There is a question being asked in more and more boardrooms, and it is not the question HR spent the last decade preparing to answer. For years the demand on the function was operational. Run the processes well, keep the service levels green, deliver the programme on time. The question this year is different in kind. It is no longer whether you can run efficiently. It is whether you can design what the workforce will need to be able to do, and ensure the design can grow and morph with unprecedented change.
This spring, at the Workforce Digital Transformation Summit in Amsterdam, that shift was visible in the room. Company after company presented its work on artificial intelligence in the HR function, and a pattern separated the ones seeing results from the ones still waiting. The organisations getting value had not bought the most technology. They had gone in with a clear view of what they wanted, treated AI as a tool to augment their people rather than a silver bullet to solve all problems. They first diagnosed the state of their own house before deploying anything into it. Their HR leaders had also started describing themselves in a particular way. Not as the function that administers transactions, but as the function that designs the conditions the workforce operates inside, and stands behind that design.
One European operator put it on a slide, as a trajectory: of first being a support function, then business partner, then, in the near future, workforce architect. That is not a vendor's coinage or a conference theme. It is a large, serious organisation naming in public the role it believes its own HR function is becoming. This essay is about that role. What separates the architect from the administrator, why the distinction now decides whether AI delivers anything at all, and, because a diagnosis without a first step is just a better-informed anxiety, what a leader can actually do about each part of it.
The architect starts from demand
The administrator begins with the workforce it has and asks what that workforce can do. The architect begins with the work the organisation will need and asks what workforce that requires.
The same company that named the architect role also inverted its planning logic to get there. Most companies, and most of the vendors selling to them, start from the supply side, as in "here are our people, here are their current skills, let us map those forward". That firm chose, after a year of deliberating, to start from the other end: the company's own demand out to 2030 first, and only then the skills to meet it. It sounds academic until you notice what it changes. A supply-first plan optimises the people you happen to have. A demand-first plan tells you which people you will need to become, and where the gap sits before it becomes a crisis. The architect designs toward the second question. The administrator spends investment cycles staffing the first.
The model that served us has been outgrown
There is a reason the administrator posture is so deeply set, and it has a name. The Ulrich model, the operating model that for decades organised the HR function into business partners, centres of expertise and shared services, gave the profession a clarity it badly needed and served the industry well. It has since outgrown the business it was built for. Designed for a world of relatively stable roles and predictable planning cycles, it made the structure of HR legible, often leaving out what the employee actually experiences. That gap is why employee experience got bolted on later as a separate concern, and, being an add-on with a soft-sounding name, why it was so often the first line cut when budgets tightened.
Now the ground the model stood on is moving under it. AI can absorb a large share of the transactional work the shared-services layer was built to handle, and the planning horizon has compressed from years to quarters. An operating model optimised for stable administration is, in that new environment, optimised for the wrong thing. The model is not wrong. The world around it is rapidly changing and calls for a more nimble approach. The pillars of the Ulrich model have become weight in themselves, outcosting their purpose. They can no longer hold up to the pace of business change.
The temptation, on realising this, is to propose a reorganisation. That is almost always the wrong first move. It is expensive, slow, and easy for a board to decline. The change has to be earned with evidence before it is argued as structure.
The measure is value, not motion
Suppose the architect is in place and the demand is clear. There is still a way for all of it to go quietly wrong, and it goes wrong at the level of measurement.
The productivity case for AI is real, but it is routinely overstated, because of what gets counted. This January, one large workforce study put a number on the gap. For every ten hours of efficiency gained through AI, nearly four are lost correcting, verifying and reworking what it produced. Only around one in seven employees consistently saw a net-positive outcome. The efficiency is not imaginary, but neither is the tax on it, and organisations that report the gross figure while ignoring the tax are measuring motion rather than value.
Underneath the arithmetic sits a subtler failure, and it is a failure of metric choice. Ask why a team reports the metric it does, and the honest answer is usually that the metric is easy to count. Time to hire is a date subtracted from a date; you can read it this afternoon. Quality of hire is a composite you cannot read for six months. So the easy one wins, and the organisation optimises the speed of a thing without ever checking whether the thing got better. A ticket is logged closed the instant it closes. Whether it stayed closed takes a week to know. The real metric is almost always the harder one to read, which is exactly why the easy one keeps winning. One organisation at the summit reframed its entire approach to workforce planning around this idea, insisting the plan be treated as a live model its leaders keep current rather than a static snapshot outdated within six months of approval. A movie, they said, not a photograph. The reframe is the one the measurement demands. Stop reading the still frame. Start reading the motion that matters.
Do one thing all the way down
There is, finally, a practical sequence the architect has to work through. What to offload to AI. Whether the data underneath is clean enough to offload it. How the introduction of capable agents changes the shape of the work and who governs it. What skills and structures must retire. What the adoption actually costs. And how the humans in the system come through it intact rather than merely displaced. It is a real list, and every item on it earns its place.
But the list is where good intentions go to die, because organisations work it the wrong way. They take all of it on at once, run every question in parallel, and produce a thin answer to each. A shallow audit of everything, and a change in nothing. The most common failure here is not neglect. It is breadth without depth, six half-answers where one whole one was needed.
The discipline is to take a single journey and follow it all the way to the floor.
What the role is for
Strip away the summit slides, the operating-model history and the measurement arguments, and the workforce architect is a posture before it is a title. It is the decision to design the conditions the workforce operates inside rather than administer the conditions it inherited, and to be accountable for the design. That accountability is the same thread this journal followed in June. The organisation that can explain and justify what it does to people is the one that will pass the reviews no one scheduled. The architect is the person who makes that explanation possible, because they built the thing on purpose.
You do not need a dedicated programme or a big budget for any of this. The largest organisations at the summit had funnels and centres of excellence. None of that was the difference. The difference was how they saw the work, and seeing is free. It is built from small, concrete acts. One capability list for one unit. One honest account of what poor design cost. One net-value count on one use case. One journey followed to the floor. The architect is made by doing those things, not by reorganising.
There is a next horizon already visible from here, and this journal will take it up in the coming month. It is the possibility that each employee comes to have a kind of personal instrument, a companion that reads their own development the way a health app reads a body, tuned to the direction the organisation is heading. One operator at the summit has already built the first version of it. That is where demand-first, human-centred architecture points next. For now, the work is nearer to hand, and it was always this work. The technology simply set a date for it.
Clover & Myrtle is a vendor-independent HR transformation practice. The work the industry forgets, after go-live and underneath the deadline.
cloverandmyrtle.com/journal