1. The demographic pressure on field service, in numbers
The retirement of the baby boomers is not an event but a period. According to the German Federal Statistical Office, by 2040 around 13.3 million members of the labour force will have passed the statutory retirement age of 67 - 30 percent of the 2025 labour force. The share of those aged 55 and over in the labour force has risen within ten years from 20.7 to 27.0 percent.
Mechanical and plant engineering sits exactly at this level. The VDMA press release of 14 August 2026 states: “A good quarter of the employees in mechanical engineering are 55 or older. There are many retirements in the industry. If these positions cannot be refilled, companies lose valuable know-how.” The same release reports around 21,500 jobs cut in the first half of 2026. What the industry faces by 2034 has been projected by the IW for the VDMA’s Impuls Foundation: around 296,000 retirements against about 118,000 newcomers.
The source that names the problem most precisely is the DIHK skilled-labour report 2025/2026 (Association of German Chambers of Commerce and Industry): 23 percent of all companies expect company-specific knowledge to be lost when employees leave for age reasons, in industry 35 percent, in mechanical engineering 41 percent - one of the highest values of all. The DIHK’s explanation reads like a description of technical service: long tenure, “through which employees acquire deep, undocumented knowledge”; and on small and medium-sized companies: “There, knowledge is often concentrated in individuals and not documented.”
One note on context: the current skilled-labour shortage analysis of the Federal Employment Agency (May 2026) no longer counts technical service staff at skilled-worker level among the shortage occupations, for cyclical reasons. So the shortage is not noticeable right now, the departures happen anyway - and according to the same analysis, a vacancy for technical service staff in maintenance and repair nevertheless stayed open for a median of 140 days in 2025.
2. What is really lost with an experienced service technician
In 1966 the philosopher Michael Polanyi coined the sentence that captures the problem: “We can know more than we can tell.” He called it tacit knowledge: skill that shows in doing but cannot be fully translated into rules. A service technician who diagnoses a machine by its sound is the textbook example.
What that means in everyday service work can be sorted into six kinds of knowledge - none of which is in the instructions for use:
| Kind of knowledge | Example | Where it lives today |
|---|---|---|
| Fault patterns without a fault code | “If the hydraulics whistle on start-up in winter, it is the check valve, not the pump.” | In heads, sometimes in a notebook |
| Customer and machine specifics | Which machine at customer X was modified in 2014, which controller version runs, who decides there | In heads, scattered across emails |
| Diagnostic paths and wrong turns | Which three causes to rule out first - and which check costs time and never yields anything | Nowhere; service reports document the fix, not the path |
| Workarounds and temporary fixes | How to keep the machine running safely until the spare part arrives | In heads; deliberately undocumented |
| Part equivalents | Which component from another series fits when the original is discontinued | In heads, in the workshop drawer |
| Network | Whom to call in engineering when it is about the controller software from 2011 | In the phone |
Typology from our project practice, not an empirical survey. The examples are fictitious, the patterns are not.
That this knowledge really does sit with individuals is shown by the Insight-Report Service 2022 by the documentation service provider kothes and the German customer service association KVD (service technicians in the German-speaking countries, 43.6 percent from mechanical and plant engineering). A good 83 percent of respondents rely on their own notes to solve service cases. Asked what happens to these notes, 70.4 percent answered: “I use them for myself.” Only 25.7 percent can make them available to everyone through the organisation, and almost 60 percent do not know where to look for the information they need in the first place. The report draws the conclusion itself: the knowledge of long-serving service technicians “often exists only in the form of personal notes and documents”.
What that does is measured by the 2026 Field Service Benchmark by the vendor Aquant (vendor study) as the “skills gap”: the difference in first-time fix rate between the best technicians and everyone else. At underperforming service organisations it is 10 percentage points, at the best 2.9 - widest where “critical knowledge remains concentrated among a few technicians”. What that means in repeat visits and costs we have calculated elsewhere. What matters here is the mechanism: a service organisation whose quality depends on two or three people has already priced in its knowledge loss - it only becomes visible when those people leave.
