ChatGPT to write emails

Use ChatGPT to write emails It means getting you to help produce the first draft of a message —a response to a customer, a payment reminder, a business proposal, a follow-up — based on little information you provide to them. You give the context and the goal, ChatGPT returns an ordered text in the right tone; you reread, correct and send. In practice you go from "white sheet" to "draft to file", which is the point where most of the time you lose.

The advantage is not to write "perfect" emails automatically, but remove the initial friction and reduce the time for repetitive emails. The rule that matters is only one: ChatGPT writes the first draft, the person decides. The more precise you are in the prompt (to whom, why, with what tone, with what constraints), the less you will have to correct later. And when the same type of email comes back every day, the manual copy-paste stops making sense: there you go to the Automation.

In this practical guide you will find ready-to-copy prompts for the most common cases, a table that connects tone and lens to the right prompt, a five-step workflow and errors that make the text sound "fake". If you want to first understand what the tool is, start from ChatGPT: what it is and how to use it.

Why use ChatGPT for emails

Most of the time spent on an email is not typing: it is decide how to get started. Finding the right tone for an angry customer, softening a reminder without looking aggressive, saying no while being courteous. ChatGPT is good at just that: you tell it the point and it gives you a structure to work on, instead of the blank sheet.

The second reason is the repetitive volume. Those who manage assistance, sales or administration write dozens of similar emails every day: same logic, different details. ChatGPT generates the skeleton in seconds and you only customize the parts that really change.

Be careful not to confuse "fast" with "good". A rough draft pasted without rereading is almost always generic, too long and full of formulas. The value comes from the combination: ChatGPT for speed, you for judgment. Without the second part, send emails that look like they were written by a robot — and you notice.

How to write the right prompt for an email

A prompt is simply the instruction you give to ChatGPT. For emails, a prompt that works almost always contains the same four ingredients. If you skip one, the draft comes out unbalanced.

• Recipient and context: who receives the email and what happened before (a customer complaining, a lead asking for a quote, a supplier late).
• Objective: what the email must get. Reassure, sell, solicit, fix a call, decline. One, clear.
• Tone: friendly, formal, direct, empathetic. The wrong tone is the first reason an email sounds fake.
• Constraint: maximum length, language, what to include and what to include not say (e.g. do not promise dates, do not quote prices).

The more specific you are, the less corrections later. If you want to become good at building effective instructions even beyond emails, the guide on prompt engineering: how to get started goes into the details of the method. Another quick trick: paste to ChatGPT 2-3 emails you wrote and ask him to imitate that tone. He will learn your style much better than with a thousand adjectives.

What tone to use for each target

Not all emails want the same log. A response to a complaint is not written as a business proposal. This table links the objective to tone and to a start prompt that you can adapt.

The logic is always the same: clear objective plus tone fair equals useful draft. Change details in parentheses and have your own prompt. For wider uses of ChatGPT in work context, take a look at ChatGPT for companies.

Ready-to-copy prompts

Below are four complete prompts for the most frequent cases. They are meant to be copied, filled and used. Expand what you need.

Cross Note: These prompts remain generic models. ChatGPT does not know your company, so always check names, amounts, deadlines and promises before sending. To understand why it can "invent" plausible but fake details, read the AI Hallucinations.

Practical examples

Two concrete scenarios of Italian SMEs, to see how it really is used.

1. Studio accountant — recurring responses to customers. During the tax return, dozens of emails arrive with the same questions: missing documents, deadlines, amounts to be paid. The answering machine pastes the email received in ChatGPT with Prompt 1, gets a clear and courteous draft, corrects customer-specific data and sends. The time per email drops, the tone remains uniform even when different people respond. When a question is always the same, however, copying by hand stops making sense: one switches to an automation that prepares the answer for itself. It is the principle of the model described in AI automation: what it is and how it works.

2. E-commerce — prompts and after-sales service. An online store needs to handle payment reminders for orders by advance wire transfer and responses on shipments and returns. With the Prompt 2 the holder prepares the reminders keeping a kind tone (important: the customer may simply have forgotten), and with variants of the Prompt 1 responds to the most common doubts. The jump in quality comes when these messages, always the same in the structure, are generated and sent automatically, leaving the operator only the particular cases.

In both cases the logic is identical: ChatGPT by hand to start, automation when volume grows. They are no alternative, they are two steps of the same scale.

The workflow in 5 steps

Here is the sequence that makes emails written with ChatGPT fast and use it as a mental checklist.

1. Gather the context. Paste the received email or write the situation in two lines. Without context, the draft is generic.
2. State goal and tone. One sentence: what the email has to get and with what I register. It's the ingredient that people skip most often.
3. Add constraints. Length, language, what not to say (dates, prices, promises). Less space left to invention, less corrections.
4. Generates and refines. Read the draft and ask for a targeted correction: "shorter", "more formal", "remove the first sentence". One or two rounds are enough.
5. Review and send. Hand-check data, names, amounts, and tone. Customize a real detail. Then Never skip this step.

Step 5 is not negotiable: it is what separates a professional email from one that makes your company seem distracted. ChatGPT speeds up steps 1-4; the judgment on 5 remains yours.

Connected mail and authorized sending

Connecting a box does not allow every action possible. With Dots, the guide to Gmail, Drive and Slack app distinguishes tools and channels of conversation; the one at permissions, approvals and emails separates the preparation of a draft from the permission to send it. The text remains to be reviewed in the context of the client.

Mistakes to Avoid

• Pasting and send without rereading. It's error number one. It comes out a generic text, long and with possibly wrong data. The draft is a starting point, not the final message.
• Not specifying the tone. Without indication, ChatGPT uses a neutral and somewhat bureaucratic register. Always specify whether you want warm, direct or formal.
• Leaving emails too long. AI tends to fill. Set a word limit in the prompt and, if needed, ask "halve the length".
• Entering sensitive customer data. Names, IBANs, personal data in a public tool is a GDPR risk. Use placeholders and re-enter real data only in your box.
• Use it by hand for emails that are repeated every day. If you copy-paste the same type of message dozens of times, you are doing a job by hand to automate. There the real savings is elsewhere.

How to apply it in the company

For an SME or a professional, the sensible way to start is small and measurable, we do not "put AI on emails" in the abstract.

The practical path is this: for two weeks use ChatGPT by hand and mark which emails you write most often. The recurring ones are the perfect candidates to automate. At that point you no longer talk about "copying prompts", but about a system that reads the mail, prepares the answer and sends it with a human check on doubtful cases — the territory of the automation of business processes. If you want to go beyond the single email to more structured texts (newsletters, descriptions, articles), the starting point is ChatGPT to create content.

Related insights, with verified data on official documentation: Claude Haiku 4.5 vs GPT-5.6 Luna and Claude Sonnet 5 vs Haiku 4.5.

Conclusion

Use ChatGPT for emails It doesn't mean delegating communication to software: it means eliminating the friction of the first draft and writing repetitive emails faster, keeping the part that matters — tone, data, final decision for you. Everything is played on two things: a prompt that says who-why-tone-bonds, and a human review before pressing send. The moment you stop copying hand prompts and the same message comes back every day is the signal that it's time to Automate and to build something more solid, perhaps starting from ChatGPT for companies.

AI becomes useful when it enters processes. Giallo Studio helps SMEs and teams build concrete, measurable and sustainable automations — or take a look at the experiments of The Lab.