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

This research investigates how using Large Language Models (LLMs) to rewrite professional emails influences the behavior of the people receiving them. While AI tools are becoming common in the workplace, it remains unclear whether using them actually changes how recipients engage with our messages. This study explores whether the specific tone chosen by an AI—playful versus professional—acts as a catalyst for better communication outcomes.

How the Study Was Conducted

Researchers conducted a randomized crossover field experiment involving 121 employees across six different companies. Over a three-week period, participants sent a total of 16,880 work emails. Each email was categorized into one of three conditions: unaided writing (the employee’s own words), GPT-5 rewriting in a playful tone, or GPT-5 rewriting in a professional tone. By comparing these conditions, the authors sought to determine if AI-assisted editing directly impacted key engagement metrics like open rates, reply rates, and response times.

The Impact of Tone on Positivity

The study found that the AI’s tone had a measurable effect on the emotional content of the emails. Playful editing significantly increased the emotional positivity of the messages, while professional editing had the opposite effect, decreasing it. Despite these shifts in tone, the AI’s intervention did not lead to a direct change in how recipients interacted with the emails; the raw open and reply rates remained unaffected by the mere fact that an AI had rewritten the text.

The Indirect Path to Engagement

While the AI’s involvement did not directly change recipient behavior, the researchers discovered an important indirect link. The study revealed that the level of emotional positivity within an email—regardless of whether it was generated by AI or written by the sender—was a strong predictor of engagement. Higher positivity was associated with a significantly higher likelihood of an email being opened and replied to.

Key Takeaways

The findings suggest that AI-assisted communication does not necessarily improve workplace engagement simply because it is used. Instead, the effectiveness of AI in professional settings depends on the emotional tone it produces. The study concludes that the primary value of AI writing tools in this context lies in their ability to shape the emotional quality of the message, which in turn influences how recipients choose to respond.