How to Talk to AI: Three Frameworks to Get Better Results with AI

What is Prompting and Why is it Important?

Prompting is the process of giving instructions to an AI system. It refers to how we talk to AI and a prompt can be a question, a task or even a set of guidelines that tells the AI what you want it to do. A good prompt sets up the conversation so that the AI knows what role you want it to play, what you want and how to deliver an answer to you.

Think of it like briefing a colleague. If you ask, “Write me a report,” you might get something very general. But if you say, “Write me a two-page report for senior management, summarizing the risks of our upcoming project, using clear language and bullet points,” you are far more likely to get something useful.

Prompting matters because:

• Context drives quality
  The AI does not know your goals unless you explain them and the more context you provide, the more accurate and tailored the response.
• Efficiency improves
  Clear prompts reduce the need for back-and-forth corrections, saving time.
• Consistency increases
  Standardized prompts can help teams get reliable outputs across different use cases.
• Creativity unlocks
  By experimenting with different styles of prompts, you can generate a wider range of ideas and solutions.

Prompting isn’t just about asking questions, it's about learning how to communicate with AI in a structured way so that it becomes a more effective partner in your work.

In this post we’ll walk through three simple frameworks you can use to start conversations with AI. These are the ones we use more often and they’re especially useful if you’re just beginning to explore how AI can support Architecture, Engineering and Construction (AEC) workflows.

Contents

1. Framework 1: Role + Goal + Format

2. Framework 2: C.O.T.E. (Context, Objective, Task, Example)

3. Framework 3: R.A.C.E. (Role, Action, Context, Expectations)

4. Try Them Out: A Practical Exercise (with answers)

Framework 1: Role + Goal + Format

This is our go-to for quick, focused tasks. It gives AI three essentials: a role, a goal, and a format. That way, it knows who it is, what it needs to achieve, and how to present the answer.

The structure looks like this:

 “You are a [role]. Your goal is to [objective]. Please respond in [format].”

It works because you’re giving the AI three essential things:

• A role so it knows who it’s pretending to be
• A goal so it knows what you want to achieve
• A format so it knows how to present the answer
  

Example:
“You are a project engineer. Your goal is to draft a site safety email to subcontractors reminding them about working near live edges. Please respond in email format.”

The result is clear, professional, and ready to refine.

Another variation:
“You are a project coordinator. Your goal is to write a client-ready project update. Please respond in bullet points.”

This framework is simple, fast, and highly effective for everyday communication.

Framework 2: C.O.T.E. (Context, Objective, Task, Example)

COTE is best when you’re dealing with more complex or creative work. It gives AI more background detail upfront so its response is aligned with your situation and intent. The structure looks like this:

• Context: What’s happening right now?
• Objective: What outcome do you need?
• Task: What do you want the AI to do?
• Example: What tone, style, or format should it use?
  

Example:

• Context: We’ve had multiple breaches of fall protection protocol on Level 4 over the last two weeks.
• Objective: I need to send a strong reminder to all subcontractors that violations will have consequences.
• Task: Write a short but firm email addressing this issue.
• Example: Tone similar to how a superintendent might talk—direct and serious, but professional.
  

This kind of detail allows the AI to produce an email that’s aligned with your intent and your audience. Instead of filling in the blanks, it has the whole picture.

You can apply the same structure to client updates, progress reports, or even technical explanations.

Framework 3: R.A.C.E. (Role, Action, Context, Expectations)

RACE shines when who you’re speaking to is just as important as what you’re saying. It’s great for stakeholder communication, toolbox talks, and client-facing updates where tone and framing matter.

The structure looks like this:

• Role: Who are you in this conversation?
• Action: What are you trying to produce?
• Context: What’s driving the need?
• Expectations: How should the output look and feel?
  

Example:

• Role: You are a site safety officer writing a toolbox talk for a foreman to deliver to the crew.
• Action: Draft an outline for a 5-minute talk on fall protection.
• Context: Several recent incidents where workers weren’t clipped in near live edges.
• Expectations: Keep the language simple and direct, with a serious but supportive tone. End with a clear action step.
  

The output is a ready-to-use toolbox talk that matches both the urgency of the issue and the audience’s needs.

Try Them Out

Prompting is personal, there’s no single right way to do it and you don’t need to memorize every framework. Use the above as a guide and starting point that can help you structure conversations with AI until you find your own style.

Here’s a quick exercise:

Ask AI to “Write a site update email to the client after a weather delay pushed back concrete pouring.”

Then try it three ways:

1. Role + Goal + Format

2. COTE

3. RACE

Compare the outputs. Each one will give you a slightly different style of response. The best framework is the one that fits your workflow.

Answers below: