Make AI Work for You

Imagine asking an AI assistant:

Plan a three-day trip to Chicago.

You would probably receive a usable itinerary. But it might include activities you do not enjoy, move too quickly, exceed your budget, or conflict with something you have already booked.

The problem is not that the answer is unusable. It is that AI had to guess what would suit you.

In this part, you will learn how to replace those guesses with useful direction.

By the end of this exercise, you will have one improved request you can save, adapt, and reuse.

The five parts of a strong instruction

Before asking AI for help, consider:

1. Goal: What do you want to accomplish?
2. Context: What should AI understand about your situation?
3. Source material: What notes, documents, examples, facts, or other information should AI use?
4. Requirements: What must the result include or respect?
5. Desired output: How should the result be presented?

You will not need all five parts for every request. Include the details that can meaningfully change the answer.

Here is a stronger travel request:

Help me create a three-day itinerary for Chicago.

Goal: Plan an enjoyable first visit for two adults. Context: We like architecture, museums, and local food. We prefer a moderate pace rather than filling every hour. Source material: We are staying in River North and already have an architecture cruise booked for 2:00 p.m. on Saturday. We have also saved the Art Institute and the Museum of Science and Industry. Requirements: Group nearby activities, include time for breaks, keep costs moderate, and do not move the existing booking. Flag any information that may have changed. Desired output: Organize the itinerary by morning, afternoon, and evening. Include approximate travel times, estimated costs, and a short list of details I should verify.

This version gives AI enough direction to create a starting point that fits the travelers.

The objective is not to write the longest possible prompt. It is to provide the details that materially improve the result.

Try it now

Choose one real request you could give AI today.

First, write the request as you normally would. Then review it using the five parts above. Add only the information that would improve the answer.

Use this template if helpful:

Help me [goal].

Context: [Relevant details about my situation] Source material: [Information, notes, examples, or documents to use] Requirements: [Priorities, limits, and anything that must be included or avoided] Desired output: [The format or structure I want]

Submit both versions—the request as you would normally write it and the improved request. Compare the two responses and notice what changed when AI had clearer direction.

Examine the result

Ask yourself:

• Does it fit my actual situation?
• Did it respect my priorities and constraints?
• Is the format easy to use?
• Did AI make assumptions I did not provide?
• What still needs clarification, verification, or my judgment?

If one part is weak, you do not need to start over. Try:

Keep the parts that already work. Improve [specific part] by [explain what is missing or needs to change]. Do not change the rest.

When using personal material, share only what the task requires. For travel plans, avoid passport numbers, full booking references, payment details, and other sensitive information. Verify current details such as prices, opening hours, availability, weather, and transportation before acting.

In the next part, you will use AI to make sense of a message or document and separate the facts, decisions, questions, next actions, and details that still require verification.

---

Imagine receiving a long message about an upcoming event. It includes a date, arrival instructions, a change of location, something you need to bring, and a question you must answer.

You finish reading but still wonder: What has been decided? What do I need to do? Is anything unclear?

AI can help, but a short summary may hide distinctions that matter. It might present a suggestion as a decision, overlook an unanswered question, or turn an uncertain detail into a fact.

In this part, you will use AI to organize information without losing those distinctions.

By the end of the exercise, you will have a clear, actionable review of one real message or document.

Separate the different kinds of information

Ask AI to identify:

1. Central message: What is this mainly about?
2. Important facts and claims: What information does the source directly state?
3. Confirmed decisions: What has actually been decided?
4. Open questions: What remains unanswered or unclear?
5. Next actions: What needs to happen, by whom, and when—if the source says?
6. Information to verify: What may be incomplete, uncertain, conflicting, or outdated?

The principle is simple: AI should distinguish what the source says from what it infers or suggests.

Information appearing in a source is not necessarily independently verified as true.

Choose how to provide the information

One option is to copy and paste the relevant text. You can also upload a document if your AI assistant supports it.

Use this instruction:

Examine the message or document below using only the information it contains.

Identify:

1. The central message. 2. Important facts and claims. 3. Confirmed decisions. 4. Open questions. 5. Next actions, including the responsible person and deadline only when stated. 6. Information that is uncertain, conflicting, or requires verification.

