From proof of concept to recurring use
Digital Today reports that Amorepacific selected DeepL Voice after an internal proof of concept as collaboration increased across operations in the Americas, China, Japan and Southeast Asia. Its uses now include global meetings, multilingual conferences and employee training.
The reported capabilities include automatic language detection, translated subtitles in participants’ preferred languages, and a glossary for consistent brand names and workplace terminology. The coverage also identifies Microsoft Teams integration, single sign-on, centralized administration and audit logs. It does not establish which administrative controls Amorepacific has configured.
The important distinction is adoption versus effectiveness. The figures indicate increased use, but the report supplies no measurement period, total license count, translation-quality results or cost comparison. “95% license utilization” should not be read as 95% of employees—or 95% of all company meetings. A busy calendar is not, by itself, a return-on-investment calculation.
What Teams administrators should know
DeepL’s own product documentation confirms that Voice for Online Meetings supports Microsoft Teams, Zoom Meetings and Google Meet. It also distinguishes translated subtitles from voice-to-voice translation, advertising subtitles in more than 40 languages and voice translation in 30. Those are product-wide capabilities, not confirmation of the languages or audio features used at Amorepacific.
That distinction matters when defining a pilot: reading translated captions and listening to translated speech are different user experiences. Evaluate the mode employees actually need rather than treating every feature under the “voice translation” label as interchangeable.
There is also a separate data-handling question. DeepL states that meeting data is processed temporarily in memory and deleted when the call ends; transcription and translation data are not permanently stored by the service. It says data is encrypted in transit, is not used for model training, and can persist locally on participants’ devices. These are vendor statements, not verification of Amorepacific’s contractual or endpoint controls.
For an IT review, the practical implication is straightforward: ask about both service-side processing and participant-side copies. A retention statement about the translation service does not answer every question about what remains on a PC.
A useful evaluation checklist
For organizations considering a similar deployment, the following are suggested tests—not a reconstruction of Amorepacific’s undisclosed proof of concept:
- Use representative conversations. Include the languages, accents, speaking speeds and interruptions employees encounter.
- Test terminology deliberately. Check brand names, abbreviations and technical terms with bilingual reviewers.
- Measure comprehension, not just attendance. Ask participants to identify decisions and action items after a translated meeting.
- Review administration separately. Establish identity, access, audit and local-data requirements before expanding access.
- Define success in advance. Track active use alongside correction rates, participant feedback and costs.
Amorepacific plans to consider broader application of real-time translation under its longer-term “AI First” strategy, Digital Today reports. System architecture director So-young Park described the goal as enabling employees to collaborate in their preferred languages. No expansion timetable was disclosed.
The takeaway for enterprise IT is measured rather than magical: this is a reported example of sustained use worth examining, not evidence that language barriers have disappeared. A sensible next step is a tightly scoped evaluation with clear success criteria—not a license purchase justified by meeting counts alone.