Challenge: Find and Use a New AI Prompt
If you talk to an AI expert, someone like Scott Schang over at Redefining Business Intelligence, you’ll learn that the quality of the output of any Large Language Model is dependent upon the quality of the prompt and the data the model has been trained on and/or provided.
That’s a mouthful, but it basically means garbage in delivers garbage out. This has pretty much been true of data processing systems since they were invented. It still holds true for AI.
AI models aren’t mind readers. They rely entirely on the data and instructions they’re given. A vague prompt or sloppy input will result in equally vague or irrelevant output. When clients provide clear, structured data and a well-constructed prompt, they dramatically increase the odds of getting something insightful or on-brand.
The difference is that, unlike data processing systems of the past, you can talk to AI and ask it for help in providing the information it needs to give you a less garbage-like answer.
Here’s a challenge: work with an AI model of your choice to refine a prompt that your team can use repeatedly to increase efficiency. Here’s how to do it.
Step 1: Identify a Repeatable Task
Choose a task that your team or partners do often. Tip: Choose a task where quality varies depending on the input or clarity of the request.
Step 2: Tell the AI What You’re Trying to Build
Start with a clear goal. Try something like: “I want to build a reusable AI prompt template I can give to my team to help them [TASK]. Help me design this.” Then explain: the desired outcome, who the end-user is (your employee? a client?), and what makes a great output vs. a bad one.
Step 3: Ask the AI What It Needs
Prompt the AI with: “What inputs do you need in order to do this task well?”
The AI should respond with a list.
Step 4: Build the Master Prompt Together
Now ask the AI: “Using the above, please write a master prompt with [brackets] or <placeholders> for the variable parts.” You’ll get something like: “Write a <length> press release announcing <newsworthy event>. The tone should be <tone> and target <audience>. Be sure to include key messages like <key message 1> and <key message 2>. Cite data such as <data point>.” You can refine this until it fits your standards.
Step 5: Test It
Use the master prompt with a real example. Fill in the placeholders and see what the AI produces. Ask: “How would this prompt perform if someone less experienced used it? What can we add or clarify?” Revise the template with the AI’s help until it’s foolproof.
Step 6: Document and Distribute
Once refined, add example inputs, note formatting requirements, store it in your team playbook, SOP doc, or project brief. Encourage employees to try it, tweak as needed, and provide feedback.
If you’d rather not do this work on your own, you can hire a prompt engineer (annual salary now c. $163,500), or call us at RGA. If we can help you, and we probably can, we know the smart folks over at RBI who absolutely can.
