Email Tone Transform: why telling AI to "make this more professional" kills your response rate — and the prompts that fix it

Two modes, two prompts, two very different outcomes. Both run in a single request to the LLM:

MODE 1 — PLAYFUL (boosts reply odds):
Original email → [prompt specifying a warm style] → email rewritten in a warm tone

MODE 2 — PROFESSIONAL (lowers reply odds):
Original email → [formal-register prompt] → email rewritten in a cold tone

Both steps happen in a single chat request. No automation needed.

Task: You need to write an email to Alex Chen, a product lead at a company you crossed paths with at a conference. You want to propose a call to discuss a partnership. The email is written, but it feels stiff and bureaucratic.

Prompt (playful mode):

Rewrite the email below in a lively, warm, slightly playful style — like 
someone who's genuinely excited to connect, not a corporate autoresponder. 

Preserve the exact meaning, all facts, and the call request. 
Don't add or remove any information. 
Change only the tone, wording, and feel.
Style: lively, warm, with a touch of wit — the way sharp people write 
in a chat when they want to come across well.

Email:
[paste your email here]

Result:

The model returns a rewritten email with the same content — same dates, names, call request — but a completely different feel. The text becomes livelier: cliches get swapped for natural phrasing, sentences get shorter and less bureaucratic. The email reads like it was written by someone who actually cares, not copy-pasted from a corporate style guide.

When told to "sound more professional," an LLM optimizes for form, not perception. It strips out conversational phrasing, adds formal constructions, and aligns the style with corporate norms. All of that lowers emotional warmth — and the recipient feels the difference, even without consciously noticing it.

People don't reply to emails because they're "correctly written" — they reply because they sense a genuine human connection. Social psychology research shows that a warm, playful tone reads as a signal of closeness and openness, which triggers the impulse to respond. This holds true both with colleagues and with strangers — the only difference is when tone kicks in (at the open vs. at the reply stage).

The key lever isn't whether you use AI at all — it's which direction the AI shifts the tone. Recipients had no idea whether an email had been AI-edited. And it didn't matter — their behavior was driven entirely by the emotional warmth of the text. AI here is just a tool for shifting tone, not a magic fix and not a risk in itself.

Prompt levers you can pull: - Style → instead of a vague "playful," get specific: "the way people write in a work Slack channel," "like a smart friend who's busy but happy to hear from you," "like a YC founder" — sharper role-play produces sharper results - Constraints → "no emojis," "no over-familiarity," "no exclamation points" — fine-tune for context - Context anchor → add "this is for a client I'm meeting for the first time" or "for a colleague I've worked with for a year" — the model will calibrate the warmth accordingly

Playful / warm mode:

Rewrite the email below in {style_description} style. 

Preserve the exact meaning, all facts, names, dates, and {key_action}. 
Don't add or remove any information. 
Change only the tone, wording, and feel.
The result should be lively and appropriate for a workplace context.

Style: {style_description}

Email:
{email_text}

What to fill in: - {style_description} → "lively and warm, like sharp people write in chat," "friendly with a touch of wit," "energetic and direct" - {key_action} → "the meeting request," "the timeline question," "the partnership proposal" - {email_text} → paste your full email

Professional mode (use deliberately — when formality matters more than response rate):

Rewrite the email below in a formal, professional tone, 
preserving the original meaning, factual content, and {key_action}. 
Don't add or remove any information, and don't change names, dates, or tasks. 
The text should be clear, concise, and appropriate for official business correspondence.

Email:
{email_text}

🚀 Quick start — paste into your chat:

Here's a template for rewriting an email in a warm tone. 
Adapt it to my situation: [describe who you're writing to and what response you want]. 
Ask me questions to fill in the blanks.

[paste the template above]

The LLM will ask about style, recipient context, and the key action in the email — because those determine how warm to make the tone without losing appropriateness.

⚠️ Email type: For official and legal correspondence (contracts, complaints, government agencies), a professional tone isn't a mistake — it's the right call. The warmth insight applies where you need a genuine response, not where format compliance is the goal.

