HR now has to design the productivity of humans and AI agents at the same time — and most HR functions are not set up for it. This is written for HR leaders and CHROs who are being asked to plan a workforce that is no longer only human.
Here is the specific thing that breaks. When agents do real work, headcount stops being the unit of capacity. Job architecture, performance management, workforce planning, succession, span-of-control benchmarks — every one of these was built on the assumption that a role maps to a person. That assumption is not slightly strained. It is now the constraint.
The shopping-list version of AI transformation
Since I started publishing on AI-native org design I have had a steady stream of questions asking for real cases — what companies have actually done, not what consultancies recommend. So let me offer one that changed how I think about where transformation should start.
First, the pattern it breaks from. Most companies treat AI transformation as a shopping list: buy the tools, provision the accounts, run the trainings, report the licence uptake. All of it is necessary. None of it is transformation. If people stay in their existing roles and simply work a little faster, nothing structural has changed. You have handed everyone one more tool while the organisation continues running on its legacy operating system.
So where should real transformation begin? I would have guessed operations, or whichever function has the most obviously automatable work. The answer from a Chinese smart lock manufacturer surprised me: start with HR.
Splitting HR in two
They divided their HR Center into two distinct divisions:
- Human Service — stability, compliance, operations. The genuinely essential work of employment: contracts, payroll, labour law, employee relations. This does not go away and should not be disrupted for the sake of appearing modern.
- HARO — the Human AI Resource Office — frontline design, human–agent collaboration, business outcomes. A function whose job is to design how work gets done when part of the workforce is not human.
What makes this more than a reorganisation is what the split protects. The two halves have genuinely incompatible operating rhythms. Compliance work must be careful, precedent-bound and slow; work design must be experimental and fast. Housed in one function, the careful half reliably strangles the experimental half — not through malice, but because risk aversion is the correct instinct for the compliance mandate and the fatal one for the design mandate. Separating them is how you stop the organisation’s own immune system from killing the thing it needs.
Why HR, and not IT
The insight underneath the structure is the part worth carrying away: AI transformation is ultimately organizational transformation.
And HR governs the most foundational variables in an organisation — who you hire, how you define roles, how you measure results, what gets rewarded. Those four levers determine what actually happens after the tools are installed, and none of them belong to IT. You can give a person the most capable agent available; if their job description, their objectives and their performance rating all still describe the work they did before, they will keep doing the work they did before. Rationally.
This is also why AI programmes owned entirely by technology functions tend to stall at the pilot stage. They can prove capability. They cannot change what the organisation asks of people, because that lever was never in their hands.
What this asks of HR
It is a substantial ask, and worth naming honestly. Designing human–agent productivity requires HR to hold questions it has not historically owned:
- If an agent performs 40% of a role, what is the role now — and what is the person’s new centre of gravity?
- How do you measure the performance of someone whose main contribution is orchestrating agents rather than producing output directly?
- What is the accountability model when an agent gets it wrong? Who is rated on that?
- How do you do workforce planning when capacity is no longer a headcount number?
- What do you owe the people whose roles thin out — and when do you owe it to them?
Most HR functions cannot answer these today, and there is nothing shameful in that; the questions are genuinely new. But the questions are being answered anyway, by default, by whoever deployed the agent. Somebody is making these decisions in your organisation right now. The only real question is whether it is being done deliberately.
One question to sit with
Is your HR function ready to design the productivity of humans and agents — at the same time?
If the honest answer is no, that is not a failure. It is the most useful thing you could know this quarter, and it is considerably cheaper to find out now than after the operating model has been set by accident.
📚 Start here: AI-Native Organization Design — the full research hub, with every article and video in one place.
More in this series
- 67% of AI Adoption Blockers Are Organizational, Not Individual
- $2M Revenue Per Employee: The Case for the Living Organization
- The Living Organization: Five Pillars of an AI-Native Operating Model
- Why Hierarchy Existed — and Why AI Just Removed the Reason
- How Much Does a CEO Really Need to Understand About AI?
- The Five Steps of AI Adoption: From Gated to AI-Native
- Your AI Isn’t Failing — Your Specification Is
- Stop Measuring Tomorrow’s Talent With Yesterday’s Rulers
Watch the full series
This article accompanies a video from AI-Native Operating Model and Org Design — a series on how organisations actually absorb AI, and where they break.
- ▶️ Watch the full playlist — AI-Native Operating Model and Org Design
- 📺 Subscribe on YouTube — @BreezeDONG
Chunfeng “Breeze” Dong is an executive coach (ICF PCC, CPCC) and founder of Springbreeze Ventures, with twenty years in organisational development inside Fortune 100 companies — Roland Berger, Siemens, ABB and Roche. She writes on AI-native organisation design, human–agent governance and change leadership.
📘 The Living Organization · 📘 A Soulful Transition · 🔗 LinkedIn
