Every operations leader the team interviews names someone like Bill within the first ten minutes. Different name, same function. He has been on the floor for 22 years. He is the one the night shift calls when a press starts behaving strangely. He has no written documentation for any of this. He retires in 18 months. Nobody has a succession plan.
In South Africa, the timing is structurally worse than elsewhere. Thirty-eight percent of SA industrial companies could not find suitable engineers in 2025 (Engineering News, 2025). Bill's replacement is not waiting in the wings. The replacement market is empty.
What Bill actually knows.
The team has spent enough hours on shop floors to recognise the categories. Bill's knowledge sorts into five buckets, in roughly decreasing order of how recoverable it is when he leaves:
- Asset history — which machines have been repaired, when, by whom, and what for. This is the most recoverable, if a CMMS has been maintained. Usually has not.
- Failure pattern recognition — Bill recognises the pre-failure signature of a bearing about to seize the way an experienced GP recognises a presentation. The recognition is real, the explanation is not. He cannot teach it. He can only do it.
- Vendor and supplier shadow knowledge — which suppliers ship out-of-spec parts under a certain order volume; which compressor brand tolerates the dust at this site; which lubrication contract has been quietly under-delivering for three years.
- Process workarounds — the seven things the engineering drawings say to do but nobody actually does because they do not work in this plant for reasons Bill discovered in 2009.
- Relationships — which technician will actually answer his phone at 2am, which buyer at the parts supplier can be persuaded to expedite. The least recoverable category. It walks out with him.
What Foshan taught us about succession.
The Inspectaman team has now audited 200+ Chinese manufacturing operations, with the largest cluster concentrated around Foshan and the Pearl River Delta. Chinese tier-2 suppliers face the same demographic curve as SA operations, with one structural difference: their succession crisis arrived ten years earlier. The plants that survived it did two things.
First, they instrumented before Bill retired. Vibration, temperature, and current sensors on the top-decile critical assets, feeding a historian, with at least 18 months of forward archival before the lead engineer's exit. The point was not to replace Bill with AI. The point was to capture what Bill had been pattern-matching on for two decades, so that the next engineer had a baseline to learn against.
Second, they ran a structured capture exercise with Bill before he left. Not a wiki, not a handover document. A sit-down with a junior engineer, an audio recorder, and a checklist — roughly a hundred questions, asked in front of each major asset in turn. The transcripts went into the same historian as the sensor data. Five years later, when an anomaly surfaces, the current reliability lead can pull both the sensor trend and Bill's recorded reasoning about that exact failure mode.
The cost of one breakdown when Bill is not there.
The team modelled this against a mid-sized SA food processing plant — 50 critical assets, three shifts, R500,000 average preventable failure cost. Pre-Bill-retirement: ~20 preventable failures per year, ~14 averted through Bill's pattern recognition and his network. Annual failure cost: R3M absorbed.
Six months after Bill retires, with no succession capture and no instrumentation: averted rate drops to ~6 of 20. Annual failure cost climbs to R7M absorbed. The R4M delta is what the company is paying for not having an inventory of what Bill knew. This number does not appear in any budget line. It is invisible until the year-on-year reliability report lands.
What to capture before they leave.
The team uses a structured Bill Inventory across audit engagements. The headline framework, with the questions each section is built around:
- Per asset — What is the failure history? What is the pre-failure signature you recognise? Which subassembly fails first? Which one is the false alarm? Which supplier do you trust for spares and why?
- Per process — What does the documented SOP say to do? What do you actually do? What was the incident that taught you the difference? What are the three things the next person needs to know before touching this?
- Per shift handover — What gets handed over that does not appear in the written log? What patterns indicate the night shift was rougher than reported? What signs say the line is about to have a bad day?
- Per relationship — Which vendor reps do you call directly? Which buyers at suppliers can expedite? Which auditor has standards we would not pass on a typical day? Who else in this plant knows what you know?
- Per failure — Walk through the last five significant breakdowns. What did the data say? What did you see? Why was the data wrong if it was wrong? What would you instrument differently?
The exercise takes roughly 40 hours of Bill's time over six weeks, captured by a junior engineer with an audio recorder and a structured worksheet. The output is a searchable transcript indexed by asset, by failure mode, and by date. If the team is honest, this is the single highest-ROI six weeks an operations leader can fund. The recovered knowledge keeps generating value for a decade.
Why Foundry Cortex was designed around this.
Predictive maintenance products usually frame themselves as alternatives to expertise. That is backwards. The systems that work best in the field are the ones that capture expertise into a substrate the next engineer can learn against. The sensor data is the substrate. The model is the codification. The dashboard is the translation layer that lets a five-year engineer reason the way a twenty-five-year engineer reasons.
Foundry Cortex is built explicitly around this thesis. The 60-day pilot starts with a Bill Inventory pass — capturing the senior engineer's pattern knowledge against the assets the sensors are going on. The dashboard surfaces both the model's forecast and the captured human reasoning behind every alert. Twelve months in, the program runs even if Bill retires tomorrow.
The Tribal Knowledge Risk Audit (free, no sign-up) scores how exposed your operation is to a Bill-shaped exit. It also generates the structured capture worksheet, asset-by-asset, so the interview can begin tomorrow.



