How workforce capability leaders can prioritize AI-enabled work

Build a work-to-capability portfolio

Use a four-column portfolio: business work system, consequential task, observable capability, and evidence owner. The NIST AI Risk Management Framework Core is useful because it separates governance, mapping, measurement, and management rather than treating AI adoption as a single training event. The ONET Content Model offers a second anchor for distinguishing tasks, work activities, skills, and knowledge.

Begin with three to five work systems where AI use is already material or likely soon. For each one, write the decision an employee must make and the artifact that makes the decision inspectable. “Use AI better” is not a capability. “Compare three supplier proposals, trace every material claim to source evidence, and document unresolved assumptions” is observable work.

• Work system: supplier selection
• Consequential task: compare proposals under changing constraints
• Observable capability: frame criteria, inspect evidence, use AI to organize analysis, reconcile contradictions, and communicate a recommendation
• Evidence owner: procurement operations owns the task standard; the capability leader owns portfolio visibility; a business reviewer judges the artifact

Sources: [1] [2]

Score opportunity before choosing an intervention

Give every work system an opportunity card. Rate frequency, consequence of error, current friction, evidence availability, and change readiness on simple ordinal scales. Keep the raw observations beside the rating so a high score cannot hide weak evidence. A frequent low-consequence task may suit lightweight enablement. An infrequent high-consequence task may require a supervised simulation and stronger verification practice.

A synthetic service-operations card might record 120 monthly case summaries, repeated source hunting, inconsistent escalation notes, and a manager review already in place. That evidence supports exploring a task-specific intervention. It does not establish that a particular tool or course will improve outcomes.

The portfolio meeting should decide one of four dispositions: observe, redesign the workflow, provide guided practice, or assess performance. This prevents every demand from automatically becoming a training request.

Define the evidence before commissioning content

For each selected task, write an evidence contract. It should name the scenario, inputs, expected artifact, observable process signals, reviewer, and review cadence. The contract should also state what is outside scope. The NIST Privacy Workforce Taxonomy illustrates why responsibility language matters: roles can be owners, performers, collaborators, or reviewers without all controlling the same budget or decision.

For the service-operations example, the contract could require a participant to review six synthetic case records, identify conflicting facts, draft an escalation summary, cite the decisive evidence, and note what remains unknown. Evidence might include documents opened, revisions made, sources saved, AI interactions, and the submitted summary. The reviewer evaluates demonstrated work against a task rubric; the exercise is not a claim about future workplace outcomes.

Sources: [3]

Run a quarterly allocation review

At each review, bring the opportunity cards, completed evidence contracts, observed failure patterns, and intervention costs. Ask whether the original task remains important, whether evidence is representative, whether employees had a fair chance to demonstrate the capability, and whether the intervention should expand, change, or stop.

Use a compact decision record:

• Continue: supplier-comparison simulation because contradictory-source handling remains a material need and the scenario captures that behavior.
• Revise: service-summary practice because the inputs omit the policy exception that drives the real decision.
• Stop: generic prompt seminar because no priority work system or inspectable artifact is attached to it.
• Observe: forecasting workflow until data access and ownership are settled.

Keep the leadership boundary explicit

The workforce capability leader owns the enterprise view: which work systems deserve attention, how investment is balanced, and what evidence is credible enough for the next decision. Function leaders own task truth. Learning teams may design practice. Risk and privacy specialists review boundaries. Managers support application in context.

Two Prune can support realistic AI-enabled work simulations and evidence-backed assessment of observed performance. It does not decide workforce actions, prove business impact from a single exercise, or replace local judgment. The portfolio method is useful when it makes those ownership boundaries and evidence limits visible before money and attention are committed.