Design direct measures that reveal student judgment when AI is available.
The Sprint helps a program decide what it needs to infer from student work, redesign one priority measure, and build the process evidence required to interpret the result with confidence. The engagement ends with a pilot-ready assessment and a clear improvement cycle.
A polished submission no longer tells the whole learning story.
AI can help produce a credible memo, presentation, recommendation, or analysis. The final artifact may therefore reveal less about who framed the problem, which claims were verified, and whether the student can explain or adapt the judgment. Programs need enough process evidence to interpret the work without turning every assignment into surveillance.
Output without ownership
- Who framed the problem
- Which claims were accepted or rejected
- What evidence was checked
- How the recommendation would change under pressure
Judgment that can be assessed
- The role AI played
- The workflow and decision points
- Verification and revision behavior
- The student’s ability to explain, adapt, and defend
Four practices make AI-assisted reasoning visible enough to assess.
The Sprint uses the disclose, document, verify, and defend framework to build professional accountability into the assignment. The practices are selected according to the learning purpose and the stakes of the measure.
Disclose
Explain the tool’s role, purpose, influence, and final accountability.
Document
Preserve enough of the workflow to examine how the work developed.
Verify
Check claims, sources, logic, calculations, assumptions, and fit.
Defend
Explain, adapt, or stand behind the final professional judgment.
Connect the competency, task, process evidence, rubric, and improvement decision.
The work begins with the inference the program needs to make about student capability. The assessment condition, AI permissions, evidence requirements, rubric, scoring process, and reassessment plan are then designed around that inference.
Move one priority measure from concern to a pilot-ready design.
The engagement can focus on a single program competency, one direct measure, or a small set of related assessments. A bounded scope keeps the work practical and gives the program a complete evidence cycle it can test and refine.
Clarify the inference
Define what the program needs to know about student capability and what evidence would support that conclusion.
Set the AI condition
Decide whether AI is prohibited, limited, permitted, required, or compared across conditions based on the learning purpose.
Redesign the measure
Build artifact, process, verification, and defense evidence into the task and rubric.
Pilot and close the loop
Set the scoring, calibration, finding, decision, improvement, and reassessment process.
Leave with the pieces needed to run and interpret the pilot.
The package connects assessment design with scoring, findings, improvement, and reassessment so the program can use the evidence rather than merely collect it.
- AI-relevant competency statement or dimension aligned with the program.
- Competency-to-direct-measure map.
- Redesigned assessment valid under the chosen AI condition.
- Process-evidence architecture using disclose, document, verify, and defend.
- Direct-measure rubric and scorer calibration guidance.
- Pilot, data-collection, finding, improvement, and reassessment plan.
What assessment leaders usually ask.
Does every assessment need to prohibit AI?
No. The AI condition follows the learning purpose. Some measures may be AI-free; others may permit or require AI with transparent process evidence.
Does the program need a separately named AI competency?
No. The Sprint can work with AI-relevant dimensions embedded in existing program competencies.
Can this fit the existing assurance process?
Yes. The goal is usually to strengthen the validity and interpretability of the current process rather than create a parallel system.
Is software required?
No. The design is platform-independent. Technology may support collection later, but the evidence model must stand on its own.
Start with one competency and one complete evidence cycle.
The first sprint can establish a model the program adapts to other measures over time.