How Different Languages Name and Frame AI Agents

Summary

The article examines how different languages name AI agents and how those choices shape expectations about what the software does and whom it serves. It begins with “agent,” a flexible term associated with action and delegation, alongside alternatives such as assistant, bot, worker, representative, copilot, and delegate. The 2027 Nashik Kumbh Mela project in India calls its personal AI an “envoy”; its examples have the system book lodging, find food, and guide pilgrims by voice, emphasizing that it acts on a user’s behalf. Portuguese creates an accidental association between the word for agent and “we,” while Arabic and Taiwanese usage can foreground the idea of a representative or intermediary. Turkey’s privacy regulator chose “AI intermediary” in a report on agentic AI after considering alternatives. Chinese usage commonly distinguishes a proxy acting for someone from an intelligent entity acting independently, with both terms found in current use by Stanford’s DigiChina researchers; the latter was already used for factory decision software in 2000. The article also notes humorous associations: an Israeli event used “Non-secret agents,” Turkish usage can suggest “spy,” and Duden lists that sense first for German. It concludes that naming is part of product design: “agent” signals broad flexibility, “envoy” suggests a mission, “representative” identifies whose interests are served, and “intelligent entity” implies greater autonomy.