Skip to content
OPQAI.
Sourced beginner / 🎓 Academic & Research Free tools

Predict AI Results Before Running with Your Assistant

Job to be done: Prevent AI errors by having it predict and verify results before execution

🇳🇬 Ways to use this in Nigeria

Ideas to get you started, adapt to your situation.

  • Student

    Before submitting an essay, ask your AI assistant to predict its main arguments and structure, then compare to your draft to catch errors.

  • 9-5 employee

    Ask your AI assistant to predict the key findings of a market research report it will generate, then verify the output.

  • Entrepreneur

    Before asking AI to draft social media posts, have it predict the key message and call to action for your product launch.

What you’ll get

You will learn a simple, two-question technique to make your AI assistant predict the outcome of a task before it performs it. This method helps catch errors early by creating a gap between your expectation and the AI’s actual result, preventing you from building decisions on incorrect information.

Tools you need

  • AI Assistant (freemium): Any AI chatbot or assistant you use for tasks like coding, spreadsheets, research, or writing.

Steps

  1. Ask the AI to predict the result: Before asking your AI assistant to perform a task, prompt it to state what it expects the outcome to be and why.

    Before you run the task I'm about to give you, please answer these two questions:
1. What do you expect the result to be, and why?
2. How would you know if this were broken?

    You should see the AI provide a prediction and a method for verification.

  2. Perform your task: Now, give your AI assistant the actual task you want it to complete.

    [Paste your task here]

    You should get the result of your task.

  3. Compare the prediction to the result: Check if the actual result matches the prediction the AI made in step 1. If there’s a significant difference, it signals a potential error or misunderstanding.

    You should see the actual output from your AI assistant.

Original source

This technique was shared by gde03 on DEV Community. It’s a simple yet effective method to improve the reliability of AI assistants by adding a verification step before execution.

Notes & variations

  • Free tier viability: This technique works on the free tier of most AI assistants.
  • Common mistake: Trusting the AI’s explanation of a result without comparing it to your initial prediction. The explanation can often sound plausible even if the result is wrong.
  • Tip for better results: Be specific in your initial task prompt. The clearer your request, the more accurate the AI’s prediction and the easier it will be to spot discrepancies.

Keep going

More Academic & Research workflows