“My coworker’s emails sound like a robot wrote them.” “My boss has embraced AI, a little too enthusiastically.” “I no longer know who I’m dealing with. Is this my colleague or their GPT?” “I prefer dealing with gen AI because there’s no drama. I almost don’t want real coworkers anymore.” These are the sorts of concerns and questions I’ve been hearing from people over the last several months. As someone who writes and talks about tricky interpersonal dynamics at work, I get a lot of unsolicited complaints and requests for advice, and AI has officially entered the room of workplace conflicts. Employers, tech experts, and futurists are encouraging people to use AI in our jobs, promising more ideas, higher productivity, and greater efficiency. At the same time, the use of AI is raising all sorts of questions about: Authenticity (“If my coworker uses AI to write me an email, am I actually talking to them, or to a machine?”) Trust (“Can I trust the work my colleague sends me if I know they used AI to create it?”) What it means to truly connect with our colleagues (“It now feels easier to collaborate with AI than with my colleagues.”) But the use of AI, especially as an intermediary in—or even substitute for—work relationships, has the potential to damage the very connections we need at work. I definitely see the allure of using gen AI to craft an unemotional response to a passive-aggressive coworker, or to quickly pull together a report for an impatient colleague, or to explain to an overly competitive colleague why they can’t take credit for an entire project. But I also have serious hesitations about uses like these. In this article, I’ll lay out what worries me most when it comes to AI and our working relationships, and then share some guidelines that you can use in your organization or with your team to determine when and how to collaborate with AI, without creating more problems than it’s worth. The Costs of Bringing AI into Our Work Relationships In the AI-fueled quest for productivity and creativity, we haven’t spent enough time thinking about what AI does to workplace dynamics. Here are some of the ways I see AI presenting problems. It increases our cognitive load. When I don’t know whether I’m dealing with a human, a human using AI, or just AI, I need to spend extra time and energy figuring out why their tone shifted or why they used a certain word or phrase that doesn’t sound like them. And I’m often left wondering. This minor emotional labor adds up: Am I talking to a person or am I not? What does it mean if I think I’m talking to a human but it’s actually a human using AI to write to me? Should I respond differently? Should I ask them to stop? I spoke with Alexandra Samuel, a tech expert who has been exploring the AI-human connection, most recently in her podcast Me + Viv. Samuel told me: “The fact that it could be AI makes us second guess our relationships.” Research supports this: When people suspect (or are told that) AI was used to write a message, they question the relationship itself and whether the other person is putting in enough effort. It creates “workslop,” not efficiency. In a wildly popular article published last year called “AI-Generated ‘Workslop’ Is Destroying Productivity,” researchers explained one reason why our colleagues’ use of AI can get on our nerves: because it’s creating useless noise in workplaces where there is already too much noise—performative busy work, unnecessary meetings, emails that could have been Slack messages. Do we really want to add robotic interactions to that list? As the article authors explain: “When coworkers receive workslop, they are often required to take on the burden of decoding the content, inferring missed or false context. A cascade of effortful and complex decision-making processes may follow, including rework and uncomfortable exchanges with colleagues.” In other words, AI promises efficiency but, when not used thoughtfully, it intensifies work rather than reducing it, and it creates inefficiencies that strain relationships. Your employee may have been able to write that report in record time thanks to gen AI, but now you or another one of their colleagues has to figure out whether it’s valuable, check sources, and spend time explaining to the employee why the report didn’t hit the mark. It erodes trust. The “workslop” research revealed some troubling findings about how AI-generated content affects our perception of colleagues. As the study authors write, “Approximately half of the people we surveyed viewed colleagues who sent workslop as less creative, capable, and reliable than they did before receiving the output. Forty-two percent saw them as less trustworthy, and 37% saw that colleague as less intelligent.” Even more concerning, they say a third of respondents are reporting the “workslop” to others, “potentially eroding trust between sender and receiver” and that same number of people say they are less likely to want to work with that colleague again. Other researchers found something similar: there are social costs to using AI to handle your part of a collaboration. People who were thought to be using AI were considered less competent and lazier than those who were receiving help from other tools. It eliminates the friction we actually need. I’m a big believer that we need tension and messiness in order to do good work. The friction, the back-and-forth, even the occasional miscommunication—these aren’t bugs in the system, they’re features. They’re how we collaborate, build understanding, and create something better together. Professor Linda Hill from Harvard Business School refers to this in her research as “creative abrasion” or the ability to generate new ideas through discussion and debate. Professors Constance Hadley and Sarah Wright recently surveyed over 1,500 knowledge workers who were regular AI users and they saw that many employees were using AI as a personal coach to help them navigate tricky interpersonal situations. Participants described using AI to diagnose conflict, script feedback delivery, and role play tough conversations. While this might facilitate smoother interpersonal interactions, when we outsource the management of our interpersonal dynamics to AI, we may lose that important friction that often leads to breakthrough ideas. We lose the productive disagreements that help us see problems from new angles. We lose the human messiness that actually makes collaboration work. Hadley pointed out another potential