Defend an enterprise AI capstone before a review panel

A capstone defence should test whether the learner can connect business scope, engineering, evaluation, security, operations and leadership under challenge. A polished demo is only one piece of that evidence.

Build the decision artifact

Provide the panel with the brief, architecture and data boundaries, runnable build, evaluation and calibration report, threat model, SLO and cost evidence, incident and recovery drills, adoption plan and limitations. Score dimensions independently and add hard gates for reproducibility, critical security failures and evidence integrity. Give the learner time to respond, then require targeted rework.

The commercial leadership lab makes the artifact inspectable with authored inputs:

from leadership_cases import capstone_defence_case

result = capstone_defence_case()
assert result["score"] == 79
assert result["maximum"] == 100
assert result["gates"]["critical_security_failure"] is True
assert result["pass"] is False

The fictional submission scores 79 of 100, but a critical security failure keeps pass false. No panel actually met and no certificate is awarded. The gate shows why aggregate strength cannot compensate for a material exposure.

Protect the decision from weak evidence

A rubric can reward presentation twice and evidence once. Anchor each score to artifacts and questions. Separate a recoverable weakness from a gate failure, document panel disagreement and re-run the failed control rather than merely revising slides.

Keep these artifacts for review:

• versioned capstone evidence index
• dimension scores with artifact references
• adversarial questions and responses
• gate decision, rework and retest record

This practice aligns with the discovery, productisation, client enablement, technical leadership and capstone sequence in the FDE for Professionals course. The course link describes the pathway; the local scenario is fictional and does not claim a client engagement, investment result, hiring decision or certificate.

Practice task

Run a sixty-minute mock panel with architecture, security, operations and sponsor roles. Require each reviewer to cite evidence for every score.

Continue learning

This article is part of the Commercial judgement and delivery leadership sequence. Use the neighbouring tasks when you need the prerequisite or the next application.

• Review the preceding task in Build a professional FDE portfolio from anonymized evidence.
• Return to the cluster foundation in Write an AI statement of work with measurable acceptance.

Reference: GOV.UK Service Manual: Service Assessments.