ENGINEER CREATES AI VERSIONS OF COWORKERS

The experiment trained large language models on email, chat messages, and meeting transcripts from real employees. Brian Bot emerged with a fixation on improv comedy, while Sophie Bot repeatedly addressed the engineer as "big dog"—speech patterns lifted directly from workplace interactions. The project highlights how AI systems absorb and amplify behavioral traits from training data. Even limited datasets can produce eerily accurate personality replicas, raising questions about workplace AI applications. The engineer shared the results to illustrate both the capabilities and risks of personalized AI systems. As companies explore AI assistants and automation, the experiment underscores how these tools can encode human quirks—sometimes problematically. The work demonstrates that creating functional AI clones requires careful consideration of what behaviors get replicated and why. It's a practical reminder that training data shapes outputs in unexpected ways.