AI for HR directors: workforce cost, attrition and succession, grounded in your own files

Personnel cost lands 3% over budget and the reflex in the MT is that HR is too expensive. But open Personeelskosten_definitief_2026.xlsx and the fte-realisatie sits under budget. The entire overrun is external hire. Nine roles running on ZZP and uitzend at 1.6x internal is roughly €48k a month, near six hundred thousand a year, paid to contractors because you can't fill the seats. That's not an HR-is-expensive story; it's a we-pay-a-premium-because-people-leave story, and until you can show it that way, you defend the wrong number in the wrong meeting.

Four high-performers on the risk list match last year's leaver pattern (salary capped at the ceiling, no recent development step) and two of them sit in the scarce care roles you're already backfilling at 1.6x. Right now that lives in Verloop_risicoanalyse_H1_2026.pdf and in your head, and the MT only listens once someone has already handed in their notice. Gallup puts the cost of replacing someone who voluntarily leaves at one-half to two times their salary, and preventable turnover at over $1 trillion a year for U.S. business. Seeing the pattern in time is the difference between a retention talk before summer and a scarce-role vacancy at premium rates.

Two long-term sick cases hit the end of their 104 weeks next month with no signed plan van aanpak: the exact gap UWV penalises with a third salary year. It's sitting in Verzuimdossiers_104wk_juli2026.xlsx as a casemanager export, but the deadline only surfaces when you go looking, and a missed one is a loonsanctie you then have to explain to the MT. The shape generalises beyond the Netherlands: every statutory deadline and scarce-role exposure bites hardest the moment it slips past unnamed.

Ask where the overrun really comes from and you get the split (formation under budget, the whole miss in external hire, sized in euros per month and per year), with the source file named. It reframes the tired "HR is too expensive" conversation into the true one: the premium is what unfilled and vacated roles cost. That's the version of the number you can put in front of the MT and stand behind.

The people you can't afford to lose, surfaced against the pattern that predicted last year's leavers, with the scarce-role exposure flagged and the euros attached. It's honest about its limit: it surfaces a pattern to act on, not a certainty about one person, so you book the retention conversation, you don't file the individual. But you have the conversation before the resignation, not the exit interview after it.

The two dossiers near a statutory deadline, the scarce roles with no bench behind them, the succession notes that only lived in your head, assembled from the exports you already run, each answer pointing back to the file it came from. You walk into the week naming the specific exposure instead of hoping nothing slipped, and you name it before a regulator or a resignation does it for you.