Practice Activity: Classify Eight Systems

Apply TRACE independently to eight made-up systems, then compare your reasoning with the model answers.

What to do

For each case, use TRACE.

1. Task: What work does the system do?
2. Result: What does it produce or change?
3. Actual method: Which rules, models, search tools, or actions are proven?
4. Controls: Is there approval, review, limited access, logging, or a stop rule?
5. Evidence gap: What important fact is missing?

Then write a short description of the system.

You may use more than one term:

• Fixed-rule automation.
• Calculation.
• Retrieval or search.
• Machine learning.
• Deep learning.
• Generative AI.
• Tool-using system.
• AI agent.
• A system with several combined parts.
• Human-reviewed system.
• Not enough information.

All cases are made up. Do not use real private or workplace information.

Case 1: Rate calculator

A web form applies published rates to the numbers a user enters. Developers change the rules when the rates change. The same inputs give the same result.

Case 2: Spam service

An email service uses a model trained on labelled messages. It gives each new message a spam score. A fixed rule blocks messages from known harmful domains before the model checks them.

Case 3: Plant finder

A phone app uses a many-layer neural network trained on labelled photos. For a new photo, it returns three possible plant names and confidence scores.

Case 4: Poster maker

A tool receives a written description and creates a new poster image with a trained generative model. No search or publishing feature is described.

Case 5: Policy Guide

A service finds passages in approved policies. A language model uses the passages to write a draft. A policy officer must approve replies about exceptions before they are sent.

Case 6: Meeting arranger

A service receives this goal: Arrange a 30-minute project review next week. It checks calendars that the user has allowed it to access. It suggests a time and waits for approval. It then sends invitations, checks replies, and suggests a new time if an important person says no.

Case 7: SmartRead Agent

A company says its SmartRead Agent extracts text from scanned forms. It gives no details about training, tools, actions, testing, or human review.

Case 8: Revenue forecast

A spreadsheet copies the latest monthly amount into the next three months. Its dashboard calls the result an AI revenue forecast.

When your work meets the standard

Your work should:

• Keep the result separate from the method.
• Give a clear description for at least seven cases.
• Say that Case 7 does not have enough information.
• Find systems that use more than one part.
• Avoid using a product name as proof.
• Notice that the meeting arranger waits for approval before sending invitations.
• Name one missing fact for at least six cases.

Think back

Choose the case that was hardest for you.

Write two short answers:

• Which label did you almost choose too quickly?
• Which fact helped you choose a better answer?