Research shows how AI can weaken human connection at work. Leaders should step in to prevent this.
Idea in Brief The Problem As AI becomes more integrated into work, employees increasingly turn to it not just for task support but for career advice, companionship, coaching, and counseling—treating it like a colleague or friend.
Yet despite these interactions, workplace loneliness persists. Why It Happens AI cannot replace the benefits of human connection at work. Without intervention, AI may foster socially isolated jobs, erode people’s motivation to work with and help others, and trigger unhealthful dependence on it. The Solution Monitor AI’s social impact, establish guidelines for its use, design it to promote human interaction, employ it for organizing social activities, and train employees in appropriate AI use. We’ve entered a new era of organizational life when, for the first time in history, people can turn to something other than a fellow human for conversation and support during the workday. They now can engage with AI. But how are employees using AI for social purposes? How is that usage affecting them? In our ongoing research we’ve been trying to answer those questions and understand where this new dynamic might take us in the future. The problem that sparked this study is loneliness—specifically, the loneliness that many employees around the world feel at work every day. It has significant business consequences, including lower job satisfaction, performance, and retention. We wondered if people might be assuaging it by talking to generative and agentic forms of AI at work. After all, we’ve seen how quickly people have become attached to their AI friends, romantic partners, and therapists. To find out, we studied knowledge workers who were relatively far along the AI adoption curve; nearly all of them used it weekly, daily, or even hourly in their jobs. Their survey responses and stories confirmed that most of them were already using AI for more than task execution. They were turning to it for personal support, including career advice and emotional validation—things that coworkers traditionally provide. Nonetheless, more than half of our study participants were lonely at work. In addition, our findings suggested that relying on AI for social support might erode company cultures and coworker cohesion in the future. To avoid fraying the social fabric of organizations, we encourage leaders to harness AI in five specific ways that will prioritize and promote connections to humans—not to the technology. How People Relate to Bots Our study set out to understand how employees interact with both human colleagues and AI and how those experiences influence their job satisfaction and well-being. We surveyed 1,545 U.S. knowledge workers who used AI regularly at work: Seven percent used it monthly; 36%, weekly; 48%, daily; and 9%, hourly. Most employed general-purpose forms of it such as large language models, and some also used specialized tools or platforms for tasks such as research, graphic design, and hiring. On average, our participants used AI more often in their jobs than the knowledge workers in Gallup’s research do. So our study offers a glimpse of a future in which more organizations’ workforces widely adopt AI. Among our research participants, we found strong evidence of AI anthropomorphism: They treated AI tools like humans. For example, 78% of participants used polite terms like “please” and “thank you” when interacting with them. When asked to pick the analogy that best captured how they viewed AI at work, 28% chose humanlike terms rather than technological ones . Using AI more often and using voice mode instead of text prompts intensified the tendency to personify AI. These findings are consistent with other research studies that have found that people were quick to attribute human qualities to even basic AI models. We anticipate this trend will continue as more advanced and agentic forms of AI infiltrate organizations. In our survey we adapted the Relationship Functions Inventory developed by the researchers Amy Colbert, Joyce Bono, and Radostina Purvanova. It measures how much human colleagues provide four forms of nontask support—career help, personal growth support, friendship, and emotional support—and we modified its questions to see if people were turning to AI for those same functions. They were. In fact, we were surprised by how common it was for employees to use AI for the kinds of social support once provided only by humans. Three-quarters of participants reported using AI for at least one form of it at work. Here’s how they employed AI in each area: Career development. Sixty-four percent of participants agreed with statements such as “AI helps me identify opportunities for development that will advance my career.” For example, an HR director told us she used AI to regain a sense of control over her career direction after her boss was unsupportive of her. Another person said, “AI showed me opportunities that I never saw before in my company. Got me a promotion.” Personal growth. Fifty-four percent agreed with statements like “AI helps me develop life skills and competencies, such as becoming a better listener, being more patient, and solving problems better.” One research administrator said that AI helped him develop his writing and analytic skills. A software coder described how she used AI to craft a “gentler and more professional” way of questioning a decision by her boss. A project manager said simply that AI “causes me to think through things and grow more.” Friendship. Fifty percent agreed with statements such as “I enjoy interacting with AI at work” and “AI is like a work friend.” An IT department head said, “It makes me feel as if I’m talking and working with someone instead of alone.” Another participant said, “Using AI makes it seem like I’m having a fun full conversation with a friend or close personnel. It lightens me up.” Emotional support. Thirty-five percent agreed with statements such as “AI helps me cope with stress” and “AI is empathetic to my needs and emotions.” A graphic designer recounted how she had a stressful situation at work and was unsure how to move forward. She vented the whole story to an AI