It’s Friday afternoon.
There’s a tray of sausage rolls in the canteen, a colourful card doing the rounds, and a grateful speech from the plant manager.
After 32 years, Mick is retiring.
‘It won’t be the same without him,’ they say.
They’re more right than they know.
Monday Night, 2am
Line 3 trips. The HMI throws up a fault nobody on nights has seen before.
The shift technician, Liam, (bright, keen, and just two years out of his apprenticeship) does everything right: checks the drawings; resets; checks the CMMS history; and shrugs.
‘Fault cleared. MB.’
That’s it. That’s the CMMS history.
Three hours later, he rings the shift supervisor at home.
‘Sorry to wake you. Line 3’s down. Didn’t Mick used to sort this? Something about the reset valve?’
‘Yeah… he used to do this thing with it. Never did say what.’
Mick, meanwhile, is on a cruise ship to Greece, sleeping off a buffet and 5 pints.
The Knowledge Iceberg
Every plant has two memories.
The first lives in your systems: the asset register, PM schedules, drawings, spares lists, work order history, small notebooks. It’s visible, somewhat searchable and – being honest – incomplete.
The second lives in people’s heads. And it’s much, much bigger.
It’s knowing that the Line 3 Filler always runs hot after a changeover. That the bearings from one particular supplier never last. That the compressor PM is every six weeks, not eight, because of a random failure ten years past that cost two days of production. That you spray WD-40 on the reset valve stem before moving it.
None of that is written down. It’s the part of the iceberg below the waterline.
And, eventually, it walks out of the gate.
Why It’s Getting Worse
Brain drain isn’t a new problem. But it is accelerating, and the usual fix of hiring someone experienced – is getting harder.
- The pipeline is thin. In the US, Deloitte and The Manufacturing Institute estimate that manufacturing could need as many as 3.8 million new employees between 2024 and 2033, and that 1.9 million of those jobs could go unfilled.
- It’s not just a US problem. In the IET’s 2025 skills survey, 76% of UK engineering employers said they struggle to recruit for key roles, and only 61% said their current workforce is fit for the future.
- People move on. Knowledge doesn’t only leave with retirements. It leaves with resignations, restructures, and contracts that end.
- More work is outsourced. When the OEM or a contractor does the job, the know-how often leaves in their van.
So you can’t simply hire your way out of this. There may not be anyone to hire.
This means the knowledge you have right now is more valuable than it’s ever been.
And the people on the front line know it. In a 2026 MAINSTREAM survey of 715 maintenance and reliability professionals in Australia and New Zealand, 35.8% named capturing and transferring expert knowledge as one of the biggest opportunities for AI over the next five years.
This reflects a very real anxiety in the profession about the loss of experienced tradespeople and the tribal knowledge they carry.
What It Really Costs
Lost knowledge doesn’t appear as a line item on the P&L. It hides in other numbers:
- Longer fault finding. Mean time to repair (MTTR) creeps up, and nobody can quite say why.
- Repeat failures. Same failure, different year. The root cause was found once but forgotten.
- PMs nobody trusts. If no one knows why a task exists, it gets deferred, skipped, or kept forever out of fear.
- Spares decisions on gut feel. ‘We always keep four of those.’ Why? Nobody knows.
- Slow onboarding. New starters shadow for months because there’s nothing structured to learn from.
- The 2am phone call.
The Handover Document Myth
The usual response goes something like this:
‘Mick, before you go, can you write down everything you know?’
Two weeks later there’s an 120 page document on the shared drive, in a folder called Mick_Handover_FINAL.
Nobody reads it.
Not because it’s bad. Because it’s disconnected. Knowledge that isn’t attached to the asset, the failure mode and the task is knowledge nobody can find at 2am.
And by the time someone has left, you’re already too late. Knowledge capture has to be continuous, not a leaving present.
‘Data is a process, not a project… We have to recognise that our processes need to support data capture as part of normal operations.’
Capture Knowledge Where It Lives
The goal is simple: make the plant remember what your people know.
Here’s how excellence-focused organisations do it.
- Start with what matters most. A criticality analysis tells you where failure hurts most – safety, production, environment, cost – and that’s where you focus first.
- Run FMECA or RCM workshops with your experienced people, while they’re still here. Mick knows the failure modes. A structured analysis turns that into function, failure mode, effect, cause and task. His 32 years become a defensible, auditable maintenance strategy.
- Make work orders tell a story. Failure codes. Cause. Action taken. ‘Fault cleared’ is not history. Make it easy – mobile devices, pick lists, photos, short videos – or people won’t do it.
- Close the loop with root cause analysis. Every significant failure should leave behind a searchable record of what happened, why, and what changed. That way the lesson is learned once, not every few years.
- Let AI do the heavy lifting. This is where things have moved on. AI can help turn a technician’s voice note into a structured work order entry, summarise years of free-text history on an asset, suggest likely failure modes based on similