7 Steps CHROs Can Take To Guide Organizations’ AI Strategy

–:– / –:–This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Dr. Timothy J. Giardino is the Founder of myWorkforceAgents.ai.

gettyIf I became CHRO tomorrow, I would ask the executive team to take a step back and consider a consequential strategy question: “Now that people, digital labor and conventional systems can be designed together, what should this organization become?”

HR leaders are becoming design architects, capability stewards, adoption catalysts and transition guardians in AI transformation. Microsoft’s “2026 Work Trend Index” report found that organizational factors like culture, manager support and talent practices accounted for 67% of reported AI impact. I believe the CHRO’s role is to make sure the enterprise has a deliberate plan for what happens to work, jobs, management, learning, trust and human accountability when AI begins performing work.

Here’s the seven-step process I’d suggest any CHRO take to accomplish this goal.

First, help the executive team define where AI belongs, where human judgment must remain and what evidence is required before autonomy or workforce changes scale. The output would be an executive transition compact that covers decision rights, named owners, escalation routes, challenge authority, workforce principles, prohibited uses and the intended operating model.

Then, identify and prioritize the most material hybrid intelligence work systems. That includes the people, digital labor, software, data, vendors, handoffs, exceptions and accountable decision makers producing consequential, trusted outcomes.​

HR should be the first department to test AI and validate the model. I’d recommend selecting one bounded HR workflow, such as employee policy resolution, recruiting intake or leave administration, and redesigning it as a governed human-and-digital work system. Map how the work actually happens, define human and AI roles, test exceptions and measure hidden work. Be sure to protect the employee experience and prove that the workflow can continue if the digital layer fails.​

HR and operating leaders should follow an outcome end to end. First, remove unnecessary steps, unclear policy and avoidable handoffs. Next, classify the remaining work as automation, AI augmentation, judgment or relationship. Then, select the least complex technology capable of meeting the requirement.

The rule of thumb here isn’t that every workflow needs AI, but every AI system needs a defined workflow. This work should produce current-state, transitional and intended-state maps. It should also preserve the crucial distinction between AI access and AI adoption, as well as the difference between work redesign, workforce redesign and organization redesign.​

I’ve written about how each digital worker should be onboarded. That means establishing its defined identity and purpose, human owner, approved sources, data and tool permissions, authority limits, escalation rules, performance expectations and retirement criteria. Material systems also need an approved operating envelope describing when the AI may continue working, plus runtime assurance to confirm those conditions remain true.

For example, the NIST calls for differentiated human-AI responsibilities and post-deployment monitoring that includes user feedback, appeal, override, incident response, recovery and change management. Similarly, the ISO/IEC 42001 treats AI governance as a management system that must be maintained and continually improved.

For externally supplied digital labor, I’d require a clear provider-client responsibility model. ​

Every major HR discipline should review what the new work requires:

• Talent acquisition should recruit for redesigned roles.

• Learning and development should protect human judgment-building experiences.

• Talent management should update skills, careers and succession.

• Total rewards should examine new responsibilities and hidden supervision work.

• Employee relations should create correction and challenge pathways.

• People analytics should measure usable capacity, rework, trust, fairness and manager burden, while also tracking differences in AI enablement across comparable roles.​

Managers will need to lead people, supervise digital labor and orchestrate the workflow between them. People will need context, coaching, development, inclusion and psychological safety. Meanwhile, the digital labor will need approved sources, permissions, monitoring, calibration and bounded authority. The work system itself will need strong handoffs, exception capacity and recovery.

Provide managers with a defined operating rhythm for readiness, exceptions, incidents, hidden work and weekly system review. Review the hybrid intelligence span of control based on the volume and complexity of exceptions, approvals, corrections, handoffs and recovery events a manager can absorb from both human and digital labor.

A three-firm OECD experiment found that participants who used consultation-produced designs could preserve productivity gains while improving job quality. So make sure employees participate in redesigning work they understand best and receive fair access to approved tools and training. Additionally, review AI form factors, including voices, avatars, accents and personas for accessibility, cultural fit, diversity, representation and psychological safety.

Note: Shadow AI may occur, and that scenario should be surfaced without punishment. Then, route it through a lightweight review that protects useful experimentation while assessing data exposure, bias, security, legal risk and workflow impact.​

Every material work system requires evidence on business outcomes, rework, hidden work, handoff quality, exception volume, manager capacity, trust and fairness, continuity and a human capability reserve to ensure business continuity. Corrections and incidents should feed redesign, not end with transaction repair. Don’t permanently remove human capacity until the workflow has proven it could continue, recover and return safely to human control.

Each review should end with a decision: continue, expand, recalibrate, restrict, pause, redesign or retire. The next step should follow the intended operating design, not the simple availability of more automation.​

HR has spent years asking to enter transformation earlier. Agentic AI creates that opportunity. But the seat only matters if HR arrives with an operating plan for integrating human and digital labor without surrendering accountability for the organization that follows.​

Forbes Human Resources Council is an invitation-only organization for HR executives across all industries. Do I qualify?