The Retirement Wave: Capture Expertise Before It Walks Out

Every year around 1.3 million people in Germany retire while only about 800,000 enter the labour market. By 2036 that leaves a gap of roughly 4.3 million workers (IW Köln, 2026). Comparable demographic curves apply across most of Western Europe.

These figures are known. What planning tends to miss is their operational translation: it is not only workers who leave. Capability to decide leaves with them.

What actually gets lost

When someone leaves after 30 years, you do not lose their job description. That exists. You lose the answers to questions written down nowhere.

How do you tell this batch will cause trouble before the readings look abnormal? Which supplier delivers reliably under pressure and which one only promises? Why is plant 3 run differently from the other two? At what point do you escalate a customer case, and to whom exactly so it does not sit for three weeks?

42 percent of job-related knowledge is documented nowhere (Panopto/McKinsey). Among long-tenured employees the share is typically higher, because their knowledge consists precisely of the cases no process anticipates.

The costs this creates

Replacing a departing employee costs between 100 and 300 percent of annual salary (NAIC). The largest share is not recruiting, it is productivity: 58 percent of total turnover cost falls on lost productivity and absence (Work Institute 2024).

Further figures that complete the picture:

• 72 percent of companies have at least one person whose departure would significantly impair operations (SHRM 2023)
• 48 percent lose institutional knowledge with every departure (Retention Report 2024)
• 54 percent report project delays caused by turnover (Retention Report 2024)
• At a company with 30,000 employees, annual knowledge loss runs to roughly 72 million euros (ProcedureFlow)

And the item that irritates most: departed experts return as consultants at two to three times the daily rate, according to HBR. You buy the same knowledge a second time, more expensively, and it still does not flow back into the company.

Why succession planning is not enough

Only 21 percent of companies have a formal succession plan (SHRM). Over 56 percent have none at all. That is one half of the problem.

The other half concerns the 21 percent who do have one. A succession plan determines who follows. It does not transfer what the predecessor knew. SAP SuccessFactors and Workday identify successors reliably and do not secure the departing person's knowledge.

You know the result: the succession is filled, and three months later the new person is on the phone with the old one, because a case comes up that is not in the handbook. That the wiki does not close this gap is shown in the piece on the Confluence alternative.

Why the usual countermeasures arrive too late

The three standard answers share one flaw: they start when the resignation is already on the table.

1 Double staffing during handover

It works, but it is expensive and only covers what happens to occur in those weeks. The rare case usually does not turn up during the handover window.

2 Asking the departing person to document

Produces process descriptions. Exceptions and decision criteria are almost always missing, because they do not strike the writer as noteworthy.

3 Bringing them back as a consultant

It works and costs two to three times the daily rate. And at the second departure you are back at the start.

The shared flaw is timing. Knowledge capture that only begins when a departure is announced has four to twelve weeks. That is enough for processes, not for experience.

The other approach: by risk rather than by notice date

Do not start with the people who are leaving. Start with the roles that are critical.

Four criteria are enough for a first prioritisation. How high is the damage if this role sits vacant or inexperienced for three months? Does the knowledge exist more than once in the company or only once? How likely is a departure in the next 24 months, through age, reorganisation or attrition? How much of it exists documented?

Those four values produce a list you can work through. Usually it is fewer roles than feared. In companies with a few thousand employees, the number of genuinely critical single-person dependencies often sits in the low double digits.

For those roles the knowledge is captured while the person is still there and working without time pressure. How such an AI-supported expert interview works methodically is covered in its own piece. Expect roughly four weeks per expert, spread across several sessions, the experience value from projects including Hörmann and BASF.

What comes out of it

Not a folder. A playbook that guides a decision: criteria, exceptions, warning signals, escalation paths, dependencies. Every knowledge object with an owner, a validity date and a verification status, so it stays visible later what still applies.

It becomes usable in several forms: as knowledge articles and SOPs, as training video for the successor, as an internal podcast for onboarding, and as context for internal AI systems via RAG API and MCP server.

The effect shows in two places. Onboarding and ramp-up time typically fall by 15 to 30 percent, queries to experts by 20 to 40 percent. Both are experience values from projects and customer feedback, not commitments.

Employee representation belongs at the start

A knowledge capture programme without early involvement of employee representatives gets expensive, because it has to be set up again later.

Three points carry the conversation: participation is voluntary. The purpose is transparent and set down in writing. Access is governed role-based with an audit trail. The framing is decisive: this is about role knowledge and operational capability, not the assessment of individuals.

In practice the concern on the experts' own side is larger than that of the committee. It cannot be argued away, but it can be placed: whoever secures their knowledge is freed from routine queries and is needed for the cases where experience genuinely counts.

The first step that shows results in eight weeks

Pick one role. One whose departure is due in the next 24 months and whose absence hurts operationally. Capture it completely, make the result available to the team, and measure one number: how many queries still reach that person after four weeks?

That single number carries the discussion in the executive team further than any market study. It comes from your own organisation.