How to Communicate With AI

Boost Your AI Skills

AI is a rapidly developing field — new tools, models, capabilities, and features develop regularly. AI can do many different things and help you with your routine, work, and business tasks: from daily planning to executing complex automated workflows.

Depending on the tool and model, you can speak to it, attach documents and screenshots, analyze data, run workflows, and generate images, audio, and video.

Yet all these options don't automatically make work easier. Without a clear system, you can waste time trying random prompts, correcting weak answers, and explaining the same context again and again. This course helps you with this. The goal is simple: get more done with less effort so you can make your work, business, and everyday life easier.

You'll walk through the full cycle: from clear communication with AI to turning your inputs into repeatable workflows. Finally, you'll set up a personal assistant designed to fit the way you work.

The first question is: how do you communicate with AI efficiently to get the most relevant output for your specific case with minimal effort? This lesson answers this exact question.

Explore AI Capabilities

There are many types of AI available on the market right now. They differ from one another by the kind of work they help you do. Many tools combine several capabilities, so one assistant may chat with you, read files, search sources, create media, and complete a workflow.

Check this table to explore which type of work AI can help you with:

For most of your tasks, you’ll need a single AI tool with many capabilities: one that supports text, voice, files, images, research, and simple workflows. It can cover many everyday needs, keep your setup simple, and reduce subscription costs. You can add an extra tool when the task genuinely needs it, for example, the specific task such as audio or video generation.

Although there is a wide variety of AI tools and their capabilities, the principles in this course apply to every type. The first rule matters most: start with a clear outcome.

Start With The Outcome

Whether you type, speak, or attach a file, the main challenge is the same: AI needs to know what useful result you want.

You don't need a special prompting technique or a complex formula — your input can be short or detailed. What truly matters is a clear result and enough relevant context to achieve it. So start with the outcome you want to make the input clear and get the relevant output. Think of briefing a colleague who's new to the task: you'd explain what you need them to deliver and share the details they need to do it well.

For example, you need to promote a workshop:

• Name: Plan Your Week in 20 Minutes
• Free 45-minute online workshop
• When: Thursday, 12:00-12:45
• For small-business owners
• Registration required; only 30 places
• Tone of my communication is calm, practical, without hype

You decide to use AI to write a post for you, and you're choosing between the two prompts to accomplish your task properly. As a result, you want to have a well-written post. Look at these two prompts and the AI outputs they produced to decide which one worked better. Compare based on what you get from the AI.

Input B worked better because its details were useful: it named the result, audience, facts, tone, and limits. Input A focused on excitement instead of the information the reader needed. The difference was relevance, not length.

A useful outcome provides the right context or source, adds the limits that matter, describes the format you want, and explains how to recognize a finished result.

A simple task may need only one sentence: "Summarize this paragraph in one sentence for a reader who is new to the topic." Add more only when the result depends on more.

Give AI the Best Source

AI can work with several kinds of input where supported. Use the format that makes your task clearest, or combine formats when each adds something useful. For example, when creating a website, a text brief can explain what a website needs to do, while a screenshot of a quick sketch can show the layout you have in mind.

Then tell AI what result you want from your input. Let’s look at what the input can be and how to best instruct AI for each type:

For each task, ask: What would AI need to read, see, or hear to work from the clearest available evidence? Choose the input that carries that information most directly. To practice, match each task to its most useful starting input.

The best input saves you from retyping and gives AI something real to work from. Don't forget the instructions: you still need to tell AI explicitly what to do.

Your instructions turn each material into a specific task: they give AI a clear direction and help it produce a better result.

For example, the same spreadsheet can support several tasks: checking calculations, comparing months, finding unusual changes, creating a chart, or writing a summary. Your instruction selects the result that matters to you.

The instructions' details change with the task:

• For a text, specify the audience and tone
• For an image, describe the composition and intended use
• For a video, define the sequence and duration
• For an agentic workflow, set permitted actions and approval points

Adding Context

Whether you're working in a chat, through a voice conversation, with a document, or on an image or a video, context is the same everywhere. The same rules apply for every tool, every input type, and every job:

1. Say what result you need.
2. Give AI the source it should use.
3. Add the context and limits that matter.

You've already defined the result and the source, so now you'll focus on the context.

