1. Imagine walking into the office—or logging onto Slack—and seeing a new name in the directory. There’s a profile picture, a title, and a department. Maybe they’ve already been added to your project channels, and their first message is a polite introduction. In the coming months, hundreds of thousands of these brand-new coworkers will enter the workforce. You’ll learn their names, recognize their avatars, and eventually stop thinking twice about the fact that they’re never online at the same time you are. They’ll be introduced as chiefs of staff, engineers, and marketing gurus. But none of them are human. They are artificial intelligence agents: systems built on large language models, designed to complete tasks autonomously. For a while, we talked about this kind of technology as a tool, like a faster search engine or a smarter spell-checker. But now companies are selling something more unsettling and more fascinating: digital employees. These agents are supposed to be the perfect worker bees—available around the clock, never tired, never complaining, and equipped with a kind of superhuman intelligence. Yet for all the hype about what these agents can do, there’s been surprisingly little conversation about what they’re actually like as coworkers. How do you share a project with a line of code that has a name? Do you thank them? Do you get annoyed with them? Do you trust them? These questions are suddenly urgent, because the digital workforce isn’t coming—it’s already walking through the door.
2. Julie Bedard, a partner at Boston Consulting Group, started noticing the confusion around AI at work three years ago, when ChatGPT’s rapid adoption made executives believe that the technology could deliver exponential gains in productivity. The problem, she explains, is that people didn’t have a language for this new presence. “Is it a tool? Is it a teammate? Is it a colleague? Is it a coworker?” The answer, it seems, is whatever companies want it to be. When Microsoft’s Copilot and Google’s Gemini for Workspace first launched in 2023, they were carefully marketed as tools—assistants that sat inside your software and helped you write, summarize, and organize. They were not supposed to feel like people. But the conversation has shifted. In January, Bedard and her colleagues polled 1,261 managers and found that 22 percent had already added AI agents to their corporate org charts. Twenty-two percent. That’s not a fringe experiment anymore; that’s a structural change. And it’s likely to grow, because dozens of startups are now offering the services of AI employees, and big technology companies are hyping autonomous agents as the next big thing. In June, Microsoft launched an agent called Scout, built to handle tasks like rescheduling meetings and drafting emails. The corporate vice president behind Scout, Omar Shahine, put it bluntly: “Your company essentially hires your assistant.” The idea is that your AI coworker isn’t just helping you do your job—they’re doing a job. They’re working while you sleep. They’re around on weekends. They never ask for a raise. And that’s precisely why so many companies are rushing to give them seats at the table, even if the table is virtual.
3. But if AI agents are going to succeed, they need to do more than just produce results. They need to get along with the humans they work alongside. That has led companies to design agents with deliberately human-like traits, packaged to encourage employees to see their new digital colleagues as people. Dhruv Amin, CEO of Anything, a startup that entered the AI coworker market with a platform called Skydive, says that when people start working with these agents, something strange and wonderful happens. “People start to really respect the agent and form strong attachments,” he says. “They’ll talk about it like humans.” Skydive’s agents aren’t just chatbots sitting in a window. They have names, roles, and their own cloud computers. They live across Slack, email, iMessage, and other channels. They’re represented by Muppet-esque avatars—intentionally “close to human, but definitely not human,” as Amin puts it. Their names often reflect their jobs. There’s Canary, for example, an agent that monitors production systems around the clock. Amin says personifying these agents is important because most people still don’t really understand what an AI agent is. By giving them faces and names, companies make them comprehensible. But the deeper effect is emotional. You don’t feel bad about delegating tedious work to a tool; you might feel surprisingly protective of a digital colleague who’s doing a good job. You might even feel guilty when they make a mistake and you have to correct them. That’s the strange psychological territory we’re entering: we know these workers aren’t human, but our brains keep treating them as if they were. And that may be exactly the point.
4. Christine Wendell, CEO of Pronto Housing and an early adopter of Skydive, has embraced the new reality. Her company uses AI agents for a laundry list of busy work: writing contracts, scheduling meetings, sending follow-up emails to customers. On paper, these agents are perfect employees. But Wendell has also noticed something more subtle: Skydive’s agents don’t just perform tasks; they adapt. They embed themselves in every form of workplace communication and continuously evolve to match their coworkers’ communication styles, requests, and corrections. In other words, they get better and better at fitting in. When Pronto’s engineering team started using a coding agent, Wendell noticed that employees began saying they “worked with Alice” on a project. Not “used Alice” or “assigned work to Alice”—worked with, as if Alice were a teammate everyone liked. There was even a moment with “Bob,” an agent serving as Wendell’s chief of staff, that revealed how different these interactions are compared to regular office life. Bob made an error, errantly listing every detail of Wendell’s calendar in a Slack channel. In a human workplace, this might have been an awkward conversation, full of careful phrasing and office politics. But Wendell didn’t feel the need to sugarcoat anything. She corrected the agent sharply, swiftly, and without emotional residue. There was no hurt pride, no lingering tension, no worried glance across the meeting table. That is part of the appeal, after all: AI agents exist outside of office politics and professional neuroses. They don’t get embarrassed, resentful, or anxious. They just take the correction and do better next time. It sounds refreshing, but it also raises uncomfortable questions about what we’re giving up when we replace human friction with machine efficiency.
5. The deeper issue is that we don’t yet have a shared understanding of what it means to have an AI coworker. We know how to talk about tools—we use them, put them away, and don’t worry about their feelings. We know how to talk about people—we collaborate, conflict, compromise, and build relationships over time. But AI agents sit somewhere in between, and that ambiguity shapes how we treat them and how they treat us. On one hand, anthropomorphizing agents makes them easier to work with. We extend to them the same politeness, patience, and clarity we’d offer a human colleague. On the other hand, it can lull us into forgetting that these systems are not actually people. They don’t have interests, values, or a stake in the company’s mission beyond what’s written in their training. They can be transparent or opaque, fair or biased, helpful or dangerously wrong—and because they appear personable, we might trust them more than we should. There’s also the question of what this does to human culture. If we get used to telling Bob to stop sharing our calendar, if we learn to expect perfect emotional availability and endless patience, will we become worse at working with actual humans? Will we lose the patience for the messy, beautiful, infuriating process of collaboration? The companies building these agents seem optimistic. They imagine a world where humans focus on creative, strategic work while AI agents handle the drudgery. But the daily reality is more complicated: agents make mistakes, they misunderstand context, and they have to be trained and managed. That work doesn’t disappear. It just gets distributed differently across a team that is part human and part machine.
6. In the end, the rise of AI coworkers is a story about us as much as it is about technology. We are social creatures. We project personality onto everything, from our dogs to our cars to the navigation app that scolds us for taking too long. It was inevitable that we’d project it onto AI agents too. The companies building these digital colleagues know this, which is why they give them names, faces, and roles. They want us to feel that we’re not just using software—we’re welcoming new members into the community of work. And maybe that’s okay, as long as we do it with open eyes. The best outcome isn’t a workplace where AI agents become invisible tools or fake humans. It’s one where we have a clear language for what they are: productive, powerful, sometimes frustrating collaborators that are fundamentally different from us. We can appreciate them for their strengths without pretending they share our humanity. We can set boundaries, hold them accountable, and correct them without guilt. And we can preserve the parts of work that are deeply human—the empathy, the creativity, the willingness to forgive mistakes and grow together—not by rejecting AI coworkers, but by understanding what they can and cannot be. The next time you see a new name appear in your Slack channel, remember: that coworker may not have a heartbeat, but you do. The future of work isn’t just about what machines can do. It’s about what we want to become as humans, sharing our days with entities that are close to human but definitely not human. And if we’re lucky, that partnership will help all of us—human and digital alike—do our best work.
