THE NOOMA WAITLIST IS NOW OPEN. Founding Members lock in 25% off. Permanently.
.png)
AI is rapidly reshaping how work gets done, and HR sits directly in its path. Automation, copilots, and increasingly capable systems are already embedded in the tools HR practitioners use every day. Some of this progress is overdue. Some of it is necessary. And some of it, if adopted without discipline, fundamentally changes what HR is.
HR does not need to resist AI. But it does need to draw a line. The most important question facing HR is no longer whether to invest in AI, but where AI should stop. Without that clarity, the profession risks either being left behind or slowly hollowing itself out.
Much of the AI conversation is dominated by technical language: agentic systems, ambient AI, neurosymbolic models, AGI. For most HR practitioners, this language obscures more than it clarifies. The distinction that actually matters is far simpler.
There is AI that helps you do the work, and AI that does the work for you.
HR is not a single activity. It operates across distinct layers of risk, judgement, and consequence. Where AI belongs depends on which layer it is operating in. Treating all HR work as equivalent is how well-intentioned technology ends up producing poor outcomes.
At the data and administration layer, “do it for you” systems are not only appropriate, they are desirable. Managing employee records, tracking leave, triggering workflows, scheduling surveys, and producing reports should be automated. There is no professional value in preserving manual effort at this layer. This is not where HR’s value sits.
The next layer involves structured HR processes that still require context and care. For example, a system that identifies a return from extended sick leave and prompts a wellbeing check-in or engagement survey is good HR. AI can initiate, prepare, and prompt at this layer, provided judgement remains human and outcomes are reviewed before action is taken.
The highest-risk layer of HR work involves judgement, decision-making, and delivery. Performance management, investigations, psychosocial risk interventions, redundancies, and terminations sit here. This is where “do it for you” systems do not belong. Accountability, discretion, and human presence are not optional in these moments.
I do not believe I will see a world in my lifetime where an employment termination conversation is delivered by a bot.
More importantly, such a system does not belong in Australian HR practice. Not because the technology could not be built, but because it would be fundamentally incompatible with our legal framework, professional standards, and expectations of fairness, accountability, and care.
This is where agentic AI is often misunderstood. In HR, agentic AI should not mean autonomous delivery. A more accurate analogy is navigation.
Agentic AI should work like Waze for HR. You usually know where you are going. What helps is the fastest route, real-time warnings, and clear signals about what’s coming next.
Waze does not decide the destination. It does not drive the car. It does not carry responsibility for the outcome. It reduces cognitive load, flags risks, reroutes around problems, and helps you arrive faster and safer while you remain fully in control. That is what good agentic AI should look like in HR.
Applied properly, agentic AI can guide practitioners step by step through complex HR processes, surface legal and regulatory risks at the point of work, and highlight decision points that require care. The practitioner still decides, delivers, and remains accountable. This is copiloting HR delivery, not outsourcing it.
Other advanced AI terms are best understood as directions of travel, not adoption goals for HR. Ambient AI may support background monitoring and signals at the data layer. Neurosymbolic approaches may improve explainability over time. AGI remains hypothetical and is not a practical or relevant consideration for HR service delivery. None of these alter the requirement for human-led HR outcomes.
Research on AI use by knowledge workers consistently shows material efficiency gains and significant improvements in output quality when AI is applied well. While these findings are not HR-specific, they are highly relevant. HR is a knowledge-intensive profession, and the same dynamics apply.
Applied thoughtfully, AI enables HR practitioners to move faster through complex work, reduce cognitive load, improve consistency, surface risk earlier, and spend more time on judgement and delivery rather than administration.
The challenge for HR is not whether to adopt AI. It must. The challenge is adopting it without outsourcing judgement, authority, and accountability to systems. AI should automate low-value work, manage data and signals, and act as a copilot for delivery. Humans must remain responsible for decisions and outcomes.
HR does not need self-driving systems. It needs better navigation through complexity.
Automation should accelerate. Copiloting should expand. But the delivery of HR outcomes must remain human-led.
That is the line.
AI should help HR do its most important work better. The moment it starts doing that work instead, HR is no longer HR.