Skip to main content

How it works

AI is good at looking at refining known information, such as prioritizing a set task schedule. In our Tasks Tool, we give the AI all the information it needs to make an assessment.

1. Input Your Preferences​

You start by giving the AI a list of tasks: Task, Duration, and Priority. By doing this, you give the AI all the information that it needs to make an assessment.

When you ask AI to find information, it doesn't always give you the right information - because it can hallucinate. But if you give it solid information to work with, you'll get better results.

2. The AI system prompt​

Before we send anything, we send a system prompt.

Our system prompt instructs the AI on what it's intended to do. Being as specific as possible at this stage is important.

A system prompt preps the AI regarding what is expected, but doesn't include any customization. It's essentially the AI's role.

You are a task prioritization assistant. Based on the tasks provided, prioritize them by importance, considering the task description, priority level, and estimated time.

Output only a JSON list in the following format:
[
  {
  "description": "string",
  "priority": "High/Medium/Low",
  "time": number (in minutes)
  }
]

Return only the JSON list, with no additional text or explanations.
info

What is JSON? JSON is a structured data format. As you can see above, it tells the AI exactly what we expect returned.

Of course, you might not want to prompt in JSON. But the most important component is telling the AI exactly what you expect back from it.

3. The user prompt​

You filled out a form. But before I sent that to OpenAI, I turned it into a single prompt:

Here are the tasks to prioritize:
${JSON.stringify(tasks, null, 2)}

Note that this only works because we already sent all the context beforehand. By preparing the AI, we've affected how the AI is going to interact.

4. Caveats and concerns​

This is a fairly low risk task. You're giving the AI all the information and all it needs to do is place it in an order. It may not always understand what every task means, but it'll probably do a fairly decent job. Compare this with the recipe generator, in which a single instruction or a single ingredient could actually be disastrous.