Add Only What Changes the Work

Context earns its place when changing it could change the answer. Useful context may include:

• The audience or user
• The purpose and where the result will be used
• Facts and source material AI cannot safely infer
• Real limits and actions AI must avoid
• The role or point of view AI should use
• The tone that fits the reader and situation
• The feedback style you want: supportive, direct, or challenging
• A reference when quality is easier to show than describe

You don't need to provide all these details. Decide based on the task and desired output.

But some details are more universal and are useful for most tasks: role, tone, and feedback style when you want AI to evaluate your input. They deserve a closer look because each one changes a different part of the output.

Check the table to see when each variable helps and how it affects the instruction:

For example, you need to write a short message containing two facts: the team meeting has moved to Thursday at 10:00, and everyone should bring their latest project status and updates. Look at how the role will change the message:

The editor prioritizes natural wording. The project manager makes the timing and action more visible. The role changes what AI focuses on and how it organizes the answer. Now keep the same task and change only the tone.

The facts stay the same, while the wording and feeling change. Tone helps the message fit its audience and situation. Guidance creates a visible change in the focus, wording, or structure while keeping the source facts intact. Now apply that test to a more detailed workplace update.

Useful Details For Various Output Types

Check the most common context details that may help different output types:

For example, the workshop notes you turned into the post earlier could also support an image or a short video. Each format needs different details and has a different definition of a finished result:

• Text: "Write a calm post for small-business owners using only the supplied workshop facts." It works when readers can find the workshop name, time, free entry, 30-place limit, and a clear invitation to register.
• Photo or image: "Create a realistic-looking horizontal image of four small-business owners planning around a table in natural daylight." It works when the visual supports the planning theme and leaves room for the facts the post needs to communicate.
• Video: "Create a six-second vertical video: an overloaded weekly calendar becomes a clear plan, then ends on an uncluttered desk. Calm movement." It works when viewers can see that shift, with room to add event details afterward.

Each of these requests explains the result, source, limits, output, and what a good result should do. AI tools support different file types and outputs, so check what your tool can handle before you rely on it.

Before sending a request, ask: What must this output help someone see, understand, or do? Use that answer to choose the details that matter and to check the result. AI tools support different inputs and outputs, so check what your tool can handle before relying on a specific format.

Polish Your Result

AI shouldn't guess an important detail. Give it a simple rule: Ask when missing information could change the result, permission, facts, or safety.

This works well with voice inputs, when you speak naturally instead of drafting a polished prompt. Just say: “Summarize what you think I need. Ask questions if anything is unclear.”

The same rule helps with documents and screenshots. If a page is missing, a column heading is unreadable, or two sources disagree, AI should point it out instead of assuming some facts that might not be true. Look at this case and decide which details are important to provide and which questions to answer first.

Iterate With Focused Feedback

The first result shows how AI understood your request. Use it to spot what's missing, unclear, or ignored.

Iteration means improving the result you already have instead of starting over. Change one or two things at a time so you can see what helped. Use a follow-up when most of the result works; rewrite the original request when the goal, source, or limits were wrong.

Instead of "make it better," give a focused follow-up:

1. Say what should remain.
2. Name the visible problem.
3. State the exact change you need.
4. Ask AI to check the new result against your rules.

Check the possible iteration prompts with focused feedback for different outputs:

Good iteration is a short loop: review the result, name the specific gap, revise, and check again. When a task has several connected parts, the next technique is to split the work into stages and check each one before moving on.

Check Before You Use the Result

AI can compare its work with your rules, but that isn't the same as checking the output yourself. A confident answer can still contain a false fact, a made-up source, a wrong number, or media that gives a false impression.

Good news: most AI models provide the sources they ground their answers in. Don't hesitate to check the sources yourself to make sure the answer is truthful and relevant.

Now you know the rules of effective AI communication. You don't have to find the perfect words or write a long prompt. Just state a clear outcome, provide the right source, add relevant context, and make a final check. That gives AI direction while keeping you in control.

These principles work regardless of the input and output type: whether you type, speak, create images, or videos. In the next lesson, you'll move from individual requests to repeatable productivity and build an AI-powered workflow from scratch.