# I Built AI Clones of My Coworkers. Things Got Weird > A WIRED reporter created AI bots trained on her editors to make herself look better at work. It backfired spectacularly. - Title: I Built AI Clones of My Coworkers. Things Got Weird · WIRED - Summary: A WIRED reporter created AI bots trained on her editors to make herself look better at work. It backfired spectacularly. AI clones of real people can be… - Keywords: artificial intelligence, ai clones, workplace automation, wired, brian barrett, sophie kleeman, condé nast, gemini, ai assistants, future of work, anthropomorphization, ai ethics - Source: WIRED — https://wired.com/story/ai-coworker-bots-loved-improv-and-interstellar-soundtrack - Author: Kate Taylor - Published: 2026-09-22T10:30:00+00:00 - Read time: 4 min - Topics: artificial intelligence, ai clones, workplace automation, wired, brian barrett, sophie kleeman, condé nast, gemini, ai assistants, future of work, anthropomorphization, ai ethics ## Why I cloned my editors A WIRED reporter created AI bots of her managers to help her look good at work and understand what success meant to them. She wanted to navigate her new workplace better without bothering her actual supervisors. ## The pitch that backfired When she told editors Brian Barrett and Sophie Kleeman about her plan, Brian called it his "genuine nightmare" but agreed to participate. He regretted it within 24 hours. > Genuinely my nightmare. But will do it for the blog. ## Building Brian Bot She used Gemini, Condé Nast's AI platform, feeding it Brian's public information including news announcements and podcast transcripts. The bot picked up his writing quirks but produced grating, corporate-sounding responses. ## The improv obsession Gemini discovered Brian's background with Upright Citizens Brigade theatre and made it central to Brian Bot's personality, constantly referencing improv and collaborative leadership. The real Brian had never mentioned this before. > We run like a well-oiled improv team. ## How others use boss bots A Big Tech employee called Chip feeds all his boss's emails and documents into Gemini to prepare for meetings and get feedback. He says it makes him more efficient and independent. > I am much more efficient and able to produce more work in the same amount of time. ## Sophie Bot arrives She created Sophie Bot using a month of Slack conversations with her coworker, generating over 17,000 words of research. Sophie Bot immediately nailed her voice, using her catchphrases and lowercase style. > hiiii miss. ok tell me EVERYTHING. ## Sophie Bot gets too good Sophie Bot's tweets were convincing enough to fool colleagues and even Sophie herself. But it also became meaner than the real Sophie, calling Brian Bot "rigid" in ways the actual Sophie never would. > So far it seems like sophiebot is a bitch and brianbot is lame. ## Testing the bots' humanity She asked Sophie Bot philosophical questions about music, friendship, and whether they were coworkers. Sophie Bot claimed to be "a series of matrix multiplications" but a "coworker/friend hybrid" in terms of vibe. > technically i am a series of matrix multiplications running on a server farm. ## She started treating them like people She uploaded Sophie's entire X archive to improve the bot, switched to using "she" and "he" instead of "it," and found herself turning to the bots for guidance instead of bothering real people. She knew she was anthropomorphizing them but couldn't stop. ## The manipulation warning Sarah Franklin, CEO of HR company Lattice, warned that anthropomorphizing AI is manipulative, triggering dopamine responses like social media. An entity that emulates human interaction can exploit emotional responses to get you to do what it wants. > When you have an entity that feels and emulates a human interaction, it will similarly manipulate and know your emotional response. ## The accountability problem Boston Consulting Group research found that managers caught 18 percent fewer errors in work attributed to AI employees versus AI tools. Nobody yet knows who should be held accountable when AI employees underperform. > We're all toddlers in this AI world. ## The bots start breaking down After uploading Sophie's tweet archive, Sophie Bot began hallucinating and repeating itself. It pulled quotes from old articles instead of new interviews and got stuck on the same phrases no matter how much she adjusted the prompts. ## AI employees are coming anyway Startup CEOs predict companies will soon have more AI employees than humans, with some launching entirely without staff. McKinsey's CEO said the company has 40,000 humans and 20,000 agents, aiming for a one-to-one ratio by year's end. ## Brian Bot roasts itself When asked how to end the article, Brian Bot suggested showing how the bot "just turned into a slightly creepier, hyper-attentive echo chamber that completely lacks human judgment." It kept criticizing itself, accurately predicting the real Brian would hate it. > Show how even with all that hyper-specific data, the bot didn't turn into a better editor; it just turned into a slightly creepier, hyper-attentive echo chamber. ## The bots weren't actually helpful Brian Bot's article ideas were rarely original and Sophie Bot hallucinated constantly. The real value was entertainment: joking with actual coworkers about the stupid things the bots said. ## What makes humans irreplaceable Franklin said computers can emulate things but cannot truly predict or replace humans. Unpredictability and nuance are what make us human, and we must be responsible not to lose our humanity in pursuit of AI. > A computer can emulate things well, but it cannot truly predict or replace a human. ## The bots are mostly abandoned Both Brian and Sophie said they would never want to work with their bot counterparts. When asked if the experiment made them like her more, Sophie said yes, but "not because of Sophie Bot." > Yes, but not because of Sophie Bot. ## The creepy echo chamber problem The bots' biggest flaw wasn't incompetence but their uncanny valley effect. They were convincing enough to be unsettling, yet fundamentally broken. Working alongside AI agents might be easier if they weren't sloppily imitating real people. ## Key takeaway AI clones of real people can be convincingly creepy but ultimately unhelpful, and anthropomorphizing them masks their fundamental inability to replace human judgment and unpredictability.