website logo
← Back to Blog

Why your prompts aren't working.

April 28, 2026 Last edited: April 28, 2026 Why your prompts aren't working.

Cover photo by Patrick Fore on Unsplash.

The Communication Gap

“This is too long,” “Make it more professional,” “Just answer my question!!”

Does prompting feel like a losing battle? If it does, you might be the problem.

Let’s get you started writing better prompts using the C.A.R.E. method: Context, Action, Role, and Evaluation.

Context

AI is not magic; it is a sophisticated prediction tool.

Have you ever noticed that the predictive text feature on your phone keyboard gets more intelligent the more you use it? Large Language Models (LLMs) operate on the same principle.

By default, a new chat window doesn’t know anything about you, your business, or your project. Your first step is to give it what it needs:

  • The Background: Give 2–3 sentences describing your project, company, or target audience.
  • The Files: Don’t just describe your data, upload it. Modern models can “read” PDFs, spreadsheets, and images.
  • The “Why”: Tell the AI the ultimate goal of the task so it can prioritize the right information.

Action

AI models need clear instructions. Imagine walking up to a colleague and handing them a stack of documents to “analyze” without communicating your goals. You wouldn’t expect a meaningful result.

Specify a clear objective:

  • Be Verb-Centric: Use strong verbs like “Analyze,” “Synthesize,” “Draft,” or “Format.”
  • Define the Output: Say “Create a 3-column table” or “Write a 500-word blog post in Markdown.”
  • Set Constraints: Mention what not to do (e.g., “Avoid jargon” or “Keep the tone informal”).

Role

AI responds well to being given a title. This is a small but powerful step which helps the model think and act like an expert. Try adding one line to your prompts:

“You are a helpful [subject] expert who [desired action].”

Evaluation

AI can reiterate on its own work. Evaluate the response and explain to the model exactly what changes to make:

  • The Critique: Tell the AI what it got right and what it missed.
  • The Pivot: Ask it to change one specific thing (e.g., “Make the tone more punchy”).
  • The Self-Correction: Ask the AI, “What information are you missing that would help you do a better job?”

Let’s Try an Example

Paste each prompt in a fresh chat window and examine the difference.

Bad Prompt:

Write a summary of our recent marketing campaign. It was too expensive but we got a lot of views. Make it professional.

Better Prompt:

Role: Act as a Senior Marketing Operations Manager known for being direct, data-driven, and solutions-oriented.

Context: We just wrapped up the “Spring Surge” campaign. Our target budget was $50,000, but we spent $62,000 due to rising ad costs. However, our goal was 1 million impressions, and we actually achieved 1.8 million. This report is for the Executive Board.

Action: Draft a “Campaign Post-Mortem” report. You must: Use a professional, executive tone; include sections for Financial Overview, Performance Wins, and Optimization Strategy; and use a bulleted list for key takeaways.

If you’re ready to chat about improving your AI skills, reach out to me on linkedin.

Learn More

Let's Stay in Touch