downside: “We could lose our personal capability to diagnose and address conflict, which is not a skill loss any of us can afford in these polarized times.” It prevents us from building real relationships. That messiness of human interaction is how we build relationships. When things go completely smoothly with someone, it may feel good but we don’t necessarily grow or bond. It’s in awkward moments, expressions of vulnerability, and the clearing up of miscommunications that we get to know others, and ourselves, better. As Alexandra Samuel pointed out to me, if we start coping with email overload by offloading the relationship work to AI, we may save time but we may also put distance between us and our coworkers. Some experts, like David Rock of the NeuroLeadership Institute, are calling out that impact on our brains as well. He pointed out in a recent article that AI reduces brain synchrony, which facilitates social interaction and communication. The more employees turn to AI instead of other people, the more likely that our relationships with others will weaken. It may be training us to be less civil. One of the upsides for some people of collaborating with AI over human colleagues is that they can spend less time on niceties and concerns about their collaborator’s reactions or feelings. Even if we say “thank you” to ChatGPT or wish Gemini a “good morning,” the reality is that we don’t have to. Plus, as has been well documented now, most gen AI programs tell us what we want to hear. And, if it doesn’t, we can always change the directions. What does it mean if we can reprogram an AI collaborator when we don’t get the results we want? Will we expect the same from our human collaborators? Will we have less patience when they don’t agree? How we behave with AI is likely to spill over into how we behave with human coworkers. When AI Can Actually Help Of course, I’m not arguing that AI has no place in our work lives, or even in our work relationships. But I do think we need to be much more intentional about when and how we use it, especially when it comes to relationships with colleagues. Kate Niederhoffer, one of the coauthors of the workslop study, told Marketplace, “The most important thing is just recognize that we’re still at work, it’s changing the way we work, but we’re still working with other people.” Here are some guidelines I’ve been using to help me navigate when and how it’s OK to use AI to solve problems that involve others. Be transparent when you’ve used AI. This practice will reduce the cognitive load for employees so they don’t have to wonder who (or what) they’re interacting or collaborating with. I recently had a colleague use AI to create questions he would ask me on camera for a video project. He sent me the plan for the interview and added this note: “ChatGPT assisted with some question ideation and the layout.” For me, this was ideal because when I reviewed the questions and they didn’t sound exactly like him—and I found one to be off-topic—I didn’t have to do any guesswork about why. Being upfront and transparent about the use of AI alleviates some of the cognitive load for colleagues, and avoids the potential that they think less of each other if they deem the work to be lower standard than the norm. If you’re a senior leader, you should encourage people in your organization to be transparent as well. Set the expectation that disclosure is the norm, not the exception. One important caveat: some people might use AI as an accommodation. Alexandra Samuel raised this point: “If your colleague is dyslexic or really struggles to write emails, AI might enable them to handle communication in a quarter of the time, freeing them up to tackle projects that are important to both of you.” If that’s the case for you and you feel comfortable disclosing it, say so. Transparency still matters, but the context changes our understanding of why someone is using AI. Reserve AI for transactional relationships. If an interaction is purely transactional and you want to maintain, not build, your relationship with the other person, AI can be useful. For example, using AI to write a firm email to your general contractor who isn’t doing the work they promised on your house? That’s a reasonable use case. But interacting with a coworker you want to build a relationship with? That’s different. If you want to enhance trust and deepen connection, AI is unlikely to help you do that. Ask yourself: Is this interaction about accomplishing a task or about building a relationship? If it’s the latter, put the AI aside and engage with your colleague directly. Use AI to strengthen human relationships, not replace them. AI shouldn’t be seen as a replacement for interactions with your coworkers. Instead, prompt it to help you connect in real life. Ask it for ideas on how to build rapport with a new team member, conversation starters for a one-on-one meeting, or suggestions on how to increase your support squad at work. The goal, as Hadley describes, is to “use AI as a bridge, not a destination.” You can even program an AI tool to prompt you to engage with colleagues, such as asking it to scan your calendar for open spots and then nudging you to invite a coworker to join you for lunch or afternoon coffee at least once a week. Remind yourself that you have different norms for people versus AI. It’s worth taking a moment to articulate how you want to show up in your in-person conversations versus in your interactions with AI. You might even write it down: “I want to be courteous and collaborative with my colleagues. With AI, I can be transactional and efficient.” Making this distinction explicit helps ensure that your AI habits (good or bad) don’t inadvertently bleed into your human relationships. . . . Here’s what I keep coming back to: our relationships at work matter. They make us happier, more productive, and more engaged. They help us weather tough times and celebrate good ones. And they’re built through thousands of small interactions, many of them messy, awkward, and imperfect. When we hand those interactions over to AI or allow AI to intermediate them, we’re not just saving time or reducing friction. We’re outsourcing the very moments that create connection. We’re asking a machine to do the relationship work that only humans can do. So before you hit that “generate” button on your next email to a colleague, ask yourself: What am I really trying to accomplish here? And is AI helping me achieve it—or is it getting in the way?