chatbot and felt it understood the situation and gave her constructive ideas for managing it, including drafting a potential email response. In his project Time Stretched, Anton Repponen stretches the pixels of digital images to show individuals suspended in their own temporal bubbles. Although most people used AI for at least one type of support, our data shows that certain people are more likely to lean on AI. For example, younger people and men had a higher propensity to use AI for social support. Managers , employees on teams , and people in office or hybrid environments also tended to use AI more for social reasons. The rate of people’s overall AI usage and the stage of AI adoption in their organization mattered as well: In general, higher AI integration in the workplace was associated with more reliance on AI for social support. That implies we’ll see more widespread use of AI for relationship functions at work as adoption increases. Generally, the people in our study were quite happy with the quality of the social support they received from AI. A human resources employee said, “AI is my best friend in my work.” A manager said, “AI responds to me in a human way with a tone that I like and makes me feel heard and important.” Given that satisfaction, employees’ reliance on AI for career, personal, and emotional support is likely to continue—and intensify—as models become more sophisticated. In the short term it may offer meaningful relief to workers who would otherwise feel isolated or unsupported. Over time, however, the very qualities that make AI such a compelling social partner may incur hidden costs, with potentially serious consequences for employees and their organizations. Why We’re Still Lonely We grouped participants into three buckets based on our Work Loneliness Scale: high ; medium , and low . As you can see, 52% of them reported feeling highly or moderately lonely while working. Note that most of them weren’t working in isolated, remote jobs. Nearly all participants worked on teams. On average, participants reported spending 56% of their workweek conversing with coworkers synchronously . In addition, 83% worked from the office full-time or at least some days a week. In other words, many of our participants were lonely despite working collaboratively with their colleagues in person. What factors did matter to people’s loneliness? The same ones that mattered in our prior study on loneliness in the workplace. They included a dearth of organization-sponsored social activities, shyness, a low or isolated status in the organizational hierarchy, and negative perceptions of coworkers’ interest and care. Significantly, even though three-quarters of participants turned to AI for social support, only 12% said using it made them feel less lonely while working. The implication: Encouraging greater use of AI is unlikely, on its own, to reduce loneliness for most employees. The participants’ perspectives on the technology and its implementation differed according to their level of loneliness. People who were highly or moderately lonely rated their managers as less effective at implementing AI than people who were low on the loneliness spectrum did. Highly or moderately lonely people also felt that their senior leaders cared less about employees after AI had been introduced in their organizations and thought that AI was more likely to worsen their jobs or replace them. Simply put, the lonely employees were significantly more pessimistic about AI’s impact on their careers and distrustful of how their managers were handling integration. Although we didn’t ask them directly whether those feelings made them use AI less, these findings imply that loneliness may partly explain why AI adoption and sentiment tend to be uneven in organizations: Employees’ social experiences are most likely influencing their behaviors and attitudes toward technological change. Paying attention to worker loneliness could thus lead to more-successful AI initiatives. Another reason leaders should take loneliness seriously is that highly lonely people reported 27% lower job satisfaction and a 90% greater intention to quit their jobs than employees low in loneliness did; moderately lonely workers had 13% lower job satisfaction and a 46% greater intention to quit. In evaluating the factors that predicted satisfaction and turnover intentions, one pattern became clear: Employees’ relationships with their coworkers affected them more than anything we measured related to AI, including using AI for social support. It’s the connection to people—not connection to AI companions—that still matters the most for creating a satisfied and committed workforce. At this juncture in the evolution of work, it’s crucial to continue focusing on building stronger relationships among people in organizations. AI has not caused employee loneliness—the rates of loneliness at work have been high for quite some time. Yet the way that AI is being implemented may weaken personal connections and collaboration, ultimately making the problem worse. Warning Signs Ahead There are four reasons human relationships in organizations could suffer as AI adoption increases. First, AI can depopulate the workplace and create more isolation. A sales and business development employee explained the change: “I no longer collaborate with my team, make phone calls to subject matter experts, or rely on junior colleagues.” A product manager said, “I’m doing the same amount of work with an algorithm and not an actual person.” As organizations downsize and restructure their workforces to take advantage of AI’s capabilities, disaggregated and independent work may increase, resulting in more people who work alone and end up lonely. Second, AI can cause individuals’ social skills to atrophy and lower their motivation to connect with humans. Talking with an always reachable, sycophantic AI chatbot can be more appealing than conversing with real people. One of our participants said: “AI is free of judgment about your personality and never gives you the negative vibe you usually get from your colleagues.” Another participant predicted: “AI could magnify social isolation because people will prefer to use the AI tool instead of actually