Your title still says "HR." Your job used to be policies, compliance, onboarding, and making sure people got paid on time. Now AI agents are making workforce decisions, and the policies that govern them don't exist yet. Nobody reassigned that problem. It landed on your desk because nobody else's desk is closer to the rules.
62% of organisations are deploying AI somewhere in their operations. 39% have already adopted it within HR itself. But only 49% of those organisations have any policy governing how AI is used in their workforce, according to SHRM's 2026 State of AI in HR report. Of the half that do have policies, only a quarter believe those policies are clear and future-proof. The rest are either too restrictive, too vague, or built for a generation of tools that's already been superseded.
That's the gap. Not whether AI will reshape work — that's settled. Whether anyone is writing the rules for how it does.
From compliance to design
HR has always been the compliance function — enforcing policies someone else wrote, ensuring labour law requirements land in practice, keeping the organisation within bounds set by legal, finance, and the board. The new problem is different. When an AI agent screens candidates, assigns shifts, flags performance patterns, or recommends who gets promoted, it's making decisions that used to require human judgment. The compliance surface isn't a form or a process anymore. It's an algorithm that nobody in HR was trained to audit.
67% of HR leaders say their biggest blocker is lack of awareness of what AI can actually do, per SHRM. And 57% of HR professionals working in states with active AI employment regulations aren't even aware those regulations exist. The knowledge gap isn't technical — it's structural. HR was designed to enforce rules. It was never designed to write the rules that govern autonomous systems.
But the regulatory landscape isn't waiting. Illinois now prohibits employers from using AI that produces discriminatory effects. New Jersey's automated decision technology regulations are already in force. Colorado repealed and replaced its original AI Act with SB 189 in May 2026, shifting from broad risk-based requirements to narrower high-risk-use obligations — proof that even the rules about AI rules are still being rewritten. These aren't future possibilities. They're current law, and someone in every affected organisation needs to know they exist, understand what they require, and build the internal policies that make compliance real — not just documented.
That someone is HR. Not because the title changed, but because nobody else sits at the intersection of employment law, workforce operations, and the daily reality of how people experience their jobs.
The governance question nobody assigned
Legal and compliance functions lead AI governance in 37% of organisations. But over half — 52% — don't involve HR in their AI strategy at all, according to SHRM. The people function is excluded from decisions about tools that directly affect people. That's not an oversight. It's the legacy of a model where HR administers policy and someone else designs it.
Josh Bersin's HR 2030 research frames the shift directly: 30–40% of existing HR jobs can be automated with relatively low effort. ADP's 2026 HR Trends Guide reports CHROs project 327% growth in AI agent adoption by 2027. The function that manages onboarding, talent acquisition, compensation, and workforce planning is simultaneously the function most likely to be transformed by agents and least likely to be setting the rules for how those agents operate.
The gap isn't "HR needs to learn about AI." The gap is that AI governance — the rules about what agents can decide, what they can't, what requires human review, and what happens when they get it wrong — is a design problem. And design problems need someone who understands both the constraints (employment law, fairness requirements, organisational culture) and the operational reality (who actually does the work, how decisions propagate, where the undocumented rules live).
From administrator to architect
The Bureau of Labor Statistics projects 5% growth for human resources managers through 2034. That number understates the shift because it measures headcount, not the nature of the work. The HR manager of 2020 ensured compliance with known regulations. The HR manager of 2027 designs the policies that govern how AI agents and humans work together — policies that don't exist yet in most organisations, for tools that are changing faster than any regulatory body can write rules about.
The people leader who writes better onboarding documentation than an AI agent is competing with the agent. The people leader who designs the workforce governance framework — which decisions require human judgment, which agents have access to which data, how bias is detected and corrected, what "fair" means when a machine is deciding — is doing the job the agent can't do for itself.
Sources: SHRM State of AI in HR 2026; Josh Bersin, "HR 2030: A Vision for Agentic Human Resources," April 2026; ADP 2026 HR Trends Guide; SHRM 2026 AI Policy Activation; Bureau of Labor Statistics Occupational Outlook Handbook.