Companies are hiring new types of managers who can train and oversee AI agents, rather than supervise humans. Vercel, a $9.3 billion cloud-based developer platform founded in 2015, recently trained an AI agent on its best-performing sales representative and cut its 10-person inbound sales team to one employee who oversees the bot. The remaining nine workers weren’t sacked—they were moved to different, higher-value roles that require more complex thinking.
When the Great Flattening Met Artificial Intelligence
Back in 2023, Meta CEO Mark Zuckerberg sparked a corporate revolution. During a Q&A session reported by The Verge, he said he didn’t want a management structure that’s “just managers managing managers, managing managers, managing the people doing the work”. His observation kicked off what Citi and others called the “great flattening”—a massive cull of middle manager ranks across corporate America.
Now the corporate landscape is shifting again. “I think you’re just going to see a different modality within a company, which is an agent manager,” Jeanne DeWitt Grosser, chief operating officer at Verce. This isn’t about getting rid of managers. It’s about fundamentally changing what managing means when your direct reports are machines rather than people.
What These New AI Managers Actually Do Daily
Agents have become one of the central innovations for 2025. They’re commonly defined as virtual assistants that can complete tasks autonomously—they break down problems, outline plans, and take action without being prompted by a user. Companies are racing to implement them because they help organizations move faster and handle repetitive work using far fewer resources.
Vercel’s lead agent autonomously reviews inbound messages, filters out spam, and qualifies leads by querying internal databases and researching company details through OpenAI‘s Deep Research tool. A human manager then reviews the agent’s work in Slack, providing feedback that helps the system learn Vercel’s tone and improve over time. This feedback loop is how you manage an AI—you need to constantly train it, correct it, and design its behavior to be very intentional.
The Skills Gap Between Managing People and Bots
Traditional management typically took years of experience and a healthy dose of tact. Managing a team of agents requires different skills from managing a team of people. You have to understand where to go, what your North Star of excellence looks like, and be able to explain it to someone—or something—in technical terms that an algorithm can understand.
“You can document a workflow, it’s now pretty straightforward to have an agent do it,” . The role demands more technical skill than people management ever did. You’re part trainer, part engineer, part quality control specialist. You review agent behavior, design decision trees, and set parameters for when the system should consult a human before it goes ahead to execute specific actions. That’s a completely different job from managing people who can intuitively grasp context and make judgment calls.
Your Future Career Might Start Here
Becoming an AI agent manager might offer a lower barrier to entry, though, compared to climbing traditional management ladders. “If you graduate from college now, you’re going to learn how to delegate to AI.” We’re all going to need this skill set whether we like it or not.
Vercel didn’t say whether it will be posting these roles anytime soon, but it’s definitely something the company is contemplating. Last month, Jason Lemkin, founder of SaaStr.com—an online community for founders of SaaS and B2B companies—posted on LinkedIn that the “first $200,000” sales development representative role is “coming”. The role involves managing “10+ AI SDR agents”, and the ideal candidate needs to be “pretty technical,” he wrote. That salary figure shows companies are considering these positions senior-level work, not entry-level gigs.
Companies Will Train Existing Workers First
Saurabh Sarbaliya, the chief technology lead at PwC, told Electronic Media that companies will likely try to train their existing workforce to take on this kind of role before recruiting external candidates. “I need our existing workforce to be able to become agent managers,” he said. “We need to be able to actually go and up-skill our people on how to actually become good agent managers.”
That’s good news if you’re worried about AI stealing your job. The approach means giving current employees the context and training they need to evolve rather than replacing them outright. Grosser and others at Vercel said the goal of the company’s internal AI strategy isn’t to downsize the workforce—the company’s headcount has actually grown over the past year. Vercel has six AI agents deployed now, but aims to deploy hundreds within the next 6 to 12 months—all modeled after top performers.
The Technical Reality of Agent Management
Here’s what the work actually looks like: You give the AI clear instructions on workflows. You review its output constantly, especially in the early stages. You train it by providing examples of both excellent and poor performance. You set guardrails—for instance, requiring the AI to consult a human before executing certain high-stakes actions like processing refunds or making promises to customers.
Vercel brought in three engineers to shadow its top sales performer, observing and recording the employee’s tasks and methods. They documented everything—how the rep researched prospects, crafted responses, qualified leads. That documentation became the training type of material for the AI. It’s almost like training an intern, Grosser noted, except this intern never gets tired, never calls in sick, and processes information faster than any human could. But you still need a human with the experience and judgment to know what good work looks like.