⚠️ Habituation effect: The experiment ran for three weeks. It's unclear whether the effect holds up if a recipient keeps getting "playful" emails from the same person over and over — the novelty factor may wear off.

⚠️ Degree of playfulness: The study used styles like "pirate" and "medieval." For most professional English-language correspondence, those specific formats would come across as too much — aim for a warm, lively tone rather than literal playfulness.

⚠️ External recipients: For strangers, warmth has a bigger effect on replies but a smaller effect on opens. For colleagues, it's the opposite — a warm email grabs attention right at the open stage, because it breaks the usual communication pattern.

The idea was simple: take real people, real work emails, and test whether AI editing changes recipient behavior. 121 employees across six companies (Israel, UK, Sweden, Ukraine) wrote regular work emails over three weeks. Each week used a different condition: no AI, AI in playful tone, AI in professional tone. The order was rotated so that no single condition got an advantage just from going first.

Three things were measured: whether the recipient opened the email, whether they replied, and how fast. Additionally, each email's "warmth" was automatically scored on a scale from −1 to +1. The dataset totaled 16,880 emails — large enough for small effects to become statistically visible.

The most interesting part happened in the analysis: when researchers looked at conditions directly (unaided vs. playful vs. professional), there was no difference in opens or replies. That could have looked like the hypothesis had failed. But when they looked through text warmth — everything clicked into place. The playful condition raised warmth, and warmth predicted replies with an odds ratio of 3.32. The professional condition lowered warmth — and through that same channel, lowered response rates. The effect was real; it was just hidden behind a single mediating variable.

Playful condition:

Transform the following email into a [STYLE] version of the same message. 
Preserve the exact meaning, all factual details, and all requested actions. 
Do not introduce new information or omit existing information. 
Only change the tone, wording, and style. 
Keep the result readable and suitable for a workplace context. 
Style: [STYLE]. 
Email: [EMAIL CONTENT].

Professional condition:

Rewrite the following email in a formal and professional workplace tone 
while preserving its original meaning, factual content, and requested actions. 
Do not add or remove information. 
Do not change names, dates, deadlines, or action items. 
Keep the message clear, concise, and appropriate for professional communication. 
Email: [EMAIL CONTENT].

Context: These are the exact prompts used in the experiment, word for word. Thousands of real work emails passed through them. In the [STYLE] slot, researchers plugged in a specific style for each week (pirate, medieval, etc.).

💡 Adaptation for everyday professional correspondence: Instead of a literally "playful" style, try "lively work tone" — the way people write in a company Slack channel rather than in a formal email. This keeps the warmth without risking coming across as unserious.

Rewrite this email the way a sharp colleague would write it in a work chat — 
lively, human, no corporate jargon or stock phrases. 
Don't touch the meaning, the facts, or the request. 
Only the tone and wording.

Email: {text}

🔧 Technique: warmth check before sending → spot where the tone falls flat

Before sending an important email, ask the model to rate its temperature:

Rate the tone of this email on a scale from "ice-cold corporate" to "warm, genuine human." 
Point out specific phrases that make it colder. 
Suggest a replacement for each one.

Email: {text}

This beats jumping straight to a rewrite — you can see exactly where the tone falls flat and fix it surgically.

🔧 Extrapolating to other channels: The "warmth → response" principle isn't limited to email. The same prompts apply to: - LinkedIn/Slack messages to potential partners - Follow-ups after a meeting - First-touch messages in cold outreach

The mechanics stay the same: preserve the meaning, shift the tone toward lively and warm.

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

Authors: Ziv Ben-Zion (School of Public Health, University of Haifa; Department of Psychiatry, Yale School of Medicine) and Teddy Lazebnik (Department of Information Systems, University of Haifa; Department of Computing, Jönköping University)

Contact: [email protected]

Related concepts from the study: AI-mediated communication framework (Hancock et al.); the broaden-and-build theory of positive affect (Fredrickson); the expectancy violation effect in communication (Burgoon)