Every CHRO I work with says the same sentence at some point: "We know training matters, but I cannot defend the number in an exco meeting." That sentence is the whole problem in miniature. The spend is defensible. The way it is presented is not.
A cost centre is any function that can only describe itself in inputs. Rands spent, courses run, seats filled, satisfaction scores. When the finance director looks at a line item that reports itself in inputs, the rational thing to do in a tight year is to reduce the input. Nothing in the reporting suggests any output will move.
What changed in the last three years
Two things, in South Africa specifically. The first is the scarcity of certain technical skills becoming a hard operational constraint rather than a recruitment inconvenience. Mines and utilities are competing for the same small pool of GMR-appointable engineers and experienced protection specialists. When you cannot buy the capability, you have to build it, and building it becomes an operating requirement with a delivery date.
The second is that the money now has strings. B-BBEE skills development spend, SETA discretionary grants, mandatory grant claims — these already force a level of documentation that most other functions do not face. Organisations that treat that documentation as a compliance chore get nothing from it. Organisations that treat it as the beginning of a measurement system get a running record of who was developed, in what, at what cost, and what happened to them afterwards.
The repositioning move that actually works
The CHROs who have won this argument did not do it with a better learning strategy deck. They did it by attaching the learning plan to a small number of operational risks that the executive already worries about.
An example from a Mpumalanga processing operation. Their engineering manager population had an average age above 54 and no bench. Instead of proposing "leadership development", HR presented a table: each critical appointment, the incumbent's expected exit date, the named internal candidate, the specific gaps between that candidate and the appointment requirement, and the cost and duration of closing each gap. Total programme cost was roughly nine percent of the cost of a single unplanned twelve-week vacancy at plant manager level, which they had already lived through.
That table survived two rounds of budget cuts. The generic leadership programme they had run for four years did not, and should not have.
Three practical changes
Report in outcomes the business already tracks. Not learning outcomes. Business ones: time to competence for a new artisan, repeat-defect rate, CoC rejection rate, first-time-right on statutory inspections, vacancy duration in critical roles. If the learning function cannot influence any measure the business already reports, that is worth knowing.
Put a named executive sponsor on every programme. Not as a courtesy. The sponsor signs off the capability gap at the start and confirms at the end whether it closed. Programmes that cannot find a sponsor are usually programmes nobody needs.
Stop running open-enrolment catalogues as the default. A catalogue is a supply-side response to a demand-side problem. It is useful for genuinely individual development, and it is a poor instrument for closing an organisational capability gap on a schedule.
What competitive advantage actually looks like
It is unglamorous. It looks like being able to staff a shutdown from your own people when your competitor is paying contractor rates. It looks like promoting from within because the bench exists, which shortens the vacancy and keeps institutional knowledge in the building. It looks like an engineering function that can absorb a new standard revision in six weeks rather than eighteen months.
None of that shows up in a satisfaction score. All of it shows up in operating cost.
The repositioning is not a branding exercise. It is a decision to hold the learning function accountable for a small number of things that matter, and to stop holding it accountable for volume. In my experience the function welcomes that trade almost every time — it is far easier to defend six programmes with owners than sixty without.