CHROs, this is your moment to co-architect the agentic enterprise

Most CHROs are being handed two mandates at once right now: cut costs and show AI progress. The response in most organizations is to reach for technology in the shape of a better portal, a new chatbot, or a refreshed shared services model. These moves can improve the surface, but they rarely touch what is underneath it. And when agentic AI lands on top of a fragmented operating model, it does not simplify the system. It scales the disorder.

The CHROs pulling ahead are approaching this differently. They are rebuilding how HR works before they automate it. IBM’s internal transformation is a clear example of what that looks like in practice and what it delivers when the sequence is right.

Adding AI to a fragmented HR model does not fix it; it scales the chaos faster

Most HR organizations still run on call center logic, even when they have layered digital tools on top. Employees come in through regional phone lines, shared inboxes, HR business partner relationships, and self-service portals that sit beside the human channels rather than replacing them. Each pathway develops its own routing logic, policy interpretation, and workarounds. Over time, the system becomes less of a designed service model and more of an accumulation.

When AI arrives in that environment, the outcome is predictable. Bots surface inconsistent answers because the knowledge beneath them is inconsistent. Employees route around the digital interface the moment it fails them. HR professionals remain tied up in the same coordination work they were already doing, now simply supported by more technology. The enterprise spends on AI, but little changes meaningfully.

The explanation often offered is that the technology needs more time; the models need more data, more tuning, or more maturity. But that interpretation mistakes the symptom for the cause. The models are not the primary constraint. The operating model underneath them is.

Automation applied before standardization does not improve a fragmented system. It accelerates it. The gap between how most HR functions operate today and what an AI-ready operating system actually requires runs deeper than most organizations have acknowledged (see Exhibit 1).

Exhibit 1: The shift that makes HR AI-ready: from tickets to orchestrated work

Source: HFS Research, 2026

IBM cut HR operating costs by 40% in four years; the chatbot did not do it, the operating model did

IBM’s transformation is built on a simple premise of rebuilding the delivery model AI will run on before deploying it at scale.

The company consolidated HR support behind AskHR, a single digital entry point for employee and manager needs. Once work was entering the system through one path, IBM could standardize workflows and govern knowledge consistently. Only then did IBM layer AI on top. Three outcomes followed:

• One front door created the conditions for containment. With all HR queries moving through a single entry point, IBM could standardize workflows and design escalation to specialists as a deliberate feature rather than a fallback. The reported containment rate sits at around 94%, meaning most employee and manager needs are resolved without escalation. That number reflects the operating model underneath it more than the technology itself.
• Containment made the cost reduction structural. IBM reports a roughly 40% reduction in HR operating costs over four years, driven by collapsing redundant channels, automating repeatable transactions, and simplifying a technology landscape that had grown too fragmented to manage. The savings did not come from deflecting tickets. They came from eliminating the conditions that generated unnecessary work in the first place.
• Reduced transaction volume changed the shape of HR work itself. As routine interactions moved into the digital tier, IBM reports that the average HR career band increased by roughly two levels, a practical indicator that capacity moved toward judgment-heavy work rather than being eliminated. IBM reorganized the remaining HR workforce into domain-focused pods with deep expertise in areas like performance and compensation that support needs across regions.

This structure also addressed a concern many CHROs raise: “What happens to entry-level development when transactional work disappears?” Entry-level employees rotate across the domain pods rather than entering a single administrative track, allowing them to build expertise across domains while still developing broad exposure to the function.

What made these outcomes possible was a deliberately tiered service architecture in which each layer handles a distinct class of needs (see Exhibit 2).

Exhibit 2: The digital-first HR architecture required to scale AI-driven service delivery

Source: HFS Research, 2026

Most enterprises are automating processes they have not standardized yet, and that’s the first mistake they need to fix

Organizations trying to scale AI in HR often start in the wrong place. They introduce automation into processes that were never standardized to begin with. The result is not transformation; it is simply a faster version of the same fragmentation.

IBM’s experience illustrates this challenge clearly, but the lesson extends well beyond one organization. AI does not repair a fragmented operating model. It amplifies it. If entry points are inconsistent, workflows vary by region, and escalation depends on informal human coordination, AI will scale those weaknesses rather than resolve them.

For CHROs, the more practical question is not how quickly AI can be deployed, but whether the underlying service architecture is designed to support it. Three priorities consistently emerge in organizations that are building HR models that can scale AI effectively:

1. Front door. Consolidate entry points into one and hold the line on it. This is often the hardest political step because regions and HRBPs are accustomed to direct access and will resist it. It is also the most important prerequisite. Multiple entry points mean multiple versions of the same process, and a digital layer that employees have already learned to bypass.
2. Flows. Identify the highest-volume HR journeys and rebuild them end-to-end before automating them. Manager transfers, employment verification, policy navigation, and performance cycle queries all need clear ownership, decision logic, and governance. Automating an unstandardized process simply accelerates inconsistency.
3. Forks. Treat escalation to human support as a designed feature rather than a system failure. Define clearly what requires human judgment and how employees transition from self-service to specialist support without losing context.

These operating model shifts reinforce one another, and organizations often progress across several at the same time.

HR is the only function where every employee is a forced adopter, making it the enterprise’s most powerful AI proving ground

HR is the only enterprise function that every employee engages with repeatedly throughout their working life: hiring, onboarding, performance management, compensation, and learning. When those interactions move to a digital-first model that works, employees develop new expectations. They begin measuring IT support, finance operations, and procurement against the experience standard HR has already set.

The CHRO who gets this right is not just running a more efficient HR function. They are conditioning the workforce for how the agentic enterprise will operate before other functions even begin the transition.

Claiming that position requires a genuine partnership with IT. CHROs own the operating model design, the workforce adoption strategy, and the business case. IT owns the infrastructure, integration capability, and technical governance. The enterprises making the most progress have both sides moving together. The ones stalling are waiting for the other to move first.

The Bottom Line: CHROs who rebuild the operating model before scaling the technology will not simply modernize HR. They will establish the blueprint the rest of the enterprise eventually follows.

HR is the only function every employee must engage with repeatedly throughout their working life. When it becomes the first function to operate on a digital-first, agentic model, it quietly resets expectations for how every other enterprise function should work.

Those who move early will shape that model. Those who wait may find themselves implementing a version someone else has already designed.