Still correcting the same phrase in every AI translation?

If the same colleague fixes the same product term in every translated email, that correction belongs somewhere besides their inbox. DeepL's current documentation says glossaries specify preferred translations for words and short phrases, adapting entries to the target language's grammar. On eligible team plans, a glossary can be shared; owners and editors can change it, and teammates see those changes. This is an existing feature, not a new launch or a result I have tested. Start with one phrase the team keeps correcting. Ask the person answering customers in that language to help choose it, then try the entry in a few real sentences before sharing it. A head-office preference can be consistent and still sound strange to the people reading it. I would keep this small. Fix the recurring nuisance before commissioning a complete terminology project. The glossary does not replace reading the rest of the message. Which correction does your team keep making that could be agreed once instead?