Perspectives

AI governance is converging around several published anchors: the NIST AI Risk Management Framework, the emerging ISO/IEC 42001 standard, federal direction on safe and trustworthy AI, and sector-specific guidance. These artifacts give organizations principles, but they do not give them operating discipline.

The gap most organizations face is not policy — it is decision rights. Who can approve a model going into a critical system? What is the rollback authority when model drift is detected mid-operation? Which workforce roles need to be defined before a capability transitions from pilot to production? Governance frameworks that survive contact with operations specify these answers; frameworks that don't, become shelfware.

AI-enabled operational systems are reaching frontline operators faster than the operator-centric design practices required to use them under real constraints. The risk is well-documented in human factors research: automation operators do not trust gets bypassed, automation operators trust too much gets misused, and the band of correct calibration is narrower than most leadership teams assume.

A practitioner approach treats trust calibration as a design requirement, not a training residual. That means presenting confidence intervals operators can act on, building explainability into operational tempo (not after-action reviews), and structuring fallback to non-AI workflows so operators retain decision-making muscle when models degrade or are unavailable.

Most published AI training curricula for public-sector and mission-critical workforces are organized around tools — what an LLM is, how a model is fine-tuned, what fairness metrics measure. This is necessary but insufficient. Operational workforces need AI literacy organized around the work itself: how to evaluate model output under time pressure, how to recognize degraded performance, how to escalate when output conflicts with mission intent.

Curriculum design grounded in adult-learning research and applied technical-training practice produces a different artifact than tool-centric training: shorter modules, scenario-driven assessment, and embedded sustainment touchpoints rather than one-shot certification. The doctoral research in strategic management behind these perspectives examines how to construct, deliver, and measure that kind of curriculum at scale across large, distributed organizations.