A hiring plan is a price list written in the wrong units. It arrives looking like a schedule, so it gets reviewed like a schedule, and the question put to it is whether the order makes sense. The question that decides what the programme actually costs is a different one, and almost nobody asks it in the room: how much of the eventual organisation has to be standing before the first thing ships.
A global staffing and HR services group building an integrated digital platform across its national markets wrote that answer down explicitly, which is rarer than it sounds. Five tables, one per skill family, forty-three distinct roles in total, and against each role a single tick placing it in one of three delivery horizons. Twenty-four roles carry a tick in the first horizon. Seventeen in the second. Two in the third.
Fifty-six per cent of the eventual role set has to exist before the first horizon closes. Ninety-five per cent before the third one begins. Read as a schedule, that is a sequencing decision. Read as a price, it is a different programme from the one the schedule implies.
The challenge
The counts are not a summary anyone wrote. They come from counting the rows: seven management and delivery roles, eleven technology, ten data, twelve creative, three operations. The technology table is the one place that needs care, because it enters the same role twice, same description and same tick, so twelve rows there resolve to eleven roles. Every remaining row carries exactly one horizon tick, so the arithmetic is unambiguous and the totals close on forty-three.
Inside the first horizon, the composition is the part worth sitting with. Nine of the twenty-four are creative roles. Six are technology. Four are data, four are management and delivery, and one is operations. So the wave that is supposed to be the beginning is mostly the people who make the thing customers see, plus enough architecture to hold it up, plus a single operational role. The family that runs the platform day to day has three roles in the whole plan and only one of them in the opening wave.
That is a coherent choice, and the stated reason for it is on the same page as the tables: the first horizon has to detail the strategy and execute it at the same time. The specification and the build are not sequential here. They are the same wave, staffed once. Which means the plan is not asking for twenty-four people to start work on a defined backlog. It is asking for twenty-four kinds of judgement to be present in the room while the backlog is still being written.
There is a second axis in the same tables that changes what the twenty-four actually means. Each role is also placed against three market sizes: central, large local, small local. Most technology and architecture roles are central only. Two creative roles are the opposite, marked as not central at all and placed in the markets instead, and both of them sit in the first horizon. So the largest family in the opening wave is also the most distributed one. Twenty-four roles in the first horizon are not twenty-four seats in one building. Some of them are the same role instantiated in several markets at once, and the plan does not say how many.
That is the honest limit of what these tables support. They give variety of skill required and when, not headcount and not cost. Anyone reading the twenty-four as a budget line is adding a number the source never wrote.
The approach
Sourcing was treated as its own decision rather than as an implementation detail, and four routes were evaluated: retraining existing staff, permanent hiring, contract hiring, and outsourcing the whole capability. The recommendation was three of the four run in parallel, with outsourcing set aside on the grounds that handing out an operation that does not exist yet is risky and might be revisited once it is running.
The written reasons attached to each route are more useful than the recommendation. Retraining builds on skills the group already has and offers a path to people already inside it, and it carries a lead time before anyone is productive. Permanent hiring brings the right skills and carries its own lead time to find and onboard the people, in several countries rather than one. Contract hiring is the only route described as quick to hire and onboard, and the only one carrying an explicit option to reduce again once upskilling is complete. The drawback recorded against it is not that the skills leave. It is fit: contractors are marked as potentially less customer-centric because they are not tied to the group's own values, and like permanent hiring they still have to be found in several countries rather than one. Access to a large talent pool at short notice is written against outsourcing, which is the route the recommendation drops.
Put that grid next to the curve and the tension is plain. Both routes that leave permanent capability behind carry a lead time written against them, one to upskill and one to find and onboard, and the plan needs a majority of its roles present in the first wave, in multiple markets, before the wave can produce anything. The only route with no lead time against it is the one the plan already expects to unwind once the other two land. A staffing group is better placed than most to absorb that. It is not exempt from it.
This is where a plan stops being a plan and becomes a procurement problem with a shape. A front-loaded curve prices differently from an even one in at least two ways that a schedule review will not surface. The first is bench time: roles that arrive before there is a backlog deep enough to keep them busy are paid for regardless. The second is concentration risk in recruitment itself, because twenty-four role types landing in one window across several markets competes with itself for the same candidates and the same interview capacity.
The outcome
What this phase produced is a blueprint: role definitions, a horizon assignment for each one, a placement against market size, an operating model choice and a sourcing recommendation. Nobody was hired inside it. The forty-three roles are a design, the twenty-four is a plan, and both are proposals awaiting a business case. Nothing here is a delivered organisation and the piece should not be read as one.
Three things would be done differently now, and none of them is about the counts.
The horizon assignment was written once, by judgement, and it has no feedback path. There is no mechanism in the blueprint for discovering that a role placed in horizon two was actually the constraint in horizon one. Process mining over the delivery workflow gives that back: which handoffs stall, which queue is always deepest, which approval is the one everything waits behind. A role sequence derived from where work actually stops is a different list from a role sequence derived from where work logically begins, and the difference is worth finding before the offers go out.
The data roles were defined before the data itself was ready to be worked on. Ten roles were written and four scheduled early, which is a reasonable sequence in isolation, but a data organisation hired ahead of a feature store, a deployment path and a lineage trail is an organisation that can build models it has no route to production for. The sequencing that matters is not which data role comes first. It is whether data readiness arrives before or after the people paid to depend on it. Our Consult work now puts that dependency on the same page as the hiring curve, because it is the cheapest correction available and the most commonly skipped.
And the early language-model pilots now being run cautiously in document handling and candidate screening change none of the arithmetic above. A pilot still needs a product owner who can decide what it may be used for, a compliance role that can say what it may be used on, and an operational path to put its output somewhere a person will act. All three of those sit in the first horizon on this plan or the pilot is a demonstration. Model capability arriving faster does not flatten a hiring curve. It steepens it.
The shape of the curve is the commercial term. Everything else in the blueprint is negotiable later.