GAO's training guide needs an AI-to-job-performance evidence chain

Write the performance chain before selecting content

The direct answer is a one-page evidence chain for each funded AI learning effort: business or service priority, affected role, real task, current performance condition, required knowledge and judgment, learning activity, supervised practice, observable workplace behavior, and bounded individual and organizational measure. Name the employee, manager, learning owner, data owner, and decision that will use the result. A generic objective such as 'increase AI fluency' cannot tell a CHRO whether a course fits a recruiter, finance analyst, service representative, engineer, or manager. Tool usage, prompts written, hours attended, completion, satisfaction, and confidence can describe participation, but none alone establishes that work became more accurate, safe, timely, accessible, or useful.

Base the need on work and worker evidence

Use job and task records, quality and service evidence, manager observation, employee input, accessibility needs, incident patterns, customer or stakeholder feedback, and planned workflow changes to identify the gap. Separate a knowledge gap from missing system access, poor data, unclear policy, workload, weak supervision, inaccessible tools, or a process that should not use AI. Include employees and affected groups early enough to change the design, not only to react after rollout. Define prerequisites and different starting points, provide alternatives for people who cannot use the selected modality, and protect learning records from unrelated employment decisions. A vendor's skills taxonomy or generated assessment can inform the inquiry but should not become a person-level capability verdict without validation.

Design evaluation before the first cohort

Choose representative tasks, an approved environment, baseline evidence, scoring criteria, assessor, timing, and follow-up period before delivery. Evaluate whether participants can recognize when AI is appropriate, protect data, inspect sources, challenge outputs, follow role policy, complete the task, and hand off or stop when needed. Where practical, compare with the prior workflow or a credible alternative and record concurrent changes such as new data, staffing, incentives, or process redesign. Do not require employees to expose protected or personal information in practice. Use aggregate program analysis for investment decisions while keeping individual results bounded to the stated development purpose and providing a correction route for inaccurate records.

Use findings to revise the work system

Review reach, completion, task performance, error and escalation patterns, accessibility, employee experience, manager support, transfer to work, service or quality measures, rework, incidents, and cost together. Investigate uneven outcomes by role, location, cohort, starting skill, and access without treating correlation as causation or using small groups in ways that expose people. If participants understand the material but cannot apply it, fix permissions, data, policy, workflow, supervision, time, or tool design before buying more content. If a task changes, expire the old assessment and update the learning job. Report what the evidence supports, what remains uncertain, and which next decision is authorized rather than converting a dashboard into a claim that the workforce is AI-ready.

Turn this source into a reviewable decision

For AI for CHROs, use this briefing as a dated decision record rather than a substitute for the source. Preserve Human Capital: A Guide for Developing and Assessing Strategic Training and Development Efforts | U.S. GAO, the exact URL, the August 26, 2026 review date, the supported facts above, the editorial interpretation, the limitations, and any buyer-specific evidence. Link that record to the decisions most directly affected: Learning and capability development; Skills intelligence and internal mobility; Workforce and capacity planning; People analytics and employee listening. State whether the source changes the scope, evidence requirement, control, sequence, or only the language used to describe the decision.

Before action, name the accountable owner, affected population and workflow, exact offering or configuration, source data and rights, human decision point, exception and appeal path, complete cost, expected benefit, failure and stop conditions, retained evidence, and next review date. Keep official facts, provider statements, buyer observations, representative tests, measured outcomes, editorial inferences, and unknowns visibly separate. Reopen the record when the source, offer, model, integration, data, policy, population, responsible person, or measured result changes.

Limitations and unknowns

GAO is the government source. Its July 13, 2026 guide addresses strategic training and development in the federal government, identifies process components and characteristics, and notes that digital learning and AI are shaping design, delivery, and evaluation. It is not an employment law, AI-product evaluation, private-employer mandate, course endorsement, skills taxonomy, or proof that a particular learning activity improves performance. It does not establish an employer's roles, task gaps, accessibility, worker views, program quality, individual learning, transfer, causality, or outcome. Current job and task evidence, worker and manager input, learning and data-purpose records, representative assessments, workplace transfer evidence, and qualified people, learning, operations, accessibility, privacy, security, employee-relations, labor, and legal review control.