talking to people face-to-face.” We found evidence that this is already happening: An IT manager described an incident in which a colleague, to the dismay of teammates, unexpectedly sent an AI avatar to a meeting instead of attending personally. Workers who struggle with social anxiety or shyness might be especially prone to using AI to avoid human interactions. Third, by removing the need to go to colleagues for help, AI can undermine opportunities to build trust. In a survey conducted in May 2025 by the printing company MOO, 65% of workers said that they turn to AI tools before asking a colleague for assistance. A core antidote to loneliness is the feeling that other people “have your back” at work. That’s because the act of giving and receiving support from others creates interpersonal intimacy and mutual reliance. AI can disrupt these crucial exchanges. As one participant said, “It’s faster to get answers from an AI agent, so I no longer feel as much of a need to ask questions of coworkers.” Another said, “I feel it makes us more disconnected because we need each other’s help less.” As dependence on AI continues to grow, organizations may find that trust in peers and leaders deteriorates. Fourth, despite its lifelike capabilities, AI is indeed artificial and thus capable of triggering a sense of existential loneliness. One participant called human-AI interactions a “false friendship.” Another declared, “AI is like a helpful ghost in the office: always there and responsive but never truly present.” A third participant noted wistfully: “AI isn’t a person; it’s not human at all. AI is just a computer program in the end, so I am definitely lonely.” Technologists, including MIT professor Sherry Turkle, have warned of the threat to our humanity that accompanies an overreliance on artificial forms of intimacy. When AI agents become our managers, subordinates, and teammates, many of us will probably find the fake relationships unsettling on a deeper level. Despite these long-term risks, only 33% of the participants in our study had received any leadership guidance or information about how AI might affect their work relationships. It appears that organizations are now so focused on AI’s instrumental gains that they’re ignoring its potential interpersonal costs. That must change. We’ve argued previously that alleviating workplace loneliness requires sustained managerial attention to structures, culture, and incentives. That imperative now extends to the impact of AI on work relationships. How to Ensure AI Doesn’t Weaken Human Connections It is possible to both integrate AI into work and protect and nurture human connections. Inspired by our research and the emerging practices of some far-thinking organizations, we recommend five measures: 1. Monitor the social impact of AI adoption. Unfortunately, the understanding of AI’s effects on human dynamics lags behind technological advances and implementation. Few employers have systematically examined AI’s impact on relationships and well-being. Firms need to conduct regular surveys of team cohesion and employee loneliness levels as AI adoption increases. Our Work Loneliness Scale is a self-reporting tool they could use to assess loneliness levels throughout their organizations. User’s Guide .) Anton Repponen In addition to collecting quantitative data, companies can use interviews, message boards, and focus groups to gather employee stories and comments to understand what’s happening at a granular level. Early signs of trouble might include increased reports of isolation or distress, decreased informal communication, and fewer collaborative problem-solving sessions. With appropriate privacy guardrails in place, companies can automate the collection of all that information with AI—taking an approach like the one Microsoft researchers used when they employed machine learning to analyze anonymized emails, meetings, and Teams chats to unearth collaboration patterns. Executives need to make sure they develop a clear policy on data sources and usage, share only group-level trends and recommendations, and prohibit individual surveillance. 2. Establish guidelines for when and how AI can be used to replace human interactions. Employers need to determine when people should set aside AI in favor of human-to-human contact, something Salesforce calls its “human-in-the-loop mandate.” We recommend that coaching, mentoring, conflict resolution, and team building remain primarily human functions and be conducted in person to build relationships. When AI is used in these areas, it should augment, not replace, employee judgment. For example, AI might help identify problems in team dynamics, but members and managers should lead interventions to address interpersonal issues. In addition, each time major staffing or work-design changes occur, organizations should set clear parameters for when employees should turn to AI versus their colleagues. As lifelike AI agents and avatars come aboard, firms need to make sure they create official guidelines for when they can replace employees and how to notify colleagues that an avatar is responding instead of a human, as Harvard University has done. 3. Design AI to promote human interaction. To minimize the likelihood that employees will develop unhealthy personal attachments to AI tools, avoid overhumanizing AI—for instance, don’t assign AI agents names and personas to make them seem friendlier. Look for ways to introduce “positive friction” into employees’ interactions with AI so that, when appropriate, the path of least resistance leads back to people instead of defaulting to AI. One way to do this is through “AI provocations,” or AI-generated prompts that promote the retention of critical thinking skills by pushing users to think for themselves instead of giving them ready-made answers. We recommend designing socially oriented prompts aimed at strengthening collaboration and relationship skills. For example, you can configure AI assistants to refer a user to a human before providing a direct answer in nuanced situations in which teamwork might be advised. Or when developing plans or analyses, you can have AI insert the names of potential colleague reviewers and provide a checklist