Skip to lesson content
Free course

AI Foundations / Module 7 / M07-U02 · 15-20 minutes

M07-U02 · 15-20 minutes

Classify Inputs and Justify Tools

Work through the explanation, apply it to the example, and complete the quick check before continuing.

Input and tool decision map showing an input register, task parts matched to tools or authorised people, rejected input conditions, and autonomy choices.
Classify every input by origin, sensitivity, permission, and minimum necessary version; then match each task part to the safest suitable tool or authorised person.
Text description of the visual

The figure shows an input register with four fields for every source: origin, classification, permission, and minimum necessary version. It warns learners to reject input when permission or classification is unclear, or when the input contains more detail than the task needs. It maps task parts to suitable choices: search uses current authoritative sources, extraction uses approved input, exact calculation uses a repeatable tool or spreadsheet, generation may draft text with review, analysis needs method and evidence logging, and formal approval belongs to an authorised person. It also reminds learners to state which tools were rejected and why, and to choose between single response, assisted steps, and agentic process. Agentic work is usually unnecessary for this capstone and requires stronger permissions, confirmation, monitoring, logs, recovery, and stop controls.

Complete the input register. For every source, record origin, classification, permission, and minimum necessary version.

Then separate task parts: search, extraction, calculation, generation, analysis, and approval. Match each part to the best tool. Current claims need current authoritative sources; exact calculations need repeatable tools; formal decisions need authorised people.

State which tools you rejected and why. Choose single response, assisted steps, or agentic process. Agentic work requires a clear benefit and stronger permissions, confirmation, monitoring, logs, recovery, and stop controls. It is usually unnecessary for this capstone.