The Mistake of the Instant Result
Most people use AI like a vending machine. You type a request and expect the final product to pop out immediately. While this works for simple questions, it often fails for complex tasks like writing a long email or planning a project. When you ask for the final result in one go, the AI has to guess your preferences while it writes. This frequently leads to generic content or mistakes where the AI makes up facts to fill in the gaps.
To get better results, you must change your mindset from a worker to a manager. A good manager does not just say "do this project" and walk away. Instead, they ask for a proposal first. By breaking the process into two steps, planning and execution, you ensure the AI understands the boundaries of the task before it begins the heavy lifting.
Step One: Creating the Recipe Card
The first step is the architect phase. Instead of asking for a finished document, ask the AI to provide a detailed outline or a recipe card of how it intends to solve your problem. For instance, if you are planning a week of healthy meals, ask the AI to first list the types of cuisines and dietary constraints it will use. This allows you to catch errors in logic or style before any actual writing happens.
Think of this outline as a contract between you and the AI. You are using the AI for its high level reasoning, which is its ability to think through the structure of a problem. If the plan looks wrong, you can correct it instantly. This prevents the AI from wasting effort on a 500 word response that misses the mark entirely. You are looking for a map of the work, not the work itself.
Step Two: Executing the Plan
Once you approve the plan, you move to the execution phase. You can now tell the AI to follow its own approved recipe one step at a time. This keeps the quality high because the AI can focus its full attention on one specific piece of the puzzle rather than getting overwhelmed by the entire project. This approach is much more efficient than trying to edit a finished, but flawed, product later on.
If the AI gets stuck during this phase, you act as the human in the loop to provide guidance. You might say, "Step one looks great, now move to step two." This method ensures that the final output is exactly what you envisioned. By acting as the supervisor who provides checkpoints for approval, you transform the AI from a simple chat box into a highly reliable assistant that handles your administrative burdens with precision.
Key takeaways
- Always ask for a detailed outline or plan before requesting the final product.
- Treat the AI as a junior staff member who needs a manager to approve their strategy.
- Correct errors in the planning phase to avoid wasting time on bad drafts.
- Break large projects into smaller steps to help the AI maintain focus and accuracy.
- Use the AI's reasoning ability to build a map of the work before you start the writing.