3. Why manuals, wikis and handover meetings do not catch it
The three usual answers fail for three different reasons - and section 4 starts from exactly those.
- Technical documentation describes the machine, not the case. Instructions for use, circuit diagram and spare parts list are indispensable, but they answer the question “What is this component?”, not the question “Why has the machine at customer X been down since this morning?”. Half of the respondents in the Insight-Report Service 2022 said that step-by-step troubleshooting instructions are missing in their daily work. With the Machinery Regulation from January 2027 documentation goes digital - easier to find, but no richer in experience.
- The wiki fails at writing, not at reading. Almost every service organisation has set up a knowledge base at some point. It fills up with what was documented anyway, and then it stops: the technician who comes back from a visit at 5 pm does not write a wiki article but three lines in the service report - “valve replaced, machine running” - without symptom and without diagnostic path. A system that demands more writing than the working day allows is a graveyard after six months.
- The handover meeting asks for what only the situation brings out. The last weeks before retirement are the worst time for knowledge retention, because experience is not available on demand. Ask a technician “What should we know about series Z?” and you get a quarter of an hour of generalities. Put him in front of a series Z machine that has shown a fault since this morning and you get an hour of substance. Polanyi’s sentence applies literally - as long as you only ask.
Most machinery manufacturers, incidentally, signed up to this requirement long ago: DIN EN ISO 9001:2015 demands in clause 7.1.6 (“Organizational knowledge”), as quoted from the standard by the guide of the German quality and knowledge management associations DGQ and GfWM: “The organization shall determine the knowledge necessary for the operation of its processes and to achieve conformity of products and services.” The informative Annex A.7 explicitly names protection against loss of knowledge, for example through staff turnover, as a purpose. In most audit reports the point is ticked off with a training matrix and document control; the experience of the service technicians does not appear there. Anyone planning the transition to ISO 9001:2026, published on 16 September 2026 - certificates to the 2015 version are expected to remain valid until 30 September 2029, according to the certification body DQS -, has a reason to take the clause seriously this time.
4. The concept: a living knowledge base with a human in the loop
If experience only becomes visible on real cases, knowledge retention has to happen where the cases are: on the service visit, in the hotline, in the workshop. Four principles follow, and they matter more than any technology.
The service report is the only moment at which experience is verbalised anyway. It has to be built so that symptom, context, diagnostic path and fix are captured in five minutes - spoken, not typed, from the car park if need be. Speech recognition and a language model that turns this into the structure from section 5 are not a convenience but the precondition for anything being captured at all.
The foreman with 30 years of experience does not write articles. Every week he is shown the cases from his machine series that younger colleagues have captured, and adds in two sentences what is missing: “Next time check the pressure switch first, the valve is almost never the cause.” That is human in the loop in the proper sense: the human curates, the system collects.
When a technician in the field asks the system and gets an answer, he has to see which service report, which colleague and which machine it comes from - and how often this fix has worked since. A knowledge base that melts experience down into anonymous rules loses exactly the context that separates experience from manual knowledge. What else such a system has to be able to do is in our decision-maker’s checklist for AI assistants in field service.
A case that could not be solved with an old fix is the most valuable piece of information in the system: the fix was incomplete, the machine has changed, or there is a second cause with the same symptom. Whoever does not collect this feedback has a manual again after two years, just a self-written one.
The role of AI in this picture is deliberately modest: among a thousand service reports it finds the one that matches the current symptom, even if that report says “whistling” and the person searching says “squealing”. That creates no knowledge that nobody captured. Where AI in after-sales works beyond that, and where it does not, is the subject of our guide to AI in after-sales and service.
