Reaching the Goal
Imagine you set a goal to clean your room. You put some toys away, picked up a few socks, and stacked some books. Then you looked around and thought: is this clean enough? Or do I need to do more? That moment — looking around and checking — is one of the most important parts of reaching a goal. Today we will learn how an AI agent knows when it has truly, really, completely reached its goal.
Done Is Not Always Done
Not every attempt at a goal ends in full success. Sometimes the work is almost done but not quite. Sometimes the plan had a step that got skipped. Sometimes the result is good but the goal was not fully met. Think about baking a cake. The goal was a fully frosted birthday cake ready to serve. You baked the cake — done. You made the frosting — done. But then you forgot to frost the cake before the party started. The cake was baked but the goal was not met! Or imagine your goal was to read five chapters of a book. You read four and got distracted. Four is close. But the goal was five — so it is not quite reached. Checking carefully at the end is how you catch these close-but-not-quite situations.
After finishing its steps, a smart AI agent checks whether the goal was truly met. It does not assume success — it looks at the result and compares it to the original goal. Only when the result matches the goal is the job really done.
How does an agent check whether it met the goal? It goes back to the original goal and asks a simple question: does the result match what I was trying to make happen? If the goal was find three articles about penguins, the agent counts: one article, two articles, three articles. Yes! Three found. Goal met. If the goal was book a hotel room for July 10 to July 15, the agent checks: is there a booking? Yes. Is it for July 10? Yes. Is it for July 15? Yes. Goal met. If the goal was write a thank-you note and send it, the agent checks: is the note written? Yes. Was it sent? It looks... oh, it was written but never sent! Goal not fully met. One more step needed. Checking is what catches that last missing step.
Flashcards — click each card to reveal the answer
Here is a story about an agent that almost skipped the checking step. Ruby asked her schedule agent to set up a reminder for her dentist appointment on Thursday at 2 PM. The agent created the reminder. It was about to say done! when it stopped and checked: is the reminder set for Thursday? Yes. Is it set for 2 PM? Yes. Is it a reminder for the dentist, not something else? Yes. Every part of the goal checked out. Only then did the agent say: done! Your reminder is set for Thursday at 2 PM for your dentist appointment. That extra moment of checking is what makes the difference between a reliable agent and an unreliable one.
Some people think checking takes too long and skip it. But a missed step can mean doing the whole job over again. A quick check at the end saves a much bigger problem later. Smart agents always check before declaring success.
Match each goal to the check the agent should make before declaring success.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
After finishing all its steps, what should a smart agent do before saying the job is done?
An agent's goal was to book a flight for Monday. It booked a flight for Tuesday by mistake. What does checking reveal?
The Goal Check Checklist
- Think of a small task you need to do this week — something like tidying a shelf, finishing a drawing, or helping set the table.
- Write down your goal clearly at the top of a piece of paper: I want to ___.
- Below the goal, write three things you will CHECK when you think you are done — specific things that will tell you the goal is truly met.
- For example, if your goal is to tidy your bookshelf, your checks might be: Are all the books standing upright? Are they sorted by size or color? Is anything sticking out over the edge?
- Do the task. When you think you are done, run through your three checks.
- Report back: did the checks reveal anything that still needed doing?