How to Write AI Prompts That Actually Work (Stop Wasting Time)

---

“I tried AI once. Didn’t get anything useful.”

Sound familiar?

The U.S. Census Bureau reports that only 5% of businesses currently use AI. In construction and design? That number drops to 1%.

But here’s what’s interesting: IBM found that 35% of businesses globally were already using AI back in 2022. So why is adoption in the U.S.—and especially in our industry—lagging so far behind?

The answer isn’t that AI doesn’t work. It’s that most people don’t know how to write AI prompts that get useful results.

There’s an old programming principle: garbage in, garbage out. AI is no exception. You need high-quality input to get high-quality output. Learning how to write AI prompts isn’t about understanding algorithms or becoming a tech expert—it’s about revealing and understanding your own thinking process.

(Which, granted, sounds more philosophical than practical. Stick with me.)

The Problem: We’re Asking AI What It Thinks

Andrej Karpathy, one of the leading AI researchers, points out something that changes how we should use these tools: Don’t treat LLMs as entities. Treat them as simulators.

What’s the difference?

When you ask AI, “What do you think about this strategy?” you’re pretending it has opinions. It doesn’t. There is no “you” inside ChatGPT.

But when you ask, “What would a CFO, a creative director, and a skeptical board member say about this strategy?”—now you’re using AI as a tool to simulate perspectives you might not naturally consider.

Same tool. Different approach. Better output.

This isn’t about the AI being smarter or dumber. It’s about us using it more effectively. Rather than asking for the model’s own view, we can use it to approximate the voices of people whose perspectives would actually challenge and sharpen our reasoning.

So how do we do that? How do we shift from asking AI what it thinks to using it as a simulator that gives us genuinely useful results?

It starts with knowing how to write AI prompts effectively.

The Framework: 4 Steps to Prompts That Work

When you’re learning how to write AI prompts, the key is breaking down the steps needed to frame any problem or task. The biggest issue is that most of us skip directly to the task description without providing adequate background information.

Think about it this way: when you delegate to your lead designer, you don’t just say “design a kitchen.” You provide context about the client, the budget, the style preferences, the constraints. AI needs the same treatment.

Here’s the framework:

Step 1: Tell AI the Role It Needs to Play

One of the most important things to understand about how to write AI prompts is that specificity matters. By default, ChatGPT and other AI interfaces assume the role of “assistant.” Which might be fine if you need an assistant, but if you’re trying to get help with research, need code written, want an article that will rank in search engines, or you’re drafting emails to clients about a price increase—being specific about the role AI is playing becomes critical.

Most AI services now have different kinds of GPTs or personas pre-programmed for you to choose from. Pick the one that represents the role you need AI to have for the output you need. If one doesn’t exist, you can inform AI directly.

For example:

• “You are an experienced design business consultant who specializes in client communication.”
• “You are a technical writer creating documentation for non-technical users.”
• “You are a sales coach helping craft difficult conversations.”

This is where Karpathy’s simulator concept becomes practical. Instead of asking AI to be a generic assistant, you’re asking it to simulate a specific perspective or expertise.

Step 2: Give AI Adequate Context to Play Its Role

We each have decades of experience and expertise that inform our thinking and decision-making. The more of this information you can share with AI, the better output it can give you.

Consider feeding in:

• Brand voice documents
• Similar examples of what you want
• Your values or company values
• Articles or reference materials
• Previous successful versions of similar work
• Constraints or requirements

This step can be enlightening because it forces you to think about all the influences on your thinking process. What information sources would you call on in your experience and expertise to perform this task? Feed that information to AI.

What context should I give AI for business tasks?

Start with the basics: your industry (luxury kitchen & bath design), your typical client profile, your business size, and any relevant constraints (budget, timeline, brand standards). Then add specifics: previous examples of similar work, outcomes you want to avoid, and the “why” behind the request. The more context AI has about your business reality, the more useful its output becomes.

How do I know if I’m giving AI enough information?

