Lindsey Pollak noticed recently that many of her Zoom meetings included an unannounced participant: an AI assistant or notetaker. “Every time I logged in, I was being greeted by Fathom or Otter or Granola,” Pollak, a career and workplace expert based in New York City, tells CNBC Make It. <\/p>\n
Adding an AI bot to a meeting without asking first makes the interaction feel “robotic and weird,” according to Pollak: “It feels like I’m under surveillance or in an interrogation interview with the police, rather than on a business call.”<\/p>\n
Integrating new technology like artificial intelligence can cause “communication difficulties” and interpersonal friction at work, according to Tessa West, a social psychologist and professor at New York University. Most workplaces haven’t yet established etiquette norms around when it’s appropriate to use AI, West says. For some workers, letting AI write a message to a coworker may seem like a harmless time-saving hack, while others view it as inconsiderate or rude. <\/p>\n
Thirty-seven percent of employees admit to using AI tools<\/a> for tasks that require emotional intelligence, according to a survey<\/a> from GoTo and Workplace Intelligence published in May. Nearly as many (35%) said they had used AI to draft or edit sensitive workplace communication<\/a>, a September 2025 survey from email platform ZeroBounce found. At the same time, people who use AI to outsource “socio-relational” tasks are perceived more negatively<\/a> by their peers, according to researchers from the University of Kent. <\/p>\n As technology evolves, workplace etiquette rules “have to evolve with it,” says Jacqueline Whitmore, the founder and CEO of the Protocol School of Palm Beach, a business etiquette training company. In Whitmore’s experience, growing pains are almost inevitable: It took years to develop clear-cut professional norms around email, for example.<\/p>\n Here’s what experts say about how to use AI in the workplace without damaging your relationships.<\/p>\n Ask permission before adding AI <\/p>\n Using AI to record or transcribe meetings has become “super common,” West says, but that doesn’t mean you shouldn’t ask first. “It feels intrusive if you’re hitting the record button without asking permission,” she says.<\/p>\n Meetings should start with “a human moment” before anyone adds an AI agent to the call, according to Pollack: “The best practice is to have a greeting, even if it’s very brief, and then say, ‘Is everyone okay if I turn on the notetaker so that I can be really present in this call?'” <\/p>\n It can be tempting to use AI to transcribe every call, but she encourages workers to be choosy. “It can hamper connection and good work,” Pollack says, adding she feels “much less comfortable” chiming in during brainstorm sessions when she’s aware that they’re being recorded.<\/p>\n It can also be unnerving to see casual chitchat included in meeting transcriptions, Pollak says. “Sometimes it included really weird personal updates about my vacation and other people’s families and pets and the weather,” she says. “It just sort of felt absurd to me.” <\/p>\n Don’t outsource ‘human’ tasks<\/p>\n It may not feel like a big deal to generate emails or messages to co-workers with AI, West says, but research shows<\/a> that many people view using AI for interpersonal communication as a faux pas. Managers who use medium to high levels of AI assistance in writing emails, for instance, are seen as less authentic and trustworthy<\/a>, according to a study published in July 2025.<\/p>\n “If people find out you’re doing that \u2014 or even if they suspect it and you’re not \u2014 they find that to be very artificial and rude,” West says. “Nothing good comes of having someone suspect that you are outsourcing any human activity with AI.”<\/p>\n According to West, the worst kinds of messages to outsource to AI<\/a> are “the ones that are meant to convey emotion: anything with sincerity, anything with an apology involved in it, any kind of mea culpa.”<\/p>\n Whitmore isn’t against using AI to generate messages and emails, but she advises workers to edit the results and add a personal touch so that it isn’t “evident that it was written by AI,” she says. “You don’t want it to sound like a canned response. You don’t want to lose your voice.” <\/p>\n And if you’re worried that an AI-generated message sounds too unnatural, Whitmore says, “ask somebody else to read it and give you their opinion.” <\/p>\n Avoid citing AI ‘advice’<\/p>\n As more people turn to AI chatbots for guidance on everything from relationships<\/a> and mental health<\/a> to money<\/a> and careers<\/a>, West cautions against relying on AI for workplace advice. <\/p>\n For example, it’s likely to rub your manager the wrong way if you open a salary negotiation meeting by saying “I asked ChatGPT what I should ask for in a raise,” according to West, or if you counter their points with, “But Claude said…” <\/p>\n From a higher-up’s perspective, “it feels like you don’t value” their experience or guidance, West says. “Why am I helping you? Why am I trying to train you if you’re going to replace me” with AI?<\/p>\n When you’re preparing for an important conversation, it’s almost always best to consult another human, Peter Stewart, a business psychologist and managing partner at\u00a0coaching firm Stewart\u00a0Leadership, told CNBC Make It<\/a> in March. A trusted colleague is better equipped than an AI bot to give you informed advice, he said: “You’ve got a history with this person, and you know they’re coming with your best interest at heart.” <\/p>\n Moreover, while AI is programmed to please<\/a> its user, a close colleague is more likely to be honest with you if they feel you’re on the wrong track. “They can lay it straight,” Stewart said. “They even ask questions like, ‘What do you mean by that?'” And they can gently flag if they think you’re misinterpreting the situation or doing something that might negatively impact your work relationships or career opportunities.<\/p>\n