Updated
Updated · WIRED · Sep 22
WIRED Writer Builds 2 AI Editor Clones, Finding More Friction Than Productivity
Updated
Updated · WIRED · Sep 22

WIRED Writer Builds 2 AI Editor Clones, Finding More Friction Than Productivity

1 articles · Updated · WIRED · Sep 22

Summary

  • A WIRED reporter built AI versions of editors Brian Barrett and Sophie Kleeman to sharpen pitches, messages and drafts, but the experiment mostly showed the bots were awkward stand-ins rather than reliable workplace helpers.
  • Gemini produced a passable Brian Bot from public material, while Sophie Bot—trained on about 17,000 words of Slack chats—mimicked her voice more convincingly, even generating posts that fooled colleagues and briefly Sophie herself.
  • The gains were narrow: the bots sometimes suggested interview targets or polished messages, but they also hallucinated facts, repeated catchphrases and grew less useful as more data was added.
  • Both editors disliked their clones, saying the imitations exaggerated traits and missed human nuance, even when the bots sounded superficially accurate.
  • The test lands amid a broader push for AI coworkers—from startups selling autonomous agents to McKinsey's stated goal of matching 40,000 humans with 40,000 agents—while experts warn anthropomorphism and blurred accountability can distort judgment.

Insights

Does giving an AI a human name secretly trick our brains into trusting it enough to overlook dangerous workplace errors?
If companies clone an employee's personality from Slack data, do workers have any rights to stop their digital ghosts from staying employed?
When an AI clone of your boss gives a directive, who is legally responsible if following it leads to a catastrophic mistake?