The rise of the bot boss
Imagine learning that the system deciding whether your work meets the mark isn't a manager—it's a bot.
As artificial intelligence becomes more common in the workplace, companies are increasingly using software to monitor and evaluate work. Whether it's AI measuring customer service performance or flagging unusual employee activity, workers are encountering more technologies that judge their work rather than help them do it.
"We often talk about AI as a tool, but increasingly we're seeing systems that provide oversight, quality assurance or monitoring functions,” said Jason Thatcher, professor of information systems at the Leeds School of Business and co-author of the study, published in January 2026 in MIS Quarterly.
The study found that when bots are given authority over people, workers push back.
"For many people, the idea of a bot evaluating their work feels uncomfortable,” he said. “You're a human expert performing a task, and suddenly a software system has the authority to say yes or no. That changes how people experience work."
Jason Thatcher
To understand how people respond when software gains that kind of authority, Thatcher and his co-authors, Nadine Kathrin Ostern and Marek Kowalkiewicz of Queensland University of Technology in Brisbane, Australia; Likoebe M. Maruping of the J. Mack Robinson College of Business at Georgia State University; and Jörg Weking of the Technical University of Munich, examined the 2010 rollout of ClueBot, Wikipedia's anti-vandalism bot.
Because Wikipedia relies on volunteers throughout the world to create and edit articles, ClueBot was designed to automatically remove harmful edits, such as spam, profanity or intentionally false information. At the time, it was catching about 70% of problematic edits identified as vandalism.
But unlike a spell-checker or other automated tool, ClueBot went beyond helping contributors do their work. It evaluated their edits and sometimes overruled them.
By analyzing weeks of public debates among editors, developers and Wikipedia leaders during the bot's approval process, the researchers examined what happened when human judgment challenged a machine's decisions. Contributors pushed back, causing developers to refine its rules, limit its authority and create formal ways for people to challenge its judgments.
What emerged, Thatcher said, was a process of humans figuring out how to work with machines. "Rather than bot regulation, it was actually a bot-human-bot negotiation," he said. In other words, contributors couldn’t change the bot themselves, but they could appeal to the people who could.
"What was cool about this study, that got me so excited, was people pushed back," Thatcher said. "While they couldn't get the bot to change, they could go to that committee that made the decisions about what the bot could or could not do, who would then go back to the developers, who would tweak the bot."
Human oversight
The lesson goes beyond Wikipedia, Thatcher said. As organizations increasingly use AI to monitor and review work or enforce standards, they’ll also need to think about what happens when those systems get things wrong and who gets a say in correcting them.
"Human voice still matters," Thatcher said. "The content creator, or the Wikipedia contributor, still has a voice in how the bots operate when the bots don't operate properly."
That's what happened on Wikipedia. Over time, those efforts led to lower error rates, better ways to review the bot's work and formal ways for humans to influence its decisions.
The future workplace may include more bots evaluating human work, Thatcher said, but the most successful organizations won't be the ones that remove people from the process. They'll be the ones that leave room for employees to challenge, correct and improve the technology.
"We have to keep humans in the loop," Thatcher said. "Users and experts need to stay engaged so we're building systems that solve problems rather than create them."