ChatGPT Prompts for Work: A Reusable Prompt Library for Everyday Tasks

Overview

A useful work prompt usually contains five parts: the role or perspective the assistant should use, the task to complete, the context it needs, the constraints to follow, and the format for the final answer. You do not need to include every part in every request, but adding the missing details is often the fastest way to improve an inconsistent result.

A simple reusable structure is:

Act as [role or perspective].
Task: [what you want completed].
Context: [relevant facts, audience, source material, or assumptions].
Constraints: [tone, length, exclusions, deadline, or decision criteria].
Output: [format, headings, table, checklist, or numbered steps].
Before answering, identify any important ambiguity or missing information.

Replace the bracketed fields rather than treating a prompt as permanent text. This turns a one-off request into an AI workflow template that can be reused by you or a team. For broader ideas, see this collection of AI workflow templates for common knowledge-work tasks.

Step-by-step workflow

1. Draft and improve an email

Use AI to create a first draft, adjust tone, or make a long message easier to scan. Keep the facts and intended outcome explicit.

Rewrite the message below for [recipient and relationship].
Goal: [request, decision, update, or next step].
Tone: [direct, warm, concise, formal, or collaborative].
Keep these facts unchanged: [facts].
Do not invent commitments, dates, or explanations.
Return:
1. A subject line
2. The revised email
3. A one-sentence summary of the requested action

Message:
[paste message]

Ask for alternatives only when they serve a purpose. For example, request a direct version and a more diplomatic version if the relationship or sensitivity of the message warrants it.

2. Summarize research without losing its limits

A text summarizer can reduce reading time, but the prompt should distinguish the source’s claims from interpretation. This is especially important for technical notes, policies, and internal documentation.

Summarize the material below for [audience].
Use only the supplied text. Do not add outside facts.
Return:
- A five-bullet executive summary
- Key terms and their definitions
- Claims that require verification
- Decisions, recommendations, or actions stated in the source
- Questions the source does not answer
Preserve uncertainty and attribute claims to the source where appropriate.

Source:
[paste text]

For a large document, work section by section and then ask the assistant to combine the summaries. Keep the original source available for checking important details.

3. Brainstorm options with useful boundaries

Unconstrained brainstorming often produces familiar or impractical ideas. Add criteria that help you compare the options.

Generate [number] options for [problem or objective].
Context: [team, users, current process, or available resources].
Constraints: [budget, time, tools, technical limits, or compliance needs].
Prioritize ideas that are [specific criteria].
For each option, provide:
- Short description
- Expected benefit
- Main risk or trade-off
- First practical step
- Information needed before choosing it
Group similar ideas and label assumptions.

Follow with a decision prompt rather than accepting the list at face value:

Compare these options against the following criteria: [criteria].
Score each criterion from 1 to 5, explain the score briefly, and show where the scoring depends on an assumption. Do not present the result as an objective decision; identify what a human decision-maker still needs to confirm.

4. Turn a goal into a project plan

AI can help expose missing steps, dependencies, and risks, but the plan should be grounded in your actual tools and ownership model.

Create a practical plan for [project or outcome].
Target result: [definition of done].
Available people and tools: [details].
Deadline or milestones: [dates or sequence].
Known dependencies: [dependencies].
Return a table with:
- Workstream
- Task
- Owner role
- Dependency
- Deliverable
- Suggested timing
- Risk and mitigation
Flag tasks that need clarification instead of assigning unsupported owners or dates.

Ask for a shorter version after reviewing the detailed plan. The first output is for thinking; the second can become a working checklist or status update.

5. Analyze structured information

For logs, feedback, tickets, or survey responses, define the categories before asking for conclusions. This makes the output easier to audit.

Analyze the records below using these categories: [categories].
For each record, return:
- Category
- Brief rationale
- Urgency: low, medium, or high
- Suggested next action
If a record fits multiple categories, list the primary category and explain the overlap. If the evidence is insufficient, mark it as unclear rather than guessing.
Return a summary table followed by recurring patterns and notable exceptions.

Records:
[paste data]

6. Rewrite for clarity

Rewriting works best when you protect the meaning before improving the style.

Edit the text below for clarity and flow.
Audience: [audience].
Purpose: [purpose].
Preserve the original meaning, figures, qualifiers, and technical terms.
Remove repetition and vague phrasing. Do not add claims or examples.
Return:
1. Revised text
2. A short list of substantive changes
3. Any sentence where the intended meaning is ambiguous

Text:
[paste text]

Tools and handoffs

Prompts are most effective when they fit a clear sequence rather than living in a chat window by themselves. A practical workflow might use voice-to-text to capture rough notes, an AI assistant to organize them, a document editor for human revision, and a task system for confirmed actions. Treat each handoff as a separate stage with a defined input and output.

For example:

1. Capture: Record meeting notes or ideas in plain language.
2. Structure: Ask the assistant to separate decisions, open questions, and actions.
3. Verify: Compare the structured result with the original notes.
4. Publish: Move only confirmed information into the team document or task system.

Choose tools by task, not by the number of features. A browser-based utility may be enough for a quick transformation, while a connected assistant may be more suitable when the workflow depends on approved internal context. This comparison of AI productivity tools by task can help you assess the handoff rather than selecting a tool in isolation.

When prompts involve company information, review the tool’s settings and your organization’s rules before pasting sensitive material. If the content is confidential, use an approved environment or replace identifying details with placeholders.

Quality checks

Prompt engineering is not only about getting a polished response. It is also about making errors visible. Before using an AI-generated result, check:

• Factual grounding: Are claims supported by the supplied material or independently verified?
• Completeness: Did the assistant address every requested part of the task?
• Invented details: Did it create names, dates, sources, commitments, or technical steps?
• Meaning: Did rewriting change a qualification, requirement, or conclusion?
• Format: Is the result structured for the next person or tool in the workflow?
• Audience: Would the intended reader understand the terminology and recommended action?

If the output is weak, revise one variable at a time. Add missing context, narrow the audience, specify the format, or provide a good example. Avoid making the prompt longer merely for the sake of length. A short prompt with relevant evidence usually beats a detailed prompt built from assumptions.

When to revisit

Review your prompt library when the underlying work changes, not on a fixed schedule alone. Revisit a prompt after changing tools, document formats, approval steps, team responsibilities, or the audience for the output. Also review prompts when users repeatedly correct the same error or when a supposedly helpful handoff creates extra manual work.

Keep a small record for each frequently used prompt:

• The task it supports and its intended user
• The required inputs and any sensitive fields to exclude
• The expected output format
• Common failure modes and the checks that catch them
• The last date it was tested

Start today with one recurring task. Save the prompt, test it on three representative examples, note where the output needs correction, and update the instructions. Then share the tested version with the people who perform the same task. Over time, this creates a practical library of AI prompts that supports consistent work without removing human judgment from the process. For more reusable examples, explore the practical prompt library for ChatGPT and AI assistants at work.