Playful AI in Professional Email: A Field Experiment on Tone and Recipient Engagement

As generative AI becomes a standard tool for drafting workplace emails, organizations often assume that using AI to improve the polish or professionalism of a message will naturally increase recipient engagement. This study investigates whether AI-assisted writing actually changes recipient behavior (open rates, reply rates, and response times) and, if so, whether this effect is driven by the fact that a message is AI-generated or by the specific emotional tone the AI produces.

The researchers conducted a randomized crossover field experiment with 121 employees across six companies. Over three weeks, participants sent their regular work emails under three conditions: unaided writing, GPT-5 rewriting in a playful tone, and GPT-5 rewriting in a professional tone. By using a crossover design, each participant served as their own control, allowing the researchers to isolate the impact of tone on 16,880 total emails while controlling for individual writing styles and recipient networks.

The study found that AI-assisted editing had no direct effect on open or reply rates. However, the interventions significantly altered the emotional tone of the emails: playful editing increased positivity, while professional editing decreased it. Further analysis revealed that this shift in tone was the primary driver of engagement. Within-sender increases in positivity were strongly associated with higher odds of an email being opened and answered. Essentially, AI-assisted communication shapes workplace engagement indirectly by altering the emotional character of the language, rather than through the mere act of using AI.

These findings suggest that the "professional" polish often sought in workplace communication may inadvertently suppress engagement by reducing the warmth of a message. For employees and organizations, this implies that the value of an AI writing assistant depends heavily on the specific tone it produces. Rather than focusing on the tool's ability to improve grammar or structure, users should consider how the resulting emotional register affects the recipient's likelihood to engage, balancing the need for professional credibility with the interpersonal benefits of warmth and positivity.