5. In practice: capturing service cases so that experience is in them
“Valve replaced, machine running” is worth nothing; the value is in what happened before. A usable case has eight fields, several of which are missing from classic service reports:
| Field | What belongs in it | In the classic service report? |
|---|---|---|
| Symptom | What the customer reported and what the technician found - in the customer’s words, not as a fault code | Partly |
| Context | Machine, serial number, controller version, known modifications, operating conditions (temperature, shift, material) | Serial number only |
| Diagnostic path | What was checked, in which order - including what turned out to be a wrong turn | No |
| Cause | The actual cause, not the replaced part (the valve was the symptom, the cause was contamination in the tank) | Rarely |
| Action and parts | What was done, which parts were fitted, including equivalents | Yes |
| Outcome | Is the machine running? Temporarily or permanently? What is still open? | Partly |
| Next time | One sentence: what the technician would have checked first with today’s knowledge | No |
| Solvable remotely? | Could the hotline have solved the case by phone with this information? | No |
No ERP service module asks for the three highlighted fields by default. The first two hold the experience; the third provides, as a by-product, the data for the question of which visits could be resolved without a trip in future.
Spoken, this takes five minutes. The language model sorts what was said into the eight fields, the technician confirms or corrects on the tablet. Arithmetically, a team of 50 technicians with one visit a day captures around 3,000 cases with context, diagnostic path and cause in three months - more than a wiki project collects in three years.
5.1 The knowledge roadmap for the last years before departure
For technicians whose retirement is in sight, a second instrument is added. The KOFA study 4/2026 names the means: “mentoring, tandem models, structured handovers or joint project work”. In a service organisation that means:
- Case reviews on old tickets. One hour a week: a younger colleague pulls five closed cases from the history that the experienced technician solved back then and lets him tell the story - with the ticket on screen. The ticket is the trigger the handover meeting lacks. What he tells is filed as an addition to the old case.
- Tandem visits with reversed roles. The younger one diagnoses, the experienced one only steps in when it goes wrong - and then explains what told him so. The service report of this visit contains both perspectives.
- The call after retirement. A consultancy contract for a few hours a month during the first one or two years after departure is not an admission of failure - provided every one of those calls is captured as a case and not just made.
6. When systematic knowledge retention is not worth it
Three situations in which the set-up described is not the right first investment:
- The cases do not repeat. Whoever builds only special-purpose machinery whose faults never resemble each other secures little with case capture: the value of a knowledge base scales with the repetition of similar cases. Here it is more worthwhile to secure design and customer knowledge - a different task with different means.
- The documentation itself is missing. If circuit diagrams and instructions for the installed base exist only on paper or not at all, experience has no anchor. Then digitising the documentation is the first step - and the Machinery Regulation sets a deadline for that anyway.
- Wrong turns must not be admitted. The most valuable field in the service case is the diagnostic path including failed attempts. Where a documented wrong turn is read as the technician’s mistake, this field stays empty and the system only learns what is in the manual anyway. Sorting out the error culture is then more important than any software.
And a fourth, more uncomfortable one: the job cuts the VDMA reports for the first half of 2026 make knowledge retention more urgent and at the same time less likely - whoever does not refill positions has no successor to hand anything over to, and so no trigger to act. Anyone in this situation should at least draw the installed-base map from section 7: one hour of work that shows which customers nobody will be able to serve in two years.
7. First steps: the starting point for your company
Four steps before any software, a few days of work in total:
- Pull the age structure of the service team: who reaches retirement age within five years, who can leave earlier? HR has this list within an hour.
- Draw the installed-base map: which series, customers and machines have ended up with a single person in the visit history of the last two years? This list is your risk list - one hour of work, the most important hour of the whole programme.
- Assess the ticket history honestly: take twenty closed cases and check whether a colleague could reconstruct the diagnostic path from them. The answer is usually no - that is the baseline.
- Start with one machine series and one experienced technician: case capture in eight fields, weekly case review, three months. After that you know whether the technicians take part and whether the cases have substance - neither can be decided on slides.
