AI Learning Lab · supplied records

The lifecycle of a school AI service

Year 9 · Lesson 1 · Content 2026.11.02.1

All supplied-record lessons · Open interactive lesson

Supplied-record alternative. Enter your responses in your school assignment. These supplied results support learning but do not show that you ran an algorithm.

Your goal

Find a service failure and connect it to a stage, a person and an action.

To be able to explain how lifecycle stages and named human roles help manage failure in an AI service.

Before you start

A calculator gives the correct number but sends it to the wrong classroom. Can the service still fail?

Choices: Yes / No

Bridge feedback after trying

A working model is only one part of a complete service.

Words for this lesson

Lifecycle
The connected stages of making, using and reviewing a system.
Deployment
Putting a tested service into its intended use.
Monitoring
Checking a service while it is being used.

The idea

A school service needs more than a model. Data must be prepared, a model trained and its results evaluated. Deployment connects it to real users and systems. Monitoring checks new results. Intervention means taking action when evidence shows a problem. The stages can loop. A useful plan names who checks, what evidence they use and what they do next.

A completed example

Practice: a menu assistant copies an outdated approved menu correctly. The failure concerns the source-update process, not necessarily the language model. The source owner should check the menu version and correct it.

Check your understanding

A high model score proves the whole service works?

  1. No
  2. Yes
  3. Only if it is 94%
Practice feedback after trying

No. Interfaces, sources, people and failure routes need checks too.

Deployment is ...

  1. Putting the tested service into intended use
  2. Only training
  3. Reusing test answers
Practice feedback after trying

Putting the tested service into intended use. It connects the service to its use setting.

A correction log with no reviewer lacks ...

  1. A monitoring owner
  2. More colours
  3. A longer model name
Practice feedback after trying

A monitoring owner. Evidence needs somebody responsible for reviewing and acting.

Task 1 · Make a prediction

Could a good model still be part of a failing service

Options: Yes / No / Not sure

Write your prediction, or state that you skipped it.

Try it using supplied records

  1. Follow one fictional request through Data, Training, Evaluation, Deployment, Monitoring and Intervention.
  2. At each stage, choose the person responsible from the supplied roles and inspect the evidence.
  3. Activate one scripted fault: old category codes, unreviewed corrections or no pause trigger.
  4. Predict the result, then reveal the service consequence. Changing the fault does not change the model's supplied test score.

Scripted service workflow · not a deployed service

Complete data and source records

Fictional service stages. Fictional teaching data.
StageExisting arrangementRole
DataAnonymous request text with approved category labelsCatalogue manager
TrainingClassifier version recordedDeveloper
Evaluation94/100 separate quiet typed English requests routed correctlyTesting lead
DeploymentConnected to queue; category code mapping checked only onceIT lead
MonitoringStaff can correct routes but nobody reviews correctionsLibrarian
InterventionManual routing exists but no written trigger to switchLibrary lead
Read as stacked records
Stage
Data
Existing arrangement
Anonymous request text with approved category labels
Role
Catalogue manager
Stage
Training
Existing arrangement
Classifier version recorded
Role
Developer
Stage
Evaluation
Existing arrangement
94/100 separate quiet typed English requests routed correctly
Role
Testing lead
Stage
Deployment
Existing arrangement
Connected to queue; category code mapping checked only once
Role
IT lead
Stage
Monitoring
Existing arrangement
Staff can correct routes but nobody reviews corrections
Role
Librarian
Stage
Intervention
Existing arrangement
Manual routing exists but no written trigger to switch
Role
Library lead
Fault consequences. Fictional teaching data.
FaultSupplied outcome
Old code mappingA correctly predicted category is sent to the wrong queue after a category-code update.
Corrections not reviewedRepeated routing errors remain unnoticed in a correction log.
No triggerStaff have no agreed instruction for when to use manual routing.
Read as stacked records
Fault
Old code mapping
Supplied outcome
A correctly predicted category is sent to the wrong queue after a category-code update.
Fault
Corrections not reviewed
Supplied outcome
Repeated routing errors remain unnoticed in a correction log.
Fault
No trigger
Supplied outcome
Staff have no agreed instruction for when to use manual routing.

Expected results · supplied

These are supplied reference results. They are separate from your own explanation. No code needs to run to read them.

1. This is a scripted service workflow, not a trained or deployed library system. Correct role matching follows the stage table. Use ordered lists or step buttons, never drag-only interactions.

2. Fault selector values mapping, review, trigger show the supplied consequences. All retain the 94/100 score label and explain that its narrow evaluation excludes speech and other untested conditions.

3. A monitoring-plan form stores owner, signal, frequency, trigger, response and reason as separate fields. An example trigger may be more than 5 wrong routes in 50 manually checked weekly cases. Label it a classroom proposal, not an operational safety standard.

Help with this step

Hint 1

Which stage contains the gap in the brief?

Hint 2

Choose a person who can act on that gap. Name evidence, not just check it.

Hint 3

Practice: an IT owner checks undelivered messages every day and uses a manual callback route if delivery fails.

Task 2 · Use the evidence

  1. For the three gaps, record the stage, a named role, evidence to check and an action.
  2. Choose the most urgent gap and justify your priority.
  3. Write a monitoring sentence that includes owner, signal, review frequency and response. Other priorities are acceptable when justified.

My explanation

Write in your school assignment. Keep these answers in your own words.

Six stages

Old codes

Unreviewed corrections

No trigger

Your priority

Monitoring plan

Explain a fresh example

A library model passed its test, but category-code updates send requests to the wrong queue. Name a responsible role, evidence and action, and explain why the test was insufficient.

My confidence

I need another explanation / I can explain with help / I can explain with my own example

My next step

Write what you will do next.

Optional depth

How could a data error first become visible in a monitoring log? Explain the difference between where a fault started and where it was detected.

What this activity can show

Following a workflow is not evidence of building a complete AI-powered solution. The classroom trigger is invented and requires real operational validation before any use.

This is a student learning record. Submission is completed through the school assignment system.