a team in Brazil reviewing calls in European Portuguese, may encounter translated content that doesn't reflect the correct regional terminology, tone, or vocabulary. Second, look-alike languages: languages that are closely related or share significant similarities — such as Spanish and Portuguese, Malay and Indonesian, Hindi and Urdu, or Serbian and Croatian — are frequently misidentified by automatic language detection. When the system selects the wrong base language, the transcript or translation can be inaccurate or misleading, eroding trust in the tool and creating extra manual work to verify what was actually said. Desired Outcome: Users should be able to manually select both the source language (what was spoken during the call) and the target language for translation directly within the interface. Giving users explicit control over language selection is especially critical for look-alike language pairs, where automatic detection is prone to error. This ensures transcripts are accurate, culturally appropriate, and trustworthy across all the languages and regions a business operates in.
