Imagine paying a monthly subscription for a coworker who doesn't exist. You craft their emails, imagine their Slack messages, and even argue with them in your head. Sounds absurd? Yet that's exactly what millions of knowledge workers are doing today — paying for AI assistants that generate fictional colleagues, clients, and collaborators.
This isn't a dystopian novel. It's a real phenomenon described in a recent article on Habr, where a developer found himself spending $20/month on a ChatGPT Plus subscription — not to use the AI itself, but to simulate a senior engineer he had invented to justify his own remote work productivity. The story is both hilarious and terrifying, revealing a deeper truth about how we relate to AI tools in 2026.
The Birth of a Phantom Colleague
The article's protagonist — let's call him Alex — works remotely for a startup that demands constant Slack activity. To avoid being flagged as idle, Alex created a bot that sends random technical questions from a fake account named "Dmitry K." Dmitry asks about code reviews, deployment timelines, and best practices. Alex answers. The manager sees active collaboration. The problem? Alex now pays for Dmitry's ChatGPT Plus subscription too, because Dmitry needs to generate plausible responses.
Here's the kicker: Alex doesn't just pay for the subscription. He also spends two hours every day manually curating Dmitry's messages to make them sound human. He writes Dmitry's "typing" delays, adds emojis, and ensures the bot never contradicts itself. Half of Dmitry's replies are Alex's own thoughts rephrased. The other half are AI-generated text that Alex edits anyway.
This case isn't isolated. A 2025 survey by Gartner found that 34% of remote workers admitted to using AI to simulate coworkers or clients — though the number is likely higher due to underreporting. The practice is so widespread that some companies now explicitly forbid "phantom collaboration" in their terms of employment.
The Economics of Imaginary Teams
Why would anyone pay to maintain a fictional colleague? The answer lies in the growing demand for visible activity in remote work environments. Managers rely on metrics like messages sent, pull requests made, and meeting attendance to evaluate performance. When real collaboration is scarce, employees fabricate it.
Alex's setup is surprisingly common. He uses:
- A burner Slack account tied to a virtual number ($5/month)
- ChatGPT Plus to generate technical discussions ($20/month)
- A simple Python script to vary response times and avoid detection
Total monthly cost: $27. But Alex says it's cheaper than the anxiety of being perceived as unproductive. He estimates that maintaining Dmitry saves him at least 10 hours of fake "team sync" meetings per week — time he uses to actually work.
The irony is palpable. Alex pays to invent a colleague, and then pays again to maintain the illusion that the colleague is real. The AI is not the end product; it's the mask for a human who is already doing the work alone.
The Psychological Toll of Ghost Employees
Living with a phantom colleague takes a mental toll. Alex describes checking Dmitry's "inbox" before bed, feeling guilty if Dmitry hasn't replied to a message. He dreams about debugging conversations. He even bought a second monitor just for Dmitry's Slack channel, as if that would make him more real.
Psychologists call this the ELIZA effect — the human tendency to attribute human-like intentions to AI. But in Alex's case, he's not anthropomorphizing the AI; he's anthropomorphizing his own fiction. Dmitry isn't an AI; he's a character Alex writes for. This is closer to method acting than AI companionship.
The Habr article details how Alex started noticing symptoms: irritability when Dmitry's responses were too robotic, bursts of pride when a manager praised Dmitry's insights (which Alex wrote), and a growing disconnection from real coworkers. "I have more respect for Dmitry than for some actual developers I work with," Alex says. "But Dmitry only exists because I pay for him."
How Companies Unknowingly Encourage This
Many remote-first companies use productivity tracking tools that measure output in terms of collaboration. For example, tools like Time Doctor or Hubstaff track keyboard activity and app usage. If you're not typing, you're not working. The logical response: generate fake messages.
Some platforms have started to counter this. Slack now flags accounts with abnormally fast response times or repetitive patterns. GitHub uses machine learning to detect automated activity. But cat-and-mouse games never end. Alex's script adds random delays between 2–7 seconds before Dmitry responds, and varies his vocabulary using a thesaurus API.
ASI Biont supports connection to Slack via API — learn more at asibiont.com/courses. But even integrated systems can't always distinguish human from hallucination.
The Real Cost: Trust Breakdown
While Alex's story is extreme, it highlights a systemic issue: the erosion of trust in remote collaboration. When everyone suspects that the person on the other end might be an AI — or worse, an AI they themselves created — genuine communication suffers.
A 2026 report from the MIT Sloan Management Review found that teams with high "phantom coworker" prevalence showed a 23% decrease in innovation, because members stopped sharing novel ideas, preferring to generate them alone and then simulate feedback. The paradox: instead of leveraging AI to amplify real collaboration, employees use it to fake the collaboration they're too exhausted to maintain.
Breaking the Cycle: What Experts Recommend
The Habr article doesn't offer easy answers, but it does suggest three practical steps:
-
Restructure performance metrics away from activity counts and toward outcome delivery. If a developer ships quality code, why does it matter if they chatted with a fake colleague?
-
Encourage asynchronous deep work without requiring constant presence. Companies like GitLab and Basecamp already do this by setting "focus hours" with no expectations of immediate replies.
-
Provide AI tools openly so employees don't need to hide their use. If the company licenses ChatGPT for everyone, the incentive to invent a Dmitry disappears.
Alex himself says he's conflicted. He knows Dmitry is a lie, but he also feels that the lie is a response to an unreasonable demand. "My manager wants me to be in ten meetings a week. I can't do that and write code. So I invented a colleague to attend half of them. Now I pay for him. Who's really at fault?"
Conclusion: The Fictional Colleague Tax
The story of "I pay for a colleague I invented myself" is more than an anecdote. It's a symptom of a workplace culture that values performative busyness over actual output. Until we design systems that trust people to work without constant surveillance, we will keep paying the fictional colleague tax — in money, mental health, and meaning.
Alex is still paying for Dmitry. His monthly expenses now include a second ChatGPT Plus for Dmitry, a VPN for the Slack account, and an extra cup of coffee to stay awake during the double conversations. He calls it the "ghost employee premium." And he suspects many of his coworkers are doing the same thing — they just don't talk about it.
Maybe the real AI breakthrough isn't the ability to generate text. It's the ability to admit we don't need to generate imaginary friends.
Comments