Keep facts separate from interpretations or suggestions. Do not invent decisions, responsibilities, or deadlines. If something is not stated, say so.

Source material: [Paste the relevant message, notes, announcement, article excerpt, or document section]

If the document is long, begin with the section relevant to your immediate need. More material is not always more useful.

Optional: work from a connected inbox

If your AI assistant can securely access a connected inbox, you may ask it to locate the relevant message before analyzing it. For example, ChatGPT Work can work with connected Gmail or Outlook Email accounts when those features and permissions are available.

Use this instruction:

Find messages from [sender] about [topic] between [dates, if known].

If several conversations match, list them and ask me which one to use. Then examine only the selected message or conversation using the six categories above.

Identify which messages support the facts, decisions, and actions. Do not reply, draft, archive, delete, label, mark anything as read, or make other inbox changes.

Connected features and available actions can vary by account and workspace. Always confirm that AI found the correct conversation before relying on its analysis.

Try it now

Choose one message or document containing several details. It could be an email, announcement, set of notes, short article, instructions, or part of a longer document.

Provide it directly or use the optional connected-inbox workflow. Then examine the result:

• Can every fact be traced to the source?
• Are decisions separated from suggestions?
• Were responsibilities and deadlines included only when stated?
• Are unanswered questions visible?
• Do the proposed actions follow from the source?
• Were qualifications, conflicts, or uncertainties preserved?

For an extra check, ask:

Show the sentence or part of the source that supports each fact, decision, and next action. Revise anything that is not supported.

Share only the information needed for the task. Remove confidential or sensitive details when possible. Access to a message does not give AI permission to reply, share it, or change your inbox. You remain responsible for verifying important information and deciding what to do.

In the next part, you will use accurate information and context to turn a complicated situation into a clear, purposeful message.

---

Imagine that your internet service keeps failing after a technician has already visited. You have an appointment confirmation, several support messages, troubleshooting notes, a reference number, and an unexpected charge.

You need to contact the provider again. The individual facts are available, but explaining the complete situation without becoming vague, repetitive, or confrontational is difficult.

AI can help turn that history into a clear message—provided that it understands the audience, purpose, relevant facts, tone, and response you need.

By the end of this exercise, you will have a message that makes a complicated situation easier for its recipient to understand and act upon.

Define the communication before drafting it

Consider five elements:

1. Audience: Who will receive the message?
2. Purpose: Why are you contacting them?
3. Required information: What facts must they understand?
4. Tone: How should the message sound?
5. Desired response: What do you want the recipient to do?

A strong message does more than sound polished. It helps the recipient understand what happened, see the relevant evidence, and respond to a specific request.

Work from the information you have

If you completed yesterday’s exercise, you can use its organized facts, decisions, open questions, next actions, and verification points as the starting material for today’s message.

You can paste the relevant facts, notes, or previous messages into your AI assistant. Remove unrelated or sensitive information first.

Use this instruction:

Help me prepare a message about an unresolved problem.

Audience: [Who will receive it] Purpose: [Why I am contacting them] Required information: [Dates, events, previous actions, reference numbers, charges, promises, or other relevant facts] Tone: [For example: calm, firm, and respectful] Desired response: [The resolution, explanation, action, or update I am requesting]

First, organize the facts into a clear sequence. Identify any missing, uncertain, or conflicting information. Do not draft the message yet. Show me the factual sequence and wait for my review.

Review the sequence and correct or clarify anything necessary. Once you approve it, continue with:

Using the factual sequence I approved, provide:

1. A useful subject line. 2. A concise message explaining the situation. 3. A specific and reasonable request. 4. A shorter version suitable for chat, if needed. 5. A checklist of evidence or attachments I may need.

Do not invent policies, rights, promises, diagnoses, or facts. Preserve uncertainty where the available information is unclear.

A polished draft built on an incorrect history is still incorrect, which is why the factual review comes before drafting.

Optional: gather the history from a connected inbox

If your AI assistant can securely access a connected inbox, you may ask it to locate the relevant correspondence. For example, ChatGPT Work can work with connected Gmail or Outlook Email accounts when those features and permissions are available.

Find messages from [company or sender] about [problem or topic] between [dates, if known].

