| Phase | Min | What happens |
|---|---|---|
| Rules and pitch | 10 | Teacher reads the pitch aloud once, like a sales call. Then read it again silently. |
| The audit | 40 | Parts A to E, on paper. No devices. |
| Proofread | 7 | Check that every claim is sorted and every guess is marked as a guess. |
| Hand in | 3 | Name and date on every page. Keep your index card. |
If you need extra time, a reader or another format, talk to your teacher before the check day.
| Evidence level | What it means |
|---|---|
| 1 | Nothing to check: an opinion, a boast, or a promise with no details. |
| 2 | Something to check, but details are missing: a number or a source with no who, how many, compared to what, or when. |
| 3 | Checkable now: specific enough that you could test it, or get it in writing, before you sign. |
| Task level | Who does the work |
|---|---|
| 1 | The tool does it |
| 2 | The tool drafts, a person reviews each one |
| 3 | The tool assists, a person decides |
| 4 | A person does it |
Stop signs: a mistake that's hard to undo means level 2 at least. No one checking means level 3 at least. A claim you can check deserves credit. Don't doubt everything the same.
| Part | Points | What you do | |
|---|---|---|---|
| A | Sort the claims | 25 | Mark how much evidence each claim in the pitch comes with, and say what you'd need to see. |
| B | Sort the tasks | 25 | Give each task the tool would do a level (1 to 4) and a reason, using the four questions. |
| C | Questions for the vendor | 15 | Write three questions that could change your mind, and pick the one to ask first. |
| D | Checks and failure | 15 | For one task you would let the tool do or help with: who checks, when, how, and what happens when it fails or has to be switched off. |
| E | Defend it | 20 | Your recommendation, two reasons from the pitch, the strongest objection and a reply, and what you don't know. |
| Part | Points | Exemplary | Proficient | Developing | Incomplete |
|---|---|---|---|---|---|
| A. Sort the claims | 25 | Every claim has an evidence level that fits the definitions, and each "what I'd need" names a specific missing detail. The checkable claim is credited, not doubted along with the rest. | Most levels fit, and most "what I'd need" entries are specific. One or two are vague ("proof") or the checkable claim is doubted. | Several levels are off, or entries say only "it's not proven." | Fewer than half the claims are sorted. |
| B. Sort the tasks | 25 | Every task has a level with a reason from the four questions (routine, cost, trust, who checks), using facts about this client. The stop signs show up where they apply, and at least one task stays with a person. | Every task has a level and a reason, but a reason is thin or skips the client's facts, or a stop sign is missed once. | Levels without reasons, or reasons like "it's easier." | Tasks aren't sorted. |
| C. Questions for the vendor | 15 | Three questions, each specific, answerable, and aimed at evidence, failures, or data. The first-to-ask choice is explained in a sentence. | Three questions, mostly specific. The first-to-ask choice has no reason, or one question is too broad. | Questions are yes/no, vague, or about price and features only. | Fewer than two questions. |
| D. Checks and failure | 15 | The check says who looks, when and how, and fits this client's time. The failure plan names one failure that would look right but be wrong, how it's noticed, and who switches the tool off and how. | A check and a failure plan are present, but one is vague ("someone will look") or the quiet failure is missing. | One piece is missing, or the plan depends on the tool noticing its own mistakes. | No checking or failure plan. |
| E. Defend it | 20 | The recommendation fits the task levels. Two reasons use facts from the pitch. The strongest objection is fair and the reply answers it. "What I don't know" names something real that a trial or a question could settle. | Three of the four are solid, and one is thin. | A recommendation with one reason, no objection, or an "I don't know" that is filler. | No recommendation. |
| You're scored on (0 to 5 each) | 5 looks like | 1 looks like |
|---|---|---|
| Recommendation fits the tasks | Matches the levels they gave and could be explained in a sentence. | Doesn't connect to the task levels. |
| Reasons use facts from the pitch | Points to a specific claim and what it did or didn't show. | General statements about AI. |
| Strongest objection, and a reply | States a fair objection to their own view and answers it. | Doesn't see an objection, or dismisses it. |
| What I don't know | Names something real and says how to find out. | "Nothing" or a vague worry. |
A practice pitch, a worked example, and the five parts, to use at home.
🗓️ Final Check · Unit 3 of 3 · 11 Weeks