Best ChatGPT Prompts for Work

Most prompt advice is written by people who do not do much actual work in ChatGPT. It is full of “act as an expert with 20 years of experience” templates that barely outperform a direct question. This is different. These are the prompts that produce output you can use, organized by the work tasks where ChatGPT has the clearest advantage.

Before the categories: the anatomy of a prompt that works.

The anatomy of a good prompt

A well-formed prompt has five parts. You do not need all five every time, but the more you include, the better the output:

• Role: who ChatGPT should be for this task (“You are a senior communications manager”)
• Task: what you want it to do (“rewrite this email”)
• Context: the background it needs (“the recipient is a VP who does not like long emails”)
• Format: how the output should look (“3 bullet points” or “under 100 words”)
• Constraint: what to avoid or emphasize (“do not mention pricing” or “keep my original meaning”)

A prompt that uses three of five consistently outperforms one that uses none of them. Keep this framework in mind as you adapt the examples below to your own work.

Email and communication prompts

Email is where most people get the most immediate value from ChatGPT. The prompts below cover the three situations that come up most.

Rewriting a bloated email

Rewrite this email to be more direct and half as long. Cut filler, keep every important point. Do not change the tone: [paste email]

This works because it gives ChatGPT a clear criterion (half as long) and a constraint (keep the tone). Without the constraint, it often makes the email colder than you intended.

Following up without sounding passive-aggressive

Draft a follow-up email to [name] who has not responded to my message about [topic]. I need a response by [date]. Polite but firm. Under 80 words.

The word limit forces it to skip the filler. Without it, follow-up emails come back as three paragraphs of softening language.

Summarizing a long thread

Summarise this email thread in 3 bullet points. Identify who needs to do what and by when. Flag anything unresolved: [paste thread]

Asking it to flag unresolved items specifically catches the things that usually fall through the cracks in a long back-and-forth.

Writing and editing prompts

Editing for clarity

Edit this paragraph for clarity. Remove filler phrases, passive voice, and anything that repeats what was already said. Keep my exact meaning: [paste]

The “keep my exact meaning” constraint is important. Without it, ChatGPT occasionally changes what you actually said while making it sound smoother.

Drafting a longer document from scratch

I need to write a [blog post / report / proposal] about [topic] for [audience]. Give me a clear structure with headings first. Once I approve it, write each section one at a time.

The two-step approach (structure first, then writing) produces dramatically better output than asking for the whole thing at once. You catch structural problems before they are buried in paragraphs.

Converting notes into prose

Convert these rough notes into a coherent paragraph. Preserve the meaning and order of the points. Professional tone: [paste notes]

Useful for turning meeting notes, voice memo transcripts, or rough bullet points into something you can send or publish.

Research and analysis prompts

Stress-testing a decision

List the 5 strongest arguments for and against [decision]. Be specific, not general. Do not hedge.

The “do not hedge” instruction matters. Without it, you get balanced diplomatic output that does not actually help you think. You want the strongest version of each side.

Preparing for a vendor evaluation

I am evaluating [software / service / vendor] for [use case]. What are the 10 most important questions I should ask them? Focus on what vendors tend to understate or obscure.

The second sentence shifts the output from generic due-diligence questions to the things that actually catch problems.

Understanding an unfamiliar concept

Explain [concept] to someone who understands [related field] well but is completely new to this one. Use one concrete example.

Anchoring the explanation to something you already know speeds up comprehension significantly. ChatGPT can meet you exactly where you are rather than starting from zero.

Code and data prompts

Debugging a formula or query

Here is a [Python function / SQL query / Excel formula] that is not working as expected. Tell me what is wrong, why, and how to fix it. Then show me the corrected version: [paste]

Asking for “what is wrong and why” before the fix teaches you something. Asking only for the fix gives you code you cannot maintain.

Writing a formula you do not know how to write

Write an [Excel / Google Sheets] formula that does [task]. Explain what each part does in plain language.

The explanation request is not optional. It is the only way to know whether the formula is doing what you think it is doing.

Converting data to a readable format

Convert this JSON into a plain-language summary of the most important values. Then show it as a readable table: [paste]

Useful when you receive an API response or data export and need to quickly understand it without parsing it yourself.

Meeting and document prompts

Turning rough notes into a structured recap

Here are my rough notes from a meeting. Format them as: attendees, decisions made, action items with owners, open questions. Do not add anything that is not in the notes: [paste notes]

The constraint “do not add anything not in the notes” prevents hallucinated action items, which is a real problem when you paste sparse notes and ChatGPT tries to fill gaps.

Writing a meeting agenda

Write an agenda for a [topic] meeting with [n] participants. Duration: [time]. Goal: [outcome]. Include time allocations and a slot for questions.

Asking for time allocations forces the agenda to be realistic. A 45-minute meeting with six agenda items is not an agenda, it is wishful thinking.

The one upgrade that works on every prompt

At the end of any prompt, add:

Review your output and identify one specific thing you would improve if you were rewriting it.

ChatGPT often catches its own weaknesses when asked to. It will flag where the output is vague, where a section is weaker than the rest, or where a constraint was not fully met. You can then ask it to fix that specific thing, or fix it yourself. Either way, the output after this step is consistently better than the output before it.

What these prompts have in common

Every prompt in this guide does at least three of the five things in the framework: role, task, context, format, constraint. None of them are magic phrases. They work because they give ChatGPT enough information to do something specific instead of something average.

The most common mistake at work is treating ChatGPT like a search engine: a short query expecting a useful answer. The model responds to context. The more you give it, the less it has to invent, and the less it invents, the more useful the output is. That is the whole strategy.

Start with the email prompts. They produce results in under two minutes and make the habit obvious before you move to anything more complex.