Prompt Engineering Assistant — Kearney AI Sandbox

What this tool does

AI tools like ChatGPT and Gemini produce much better results when given clear, well-structured instructions. Writing those instructions (called "prompts") is a skill that takes time to develop.

The Prompt Engineering Assistant skips that learning curve. You describe what you are trying to accomplish in plain, everyday language. The assistant analyzes your intent, identifies what is missing, and returns a complete, structured prompt that you can paste directly into ChatGPT or Gemini to get a high-quality response.

The basic idea

You know what you need. The assistant knows how to phrase it so AI tools respond well. You provide the subject matter knowledge; the tool handles the structure.

Step-by-step walkthrough

Go to the Live Demo section

From the home page, click Live Demo in the navigation bar or scroll down to the demo section. You will see a text box on the left labeled "Your Rough Idea" and an output panel on the right.

Try one of the examples or write your own

Four Kearney-specific examples are available at the top of the input panel. Click any of them to load a pre-written idea relevant to federal financial management work. Or type your own rough idea in the box. Incomplete sentences and rough descriptions work fine.

Example rough idea you might type

I need to explain to a non-technical client what a FISMA audit actually means for their agency and why it matters for their budget request next year

Click "Refine Prompt"

The assistant sends your idea to Gemini and returns two things: an Analysis (a brief explanation of how it interpreted your request and what assumptions it made) and a Refined Prompt (the structured instruction ready to use). This usually takes 5 to 15 seconds.

If the analysis shows that the assistant misunderstood something, adjust your original idea and try again. Two or three iterations almost always produces a better result.

Open in ChatGPT or Gemini

Once you have a refined prompt you are happy with, use the buttons at the bottom of the output panel:

Open in ChatGPT ↗ opens a new tab with your prompt already submitted. You land directly on the response.

Open in Gemini ↗ copies your prompt to the clipboard and opens Gemini in a new tab. Click the message box, press Ctrl+V (Windows) or Cmd+V (Mac) to paste, then press Enter.

Use the AI response in your work

Read the response carefully. Treat it as a strong first draft, not a finished product. You can continue the conversation in the same ChatGPT or Gemini window. Ask follow-up questions, request different formatting, or ask it to expand specific sections.

Practical examples for Kearney work

Summarizing audit findings for a non-technical audience

I have a list of FISMA audit findings and I need to explain them to an agency CFO who is not a security expert. They care about financial risk and timeline, not technical control numbers. I want a clear summary that prioritizes what matters most for their budget.

Drafting a client briefing

I need talking points for a 15-minute briefing to a senior executive about a cybersecurity maturity assessment we completed. They scored low and need to improve within 18 months. I want direct language that explains where they are and what the first steps should look like without overwhelming them.

Analyzing a regulatory requirement

Explain what CMMC Level 2 actually requires in practical terms for a small defense contractor who has never gone through a formal cybersecurity assessment. What are the most common gaps and how long does remediation typically take?

Writing proposal language

I need to write the scope section for a task order proposal. The engagement is a 12-month financial management improvement project for a DoD agency. I want scope language that is specific enough to protect us from scope creep but flexible enough to adapt as we learn more.

Data policy reminder

Do not enter real client names, contract numbers, classified information, or any data covered by Kearney's AI Use Policy into this tool. Use descriptions and generic language when testing. Synthetic and hypothetical scenarios are always safe to use.

Try the Prompt Assistant now →