1
Define the decision
Name the user, decision, current workflow, intended benefit, non-goals, and a non-AI alternative.
Required evidence: A one-page problem contract with measurable success and a defensible baseline.
0/80 minimum characters
Foundation project·3-4 hours
AI Decision Brief Capstone
Evaluate one proposed AI feature and make a launch, narrow, revise, or stop recommendation grounded in evidence, risk, and operating controls.
Scenario
A service team proposes an AI assistant that summarizes customer cases and recommends the next action. Leaders need a decision brief that separates useful capability from fluent demos and sets a bounded evidence plan.
You will demonstrate
- Frame an AI opportunity without starting from a model name.
- Evaluate capability, grounding, safety, and workflow evidence.
- Defend a bounded decision with human authority, monitoring, and rollback.
Project evidence
Show the work, not a checked box.
Each response is stored in this browser as you type. Include metrics, test output, or a decision rationale wherever the deliverable asks for it.
2
Design the evidence plan
Create representative successful, ambiguous, unanswerable, and high-cost failure cases. Define offline and pilot metrics.
Required evidence: A test matrix with expected behavior, metric definitions, and acceptance thresholds.
0/80 minimum characters
3
Map material risks
Trace privacy, unsupported claims, unequal performance, automation bias, and operational failures to controls and owners.
Required evidence: A risk-control-evidence table with severity, detection, recourse, and stop conditions.
0/80 minimum characters
4
Compare the operating choices
Estimate quality, latency, review effort, and model cost for at least two designs and the non-AI baseline.
Required evidence: A compact decision matrix with assumptions and uncertainty ranges.
0/80 minimum characters
5
Write the decision brief
Recommend launch, narrow, revise, or stop. State what is known, unknown, reversible, and required before wider exposure.
Required evidence: An executive brief tied to the test matrix, risk register, and rollback trigger.
0/80 minimum characters
Runnable Python lab
Calculate a transparent decision matrix
Implement weighted evidence scoring while preserving a hard safety gate that a high average cannot hide.
Project defense
A prototype is fluent on five hand-picked demos but has no unanswerable cases or baseline. What is the defensible recommendation?
Rubric self-review
Rate the evidence, not your effort: 0 missing, 1 weak, 2 adequate, 3 strong. All criteria must be reviewed, but a low honest score does not get hidden.
The problem contract begins with a user decision and includes a non-AI baseline.
Evaluation cases represent normal, ambiguous, unanswerable, and high-cost failures.
Material risks map to testable controls, accountable owners, and recourse.
Cost and latency assumptions are explicit enough to challenge.
The recommendation remains bounded by evidence and includes a stop condition.
Artifact
Attach the portfolio evidence.
Submit the decision brief as Markdown or PDF, or attach a notebook containing the decision matrix. Files are validated but never executed.
Anonymous mode stores only metadata and a fingerprint locally. Private upload requires an account and an explicit file selection. Uploaded notebooks are never executed.
Useful references
Project completion gate
Completion is controlled by stored evidence, deterministic tests, a decision defense, rubric review, and the artifact when required.
Evidence pendingCode pendingDefense pendingRubric pendingArtifact pending
Optional cloud portfolio
Submit evidence across devices.
An account is required. Submit only when the local completion gate passes. AI review is advisory and separate from deterministic completion.