- LECTURE
- The Autonomy Dividend — 30 September 2026
- LENGTH
- 1,200 words ±10%, excluding cover block, tables and references
- DUE
- 14 October 2026, 23:59 AST
- FORMAT
- PDF only, produced from the answer template. Other structures are returned unmarked.
- FILE NAME
ACTAM_StudentID_LastName.pdf- SEND TO
- ali@alasiri.net — subject: ACTAM assignment — StudentID — LastName
What this asks of you
The lecture argued one proposition: an AI system’s governance obligation should follow its position on two axes — how much autonomy it holds, and how much consequence it carries. This assignment asks you to apply that proposition to a real system and defend your reasoning.
You are not asked to agree with the model. A submission that applies it carefully and then shows where it fails will score higher than one that applies it uncritically.
Learning outcomes
- Translate a governance framework into a repeatable assessment procedure and apply it to a real organisation.
- Score a system on defined dimensions and justify each score from evidence rather than impression.
- Convert an assessment into decision rights: who approves, who owns, what evidence is kept, and by when.
- Evaluate the limits of the framework you used, and cite primary sources for your claims.
Choose one brief
Pick the one closest to your track. All three are marked against the same rubric and use the same template.
Pick one AI system a real Saudi or international organisation has deployed. Score it on both axes with the five-step method, place it in a zone, name the accountable owner, and recommend three proportionate controls.
Build the business case for an AI governance programme at a named company: expected loss avoided, value delivered earlier, market access, minus programme cost. Source every input and give a sensitivity range.
A Saudi firm sells an AI product into the European Union. Map its obligations on both the EU and Saudi timelines, state what it must be able to evidence, by when, and who inside the company owns each item.
Every submission must contain
- The completed cover block: name, student ID, section, brief chosen, word count, date, and the signed declaration.
- The six-row assessment table — decision, autonomy, consequence, zone, owner, evidence. This is the spine of the work.
- A named accountable human and a review date. An assessment without an owner and a date is incomplete by definition.
- At least five sources, of which three are primary: the standard, the law, the company’s own disclosure, the regulator’s decision. Not this brief, not the lecture slides, not an AI chatbot.
- A section stating what the framework failed to surface about your system.
- The AI-use disclosure, completed honestly. Using an AI tool is permitted; not disclosing it is an integrity violation.
The method, in five steps
Reproduce this reasoning in your answer. Each sub-dimension is rated 1 to 5; the axis score is the average.
| Step | What you do |
|---|---|
| 1 · Describe the decision | What decision does the system touch, for whom, how often, at what value? |
| 2 · Score autonomy | Speed, scale, depth — recommends, decides or acts. |
| 3 · Score consequence | Severity, reversibility, breadth, exposure — from the affected person’s side too. |
| 4 · Place it, name the owner | Read the zone off the matrix. Name one accountable role and the escalation path. |
| 5 · Set the evidence | What proof, at what cadence, reviewed by whom, by what date. |
The four zones
Sandbox, logged, fast iteration.
Release gates, monitoring, a human escalation path.
Risk-committee sign-off, meaningful human review of each decision.
Board or ministerial approval — or do not deploy.
How your work is marked
| Weight | Criterion |
|---|---|
| 30% | Correct use of the method |
| 25% | Evidence and sourcing |
| 25% | Business judgement |
| 20% | Accountability |
Marks are for reasoning, not conclusions. Two students can score reversibility 2 and 4 and both score well if each argues it from evidence. Marks are lost for an unnamed owner, a missing review date, an unsourced figure, or a word count outside the range.