Identify the conversations you plan to use and ask me if the scope is unclear. Reconstruct the history, including important dates, actions, charges, promises, reference numbers, and anything still unresolved or conflicting.

Show me the factual history for review. After I approve it, draft the message using the five communication elements above.

Do not send anything or archive, delete, label, or otherwise change any message.

Try it now

Choose a real communication that requires more than a simple reply. It might involve an unresolved service problem, a request supported by evidence, a correction, or an update that depends on previous conversations.

Define the five elements, provide or retrieve the relevant history, and review the factual summary before asking AI to draft.

Then ask:

• Can the recipient understand the situation without reading the entire history?
• Are the facts accurate and relevant?
• Is the requested response specific and reasonable?
• Does the tone support the purpose?
• Has uncertainty been preserved?
• Does the message still sound like something I would send?

For an optional improvement, use:

Read this from the recipient’s perspective. Identify anything unclear, unsupported, unnecessarily long, or difficult to answer. Revise only those parts.

Share the minimum personal or account information needed. Do not provide passwords, payment-card details, security answers, or complete identity numbers. AI can organize and draft, but you remain responsible for verifying the facts, approving the final message, and deciding whether to send it.

In the next part, you will turn a real goal into a practical plan that respects your priorities and constraints.

---

Suppose you want to improve your spreadsheet skills so you can use them for a real personal or professional task.

You ask AI to create a learning plan. It returns a long list covering formulas, charts, dashboards, automation, and advanced analysis. The plan looks impressive, but it assumes more time than you have and does not help you decide what to do first.

A useful plan must fit your actual situation.

In this part, you will use AI to turn a goal or unresolved problem into a realistic starting plan.

By the end of the exercise, you will have a clear outcome, an ordered set of steps, and an immediate action you can take.

Give the plan something real to work with

Define six elements:

1. Outcome: What should be different when the plan succeeds?
2. Constraints: What limits must it respect?
3. Priorities: What matters most, and what can wait?
4. Stages: What are the main parts of the work?
5. Immediate actions: What can you realistically do first?
6. Risks and missing information: What could affect the plan or still needs to be determined?

The principle is simple: a useful plan connects a clear outcome to your real constraints, priorities, and next actions.

As on Day 1, include only the context that can meaningfully change the result. As on Day 2, distinguish known information from what still requires verification.

In this example, the goal might be to organize household expenses and create a clear monthly summary. You might have time for three 30-minute study sessions each week, want to use only free resources, and need a working result within four weeks.

Those details change the plan. AI can now prioritize the skills needed for the real task instead of proposing an entire spreadsheet curriculum.

Turn your goal into a plan

Use this instruction:

Help me turn this goal into a realistic plan I can begin.

Goal: [What I want to accomplish] Why it matters: [Why I want this result] Current situation: [What I have already done or know] Constraints: [Available time, budget, tools, responsibilities, experience, or deadlines] Priorities: [What matters most]

Before creating the plan, identify any missing information that could materially change it. Ask only the questions you need, and wait for my answers. If no material information is missing, say so and proceed.

Then create a plan that includes:

1. A clear and realistic outcome. 2. The main stages in a sensible order. 3. The first three actions I can take. 4. What can wait until later. 5. Likely obstacles or risks. 6. Information I still need before relying on the plan.

Do not assume that I have unlimited time, money, knowledge, or access to tools.

AI can organize the path, but you still decide which tradeoffs are acceptable.

Try it now

Choose one real goal or problem that requires several steps. It could involve learning a useful skill, completing a home project, preparing for an important responsibility, or moving a personal, professional, or creative project forward.

Complete the five fields in the instruction and answer any necessary questions. Then examine the plan:

• Is the outcome specific and realistic?
• Does the plan respect my constraints?
• Did my priorities affect what was included?
• Do the stages follow a useful order?
• Are the first actions clear enough to begin?
• Are risks and missing information visible?
• Did AI make assumptions I should correct?

For an optional improvement, use:

Review this plan against my stated constraints. Identify anything that is too ambitious, vague, unnecessary, or out of order. Revise only those parts.

Choose one immediate action from the revised plan and decide when you will take it. A plan becomes useful when it leads to action.

AI cannot determine how much time, money, effort, or risk is acceptable to you. Verify current information and seek appropriate professional guidance before relying on plans involving health, finances, legal matters, safety, or other consequential decisions.

