Polished yet impersonal: The unintended consequences of writing your emails with AI

People increasingly use AI to write their emails at work. AI can help us write them faster, clearer and more politely, while saving time and energy for more essential tasks. But what are the unintended consequences of AI-mediated communication? As more professionals at all levels, from CEOs to interns, rely on AI to handle their emails, what happens to the way we think, connect, and express ourselves? Are we outsourcing more than just words?

Writing emails is an essential part of work. On average, people spend about 28 percent of their working time on reading and answering emails from colleagues, clients, and collaborators. To deal with this load of emails, workers use strategies such as email batching (which involves checking and writing emails only at set times during the day) and technological fixes such as smart replies. With generative AI, workers can now choose between a range of tools to generate entire emails from scratch. 

Organisational members across all levels, from CEOs to young professionals, have described how AI saves them time and energy when writing emails so that they can focus on more essential tasks. They also frequently mention that it enables them to communicate more clearly and politely. Some of this is supported by scientific research. An experiment showed that using ChatGPT decreased the time that participants spend on writing tasks such as emails by 40 percent (Noy & Zhang, 2023). Exploratory interviews with early adopters of ChatGPT yielded that they frequently draw on the tool to embellish emails and other forms of text they have to write (Retkowsky et al., 2024). 

Despite the benefits of AI-mediated communication, there may be unintended consequences of using AI for emails. To start with the obvious: colleagues and clients can get easily annoyed by AI’s formulaic and stilted tone. This phenomenon is so widespread that experts recommend workers to add a personal touch to their AI-generated email. In addition, there is the paradoxical effect that by answering emails faster, we may end up with more emails to write and answer. But there are other more subtle and yet concerning consequences of using AI for emails: deskilling, dissonances, and detachment.

Deskilling

The debate on AI is often framed as one of upskilling. Upskilling involves acquiring new and advanced skills to advance one’s career or become more effective in the current position (Morandini et al., 2023).  In contrast, the possibility of deskilling (i.e. the loss of skills due to AI use) receives much less attention (Dries & Luyckx, 2025). While using AI for emails can make us more efficient, it does carry the risk of eroding our communication skills. First, relying on ChatGPT may decrease our ability to find the right words and tone ourselves in cases in which we cannot rely on AI. Either because the situation is so specific or sensitive that AI-generated replies are inappropriate or when communication takes place in person. Second and perhaps more importantly, writing is thinking. In some cases, we may have a thought in mind that we want to put on paper. In other cases, it is through writing that we become aware of our own thoughts and feelings – which is why journaling is such an effective method for self-reflection. Not every email that we write will involve a lot of thinking, but some do. When I write an email to the coauthors of a paper, for example, I might consider which parts of the paper still need work, reflect on ways to improve them, identify concrete tasks, reflect on the strengths of my coauthors, and suggest how to divide the work in a way that integrates all of the above. The more we rely on AI in writing our emails, the more we miss out on opportunities to think and reflect.  

Dissonances 

A second unintended consequence of using AI for your emails is that it can trigger dissonance both on the side of the sender and of the receiver. In a broader sense, dissonance refers to “a lack of harmony or agreement between things”. Two types of dissonance may occur in this context: cognitive dissonance and emotional dissonance. Let’s begin with an example that illustrates the experience of cognitive dissonance.

Cognitive dissonance refers to the internal discomfort one feels when encountering conflicting perceptions or information. For instance, receiving an email from a person you know whose  tone does not match their usual personal voice can be jarring. A vivid example of this was reported by a New York Times journalist who recently engaged in an experiment that involved responding to all of her work emails with AI for an entire week – with the result that some colleagues wondered whether she was annoyed at them. However, cognitive dissonance may also occur when interacting with strangers that use AI in their messages. I recently exchanged emails with an interview participant for a study to discuss his use of AI at work. The person had volunteered to participate and there was no financial incentive. In such cases, participants are usually motivated to share their viewpoints. At the beginning of the interview, however, the person seemed disinterested to answer my questions. Initially, I was irritated by the contrast between the participant’s enthusiastic emails before and lack of interest during the interview. But at some point, the person shared that he relies heavily on ChatGPT in his everyday work, which includes in his own words that he dumps bullet points into ChatGPT and asks the program to turn them into a friendly and easy to understand email. Through this comment, I realised that he was not necessarily disinterested, but there was simply a large discrepancy between his tone in his emails and in-person behaviour which I interpreted as being unmotivated. 

