Retirement wave planning

The mainframe is not your problem. The four people who understand it retiring in March is your problem, and unlike most modernization risks, this one comes with a countdown that HR can print. Estates age in two ways at once: the systems get older, and the people who hold them get closer to a send-off cake. Only one of those trends is on your roadmap.

Run the overlay exercise. Two lists, one afternoon:

1. From the inventory: every system with exactly one knowledgeable human. Not the listed owner; the person the listed owner calls.
2. From HR, handled with the discretion it deserves: tenure and age bands for those names. You are not asking for anyone's plans. You are asking which single points of knowledge are, actuarially, on the clock.

The overlap is the real modernization backlog, sorted better than any technology-based sort. A creaking system whose expert is 34 is a someday problem. A stable system whose expert is 63 is a this-fiscal-year problem, no matter how quietly it runs. Stability is not safety when the stability is a person.

What to do with the list, in order of yield:

1. Understudies, formally. Not "shadowing when convenient": a named second, with production access, doing the monthly close and the quarterly weirdness under supervision, twice. Paired on-call is the fastest knowledge transfer ever invented, because the system chooses the curriculum.
2. Extraction by doing, not by interview. "Write down what you know" produces a document about what is easy to write down. Instead: the understudy executes, the expert corrects, someone captures the corrections. Runbooks written this way smell like the real system.
3. Rehearse the absence. The expert takes an unreachable two-week vacation, on the calendar, this year, while still employed. Every gap that surfaces is a gap found at rehearsal prices instead of retirement prices.
4. Retention, priced in the open. Sometimes the right move is a part-time contract that keeps the expert attached through the migration. Pay for it like the insurance it is, and set an end date, or the arrangement becomes the new single point of failure with a day rate.

What not to do: schedule the knowledge transfer for the expert's last month. Notice periods are for handing back the laptop, not for downloading a career. The transfer takes a year of overlap done properly, which means the trigger date is not the retirement date. It is retirement minus eighteen months, which for the names on your overlap list is, plausibly, now.

The wave is not a surprise. It is the most predictable event in the estate. Being surprised by it anyway is a choice, made annually, by omission. Make a different one this budget cycle. And plan the work around the fourteen people who can safely do it, because modernization is a staffing problem long before it is a technology one.