In the next part, you will combine everything you have practiced into a personal AI workflow you can use again.

---

Imagine that you need to compare several services, subscriptions, courses, products, or work tools before choosing one.

If you simply ask AI, “Which option is best?” it must guess what matters to you. It may also treat promotional claims as facts or rely on information that is no longer current.

A better approach is to use AI as part of a repeatable process: define the decision, provide useful information, create a comparison, examine it, improve it, and then decide what to do.

That process applies far beyond comparisons.

In this part, you will combine the skills from the previous four parts into a personal workflow for a task you handle repeatedly.

By the end of the exercise, you will have a workflow you can save, adapt, and use again.

The complete workflow

1. Define the outcome

Decide what success means and what decision or action the result should support.

1. Provide useful context and material

Include the priorities, constraints, messages, notes, documents, examples, or facts that can materially change the result.

1. Generate a starting point

Ask AI to organize, compare, explain, draft, or plan.

1. Examine and verify

Check the result against the source material. Look for unsupported assumptions, missing information, and claims requiring verification.

1. Improve the result

Correct errors, add necessary context, and request focused changes while preserving what already works.

1. Turn it into action

Decide the next step, take it, and save the useful parts of the process for later.

This is what you have practiced throughout the five parts of this lesson: giving AI better direction, making sense of information, communicating clearly, and creating realistic plans.

The goal is not to remove your involvement. It is to make your involvement more deliberate and effective.

Build your personal workflow

Choose one task you expect to handle again. It could be:

• Understanding information and identifying actions.
• Preparing a message from previous correspondence.
• Planning a project within real constraints.
• Comparing products, services, courses, or tools.
• Turning notes into a useful summary.
• Preparing for a recurring responsibility.

Then use this instruction:

Help me create a repeatable AI workflow for this task:

Task: [What I regularly need to do] Trigger: [What tells me it is time to do it] Outcome: [What useful result or action I need] Available information: [Notes, messages, documents, examples, or facts AI can use] Constraints: [Limits, requirements, or priorities] Important checks: [What I must review or verify before acting]

Before designing the workflow, identify any missing information that could materially change it. Ask only the questions you need and wait for my answers. If nothing material is missing, say so and proceed.

Then organize the workflow into these stages:

1. Define the outcome. 2. Provide useful context and material. 3. Generate a starting point. 4. Examine and verify it. 5. Improve it. 6. Turn it into action.

For each stage, explain what I provide, what AI does, and what remains my responsibility. Finish with a reusable instruction and a short review checklist. Do not automate or take external action.

Test the workflow

Use it once with a real task. Then ask:

• Did it begin with the outcome I actually needed?
• Did AI receive the right information?
• Were source facts separated from assumptions?
• Did the result help without replacing my judgment?
• Were important claims identified for verification?
• Did the process lead to a concrete action?
• Could I repeat it without rebuilding everything?

If the workflow still depends on vague instructions or information you might forget next time, use:

Review this workflow as if I needed to use it again next week. Identify any step that depends on memory, vague judgment, or missing information. Revise only those parts so the workflow is clearer and easier to repeat.

Give the workflow a short, recognizable name. Save the final reusable instruction and review checklist somewhere you can easily find them the next time the trigger occurs.

If you would like help naming it, ask AI to suggest a few clear options. You make the final choice.

Avoid saving sensitive information unnecessarily. Connect and use only the tools needed for the task. AI can help retrieve, organize, compare, draft, and plan. You remain responsible for verifying important information and approving consequential actions.

You have completed all five parts of the lesson—and created a practical process you can continue improving as you use it.

You can now define what you need, give AI useful direction, examine the result, improve it, and turn it into action.

Find a starting point for your next task

When you face a different task, you do not always need to begin with a blank page. The Mastering AI Prompt Finder helps you search for what you want to accomplish and find a practical prompt you can adapt using the methods from this lesson.

Search. Find. Adapt. Use.

Explore the free Prompt Finder

Choose the area you would most like to explore next, and we’ll point you toward the most relevant Mastering AI resources.

Select one:

• Work and productivity
• Everyday life
• Career development
• Business or income
• Creativity and projects
• Understanding AI confidently