Emotional dissonance, by contrast, refers to a tension that workers feel when there’s a mismatch between the emotions they truly experience and those they are expected to display (Wharton, 2009).  Consider the following scenario: a student misses several classes in a row and then, without offering any valid explanation, asks me to elaborate on some of the concepts that were covered in their absence. In such cases, it may be beneficial – for both me and the student –if the tone of my email reflects some of the frustration I feel, rather than opting for the more convenient and bland email reply that AI generates. Why? Because for me as the lecturer, there is value in expressing my emotions and in communicating authentically– studies indicate that emotional dissonance is associated with higher likelihood of emotional exhaustion (Wharton, 2009). At the same time, the student in turn benefits from learning which behaviours are more or less acceptable in professional settings.

Detachment 

A third unintended consequence of using AI for emails is that it detaches (or disconnects) us from the receiver. Emails help us to connect and build relationships with colleagues, clients, and collaborators (e.g. Russell et al., 2022). It entails considering the receiver’s perspective. For instance, members of international teams might think about how their message will make the receiver feel and look for the right words to avoid offense (Parush & Zaidman, 2023). While our communication is already computer-mediated and occasionally supported through smart replies features (some of my last suggestions were: “Yay!” / “Talk to you then!” / “Woo hoo!” ), letting AI write our emails detaches us even further. Editor and writer Madeleine Holden argued in an essay that treating someone well involves considering their perspective in terms of what they want and need and that this can be a lot of work, because people are so hard to figure out. Therefore, it is ultimately not the final perfectly crafted polite and polished message that matters, but the consideration behind it. (Thus the phrase: “It’s the thought that counts”). When we let AI write our emails, we will be inclined to engage less in this process and in consequence risk to miss out on the opportunity to connect with the receiver and to engage in the perspective-taking that is involved in finding the right words to address someone and convey something.  

Each email presents a new opportunity

The public debate on AI and the future of work tends to focus on disruptive changes such as job losses (Dries et al., 2023). Although AI can be used to replace workers and threaten jobs, the majority of changes that we see in the workplace tend to be more nuanced and gradual and unfold their implications in the long run. AI may not be able to take over entire jobs, but it can perform certain tasks (von Richthofen et al., 2022). And while AI will not be able to take over our communication for us any time soon, relying on AI heavily in our work emails risks to erode our communicating skills, result in cognitive and emotional tensions, and decrease our connection to colleagues.

The goal of this blog article was not to question that AI can help you to write clearer and more effective emails and in a much more convenient way, but to shed light on such unintended, adverse consequences. Whenever we feel overwhelmed at work, we will feel tempted to turn to AI. And yet, each email that we receive presents us with a new opportunity to sharpen our ability to think and sharpen our communication skills, communicate authentically, and to make sense of another human’s perspective. When making such decisions, we should keep in mind that it is not only the final output – but the work that went into making it – that makes something valuable. 

The author is grateful to Charlotte Bernstorff and Mariam Humayun for feedback on earlier drafts of this blog post. 

References 

Dries, N., Luyckx, J., & Rogiers, P. (2023). Imagining the (Distant) Future of Work. Academy of Management Discoveries. 

Dries, N., Luyckx, J., & Bogaert, M. Media Review: Neo-Luddites, Unite! Worker resistance in an era of real dystopian threats. Organization Studies, 0(0), 01708406251321609. 

Morandini, S., Fraboni, F., De Angelis, M., Puzzo, G., Giusino, D., & Pietrantoni, L. (2023). The impact of artificial intelligence on workers’ skills: Upskilling and reskilling in organisations. Informing Science: The International Journal of an Emerging Transdiscipline, 26, 39-68. 

Parush, T., & Zaidman, N. (2023). Collective emotional labor and subgroup dynamics in global virtual teams: a dramaturgical perspective. International Studies of Management & Organization, 53(4), 239–262. 

Russell, E., Jackson, T. W., Fullman, M., & Chamakiotis, P. (2024). Getting on top of work-email: A systematic review of 25 years of research to understand effective work-email activity. Journal of Occupational and Organizational Psychology, 97(1), 74-103. https://doi.org/10.1111/joop.12462

Shakked Noy, Whitney Zhang, Experimental evidence on the productivity effects of generative artificial intelligence. Science 381, 187-192(2023).

von Richthofen, G., Ogolla, S., & Send, H. (2022). Adopting AI in the Context of Knowledge Work: Empirical Insights from German Organizations. Information, 13(4), 199. 

Wharton, A. S. (2009). The sociology of emotional labor. Annual Review of Sociology, 35, 147-165.