You trained to write, to code, to design, to teach. Instead you now spend a chunk of every day setting expectations, delegating tasks, reviewing output, demanding revisions and stepping in when things go wrong. That is a job description for a manager. The catch, laid out in a Business Insider report this week, is that a growing number of workers are doing exactly this work without the title and without the pay. They are not managing people. They are managing agents.
The numbers behind the anecdotes are striking. In a Boston Consulting Group survey of nearly 12,000 employees, 47 percent said they now spend more time managing and directing AI than doing the work themselves, and 67 percent said the tools had swallowed the simpler tasks, leaving them with the harder ones. Sinda Khenine, a software engineer at Electrolux, says agents let her produce far more, but the effort has shifted from the engineering problem to the coordination itself. "It's like a loop," she says. "You need to monitor the results. You need to decide if you need to rerun or finish the job." She does not think this entitles her to a raise. She does think outsiders badly underestimate the labour involved.
What makes this more than a grumble is that the responsibility is arriving without the recognition that has always come with it. Data from the workforce analytics firm Visier shows 54 percent of full-time US employees say their role changed significantly because of AI in the past two years, yet of those, only 17 percent were promoted. Previous research using the same data found individual contributors earn about a third less on average than team leaders at the same firm. "If you are a manager of people, you have the financial advantage," Visier's Andrea Derler notes. "If you are managing agents, nothing changes for you. You still have the same role."
There is a quieter cost too. The skills gap is widening faster than training can close it: 72 percent of BCG's respondents said AI had considerably changed what was expected of them, but only 36 percent felt they had been adequately reskilled, a figure that has stayed flat across successive surveys even as usage climbs. BCG's David Martin describes the result among engineers and marketers as AI "brain fry," the fatigue of people who never chose to manage suddenly owning a roster of unpredictable digital reports. Much of the corporate training on offer, he adds, is generic classroom material that does little for someone learning to judge an agent's work in real time.
Some workers have made peace with it, even found identity in it. A product marketing consultant told Business Insider she names her agents, writes them configurations and speaks to them "in a very human way," trying to move her judgment out of her own head and into systems that can carry it. Others were conscripted. An American professor teaching in China described being told to train students to use AI while also policing them for using it on assignments, and to teach "ethical" use that nobody had defined. "That's dumped on me to manage," she said, "because no one knows how to use it, why it's useful, or what it's actually going to do."
The compensation experts see the gap and mostly shrug. Maria Colacurcio, who runs a pay-equity software firm, says she is "100 percent" watching individual contributors absorb "management light" duties, and that when companies do start rewarding it, the money will likely arrive as bonuses rather than higher base pay. Ernst & Young is handing out $100 million in bonuses for the human skills that make AI work, but that is the exception. For now the arrangement is lopsided in a familiar way: the output of the agents flows up to the organisation, the responsibility for that output settles on the worker, and the pay stub stays exactly where it was.