of key questions to ask them. The goal is to use AI to knit coworkers together in meaningful ways. 4. Use AI to organize relationship-building activities. In our study, lonely employees reported that their companies sponsored 38% fewer social activities than the companies of nonlonely employees did. The obvious solution is to get more employees involved socially, and AI can help. To start, the time AI saves on work tasks could be redeployed toward human-bonding activities, such as monthly team outings. AI can help plan and coordinate these events, taking the hassles of scheduling and logistics off managers’ plates. Groups could use AI to implement regular “connection rituals” like check-ins, walking meetings, or shared meals that prioritize relationship building. Companies can also use AI tools like Chronus to match mentees to mentors and coach them through their first few one-on-ones. In meetings AI could rotate facilitators and generate short icebreaker questions or exercises that spur reflection to get the conversation going and help remote and hybrid employees build rapport. The possibilities for harnessing AI’s design and execution skills for organizing social activities are endless. 5. Train employees how to apply AI in healthful ways. Currently, employees are hearing one message from employers: Use AI. What they’re not hearing is how to employ it in ways that improve their social and psychological well-being. Employers need to step up their education and communication efforts in that regard. That includes establishing programs to help employees recognize the warning signs of overreliance on AI for emotional support, understand the limitations of AI-based relationships, and develop strategies to maintain human connections while realizing AI’s benefits. Companies such as WellSteps offer “digital wellness” programs that do this. In addition, leaders should model balanced AI use themselves, demonstrating when to leverage it for efficiency and when to prioritize human interaction. That includes being transparent about their own AI patterns, sharing both successes and limitations, and consistently reinforcing the value of human relationships throughout the organization. The Choice Is Ours Used thoughtfully, AI can give employees more time to connect and help them find ways to do so. But as AI becomes an ever-present companion in the workday, leaders must remain clear-eyed about what’s gained—and what may be lost—for the mental health of the organization. Left unchecked, AI can deepen work isolation, dull social motivation and skills, and quietly displace the small acts of help and empathy and the shared experiences through which coworkers build trust and belonging. It’s up to leaders to safeguard their organizations against those dangers. About the Research We studied 1,545 full-time U.S. knowledge or office workers who used AI for work at least monthly. They were 25 to 54 years old and on average were in their early forties; 50% were women and 50% were men. They worked in various modes: 46% were fully in-person, 37% hybrid, and 17% fully remote. Seventy-three percent managed other employees. Seven percent used AI monthly; 36%, weekly; 48%, daily; and 9%, hourly. The most popular AI tools were large language models such as ChatGPT, Microsoft Copilot, and Claude ; research and knowledge tools such as Perplexity and NotebookLM ; and organizational or note-taking tools such as Notion and Fireflies.ai . The market research firm Dynata collected data in June 2025 and December 2025 using a 10-minute survey on work loneliness, AI views and usage, demographics, personality, work mode, job features, and organizational design factors. AI-provided social-support functions were measured by adapting a shortened version of the Relationship Functions Inventory to ask how AI was providing career advice, personal growth support, friendship, and emotional support. We conducted a follow-up survey in August–September 2025 with a subset of the June participants who agreed to be recontacted. As part of that survey, we asked participants to record stories on the Voiceform platform and elaborate on their AI experiences at work. Only the December participants were asked if their organization’s leaders had communicated with them about the impact of AI on their relationships at work. The usage rates of different AI types were collected from the follow-up survey and December participants.
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Personal Perspective: The recent verdict invalidates personal responsibility.Personal Perspective: Holding companies liable for patterns of engagement and behavior outcomes is a slippery slope.
Read more »
Trump signs executive action to pay TSA employees after Congress fails to agree on DHS fundingThe Senate voted early Friday morning to fund much of the Department of Homeland Security, taking a step to reopen large parts of an agency that has been partially shut down for more than a month.
Read more »
TSA employees to get paid Monday following Pres. Trump directive as shutdown continuesPassengers traveling out of Oakland San Francisco Bay airport this weekend saw small lines and a few delays. But many still took precautions to come early. 'Morale is really, really, low at Oakland airport right now, nationwide as well,' said a TSA officer.
Read more »
Navigating Career Stagnation: Helping High-Performing Employees Thrive Without a PromotionThis article explores strategies for managers to support high-performing employees who face career stagnation due to limited promotion opportunities. It offers advice on managing expectations, fostering growth through expanded responsibilities and skills development, and maintaining employee engagement in a challenging job market.
Read more »
Personal Perspective: I'm not building a brand, I’m raising a human.Personal Perspective: Moms manage narratives, guard reputations, and perform like publicists. But I’m not building a brand. I’m raising a human.
Read more »
‘The Pitt’ Shows Burnout Is A System Failure, Not A Personal OneBurnout in health care isn’t personal—it’s systemic. What a recent episode of 'The Pitt' shows about ER stress also applies to the strained public health workforce.
Read more »