8. Frequently asked questions
How much knowledge is really lost when an experienced service technician retires?
It cannot be measured in percent, but it can be seen in symptoms: cases only one person could solve, customers who accept only one contact, machine series whose quirks are in no document. To find out for your own company, pull the ticket history: what share of the hard cases of the last two years ended up with the same two or three people?
Isn't it enough to have the technician write everything down before retiring?
No. Experience is retrieved by situations, not by the request to write it down. A technician who has repaired machines for 30 years does not know what he knows until a machine is in front of him. Knowledge retention therefore only works through real cases: structured service reports, joint case reviews on old tickets, tandem visits.
What role does AI play in knowledge retention in field service?
Two roles, both unspectacular: it lowers the effort of capture, because a technician can talk for five minutes after a visit instead of typing for twenty. And it makes the collected cases findable when the colleague who experienced them is no longer there - with a reference to the original report. What AI cannot do: create knowledge that nobody captured.
When should knowledge retention start if a technician is going to retire?
Not at the retirement date but at the last hard case this person solved - and that is usually years before the farewell. In practice that means capturing service cases becomes the standard for everyone, not a special programme for those leaving. Whoever starts only in the last year can still set up case reviews and tandem visits; whoever starts only in the last month should think about a consultancy contract after retirement.
9. Sources
Studies by vendors with a commercial interest are marked as such. Most sources are German-language publications; they are the actual evidence and are linked as such.
- German Federal Statistical Office (Destatis), press release No. N039 of 23 June 2026 (first results of the 2025 microcensus): labour force members reaching retirement age by 2040; age structure. In German.
- VDMA (German Mechanical and Plant Engineering Association), press release “Stellenabbau im Maschinenbau in Deutschland” of 14 August 2026 (republished at verbaende.com): share of employees aged 55 and over, job cuts in the first half of 2026. In German.
- German Economic Institute (IW) for the VDMA’s Impuls Foundation, press release of 11 November 2024: 296,000 retirements and 118,000 newcomers in mechanical engineering by 2034 (simplified projection). In German.
- KOFA / IW, KOFA study 4/2026 “Wertvolle Erfahrung sichern” (6 July 2026): cohort sizes, instruments of knowledge transfer. In German.
- Association of German Chambers of Commerce and Industry (DIHK), DIHK-Fachkräftereport 2025/2026 (December 2025, more than 21,000 company responses): expected loss of company-specific knowledge by sector. In German.
- Federal Employment Agency, Fachkräfteengpassanalyse 2025 (skilled-labour shortage analysis, May 2026; vacancy times in the tables annex): shortage occupations, median vacancy time of technical service staff. In German.
- Michael Polanyi, The Tacit Dimension (1966): “We can know more than we can tell.”
- kothes, KVD, KVA and SKDV, Insight-Report Service 2022 (survey April to July 2022, number of participants not stated; kothes is a technical documentation service provider): own notes (summary 83, detailed evaluation 89.5 percent), their reuse, search problems, missing instructions. In German.
- Aquant, 2026 Field Service KPI Benchmark Report (19 February 2026; vendor study by an AI provider for field service, 161 service organisations, about 30 million service events): skills gap in first-time fix rate.
- DGQ and GfWM, Wissensmanagement in der Norm ISO 9001:2015 (January 2016, CC BY-ND 4.0): wording of clause 7.1.6 and Annex A.7. The standard itself: DIN EN ISO 9001:2015-11 (paid). In German.
- DQS, ISO 9001:2026 published (16 September 2026): the new edition and the validity of ISO 9001:2015 certificates.
Experience cannot be asked for, only collected - on the case that brings it out. Whoever starts today to capture every service visit in five minutes so that context, diagnostic path and cause are in it will have, in the year the best technician leaves, not a handover meeting but two thousand cases in his handwriting. And whoever finds along the way that documentation or error culture come first has, in our view, started just as correctly.
