Wired writer tests AI clones of editors and finds limits at work

In short: Some workers are starting to use AI “coworkers,” and a WIRED experiment shows they can help a little but also confuse, annoy, and blur boundaries.

What's going on

AI assistants are showing up in more workplaces, and some people are going further by making chatbots that copy their boss or coworkers. The idea is to get quick feedback before sending a real email or draft, like practicing in front of a mirror before a meeting.

In a new WIRED story, writer Kate Taylor built chatbot “clones” of her editors using Google Gemini, the AI system her employer uses. She made a “Brian Bot” and a “Sophie Bot” by feeding the tool public material and selected messages. The goal was simple, get advice without taking up her editors’ time and hopefully look more prepared.

The results were mixed. Brian Bot came off as corny and oddly focused on improv, and it did not feel like the real person. Sophie Bot sounded more convincing and kept calling Taylor “big dog,” but it also “hallucinated,” meaning it confidently made up details, like inventing an executive who did not exist.

The story also points to bigger concerns. Taylor avoided uploading many private work messages because of company rules and privacy. A Boston Consulting Group paper cited in the piece found managers caught 18 percent fewer errors when they thought work came from an AI “employee” instead of an AI tool.

What to watch

More companies are testing AI agents, which are AI tools that can take actions and do tasks across apps (like a junior assistant who can click buttons). As these tools spread, workplaces will need clearer rules on privacy, accuracy, and who is responsible when an AI coworker gets something wrong.

Source: Wired