If the output feels generic or misses important nuances of your business, you haven’t given enough context. A good test: could this output apply to any business in any industry, or does it feel specific to your situation? Also, use Step 4 (asking AI what else it needs) to let it tell you what’s missing.

Step 3: Tell AI the Task You Want to Perform

This is typically the most obvious step for users, but even an expertly crafted request won’t get you anywhere without those first two steps.

If you’re a good delegator, you probably have this one nailed. If you aren’t great at delegating, you may need to think through the various steps in achieving the outcome you want and define your task accordingly.

You might even need to break the steps into separate requests:

• First request: “Help me identify the key decision points in this process”
• Second request: “Now create a flowchart showing those decision points”
• Third request: “Write the email template for decision point #3”

The task should be specific and actionable. Not “help me with marketing,” but “write three email subject lines for announcing our new showroom to existing clients.”

Step 4: Ask AI What Else It Needs to Perform the Task

This is the key that unlocks the next level when you’re figuring out how to write AI prompts. Use AI to tell you how to get what you want from AI.

When you tell AI, “Ask for anything you need before you start this task,” it will kick back a list of additional information it needs or questions to answer. Providing this additional data almost always results in better output.

This is where the bulk of the work for the human comes in. AI is a powerful tool, but it requires high-quality input from a human. It’s not replacing your expertise—it’s extending it by forcing you to articulate what you know so it can simulate perspectives and outputs you need.

Where AI Becomes Practical for Design Businesses

Here’s where this gets interesting for your business specifically.

Meeting notes. Client conversations. Vendor calls. That showroom discussion where the client mentioned three different things they wanted but you were focused on cabinet configurations and now you’re not sure if they said they wanted the farmhouse sink in the island or the perimeter.

AI meeting note tools like Fireflies, Plaud, or Otter can capture those conversations automatically. But the real power isn’t just in capturing—it’s in what happens next when you know how to write AI prompts that extract exactly what you need.

With good prompts, you can tell AI: “You are a luxury design project manager. Extract all decision points, open questions, and action items from this meeting transcript. Format them as a follow-up email to the client that confirms what we discussed and what we’re doing next.”

That’s the 4-step framework in action:

1. Role: luxury design project manager
2. Context: this is a meeting transcript from a client consultation
3. Task: extract decisions, questions, action items and format as follow-up email
4. What else needed: the transcript itself

No more 2 AM panic wondering what you forgot. No more “he said/she said” confusion about what was decided. No more depending on your memory to track every detail across multiple projects.

The AI isn’t doing your thinking. It’s capturing your thinking so you can focus on the strategic work only humans can do—understanding what the client actually wants, making design decisions, building relationships.

Building Your AI Skill Set

Understanding how to write AI prompts using these four steps will build your skill and the usefulness of AI to your business. It takes practice and iterations. But pretty soon, you’ll be creating a prompt library, updating your processes with which personas to use, and wondering how you ever managed without it.

The construction industry sitting at 1% AI adoption isn’t a badge of honor—it’s a massive opportunity. While others hesitate, you can be implementing systems that give you back time, reduce errors, and let you focus on the design work you actually love doing.

Start simple. Pick one repetitive task this week—maybe it’s drafting project update emails, or creating social media captions, or summarizing vendor pricing. Use the 4-step framework to practice how to write AI prompts. See what happens.

The technology is here. The tools work. The question is whether you’re going to learn how to write AI prompts effectively, or whether you’ll be part of that 99% still saying “I tried it once and didn’t get anything useful.”

Your choice.

Ready to see how AI can transform your business operations? Our AI Meeting Notes Implementation Guide walks you through exactly how to capture every client conversation, vendor discussion, and project decision automatically—without changing how you already work. Get the guide here.

---

Want to explore more ways AI can transform your business operations? Subscribe to the Elevated podcast for practical insights on leveraging technology and systems for business growth. New episodes provide actionable strategies you can implement immediately.

Prefer video content? Subscribe to our YouTube channel where we break down complex AI applications into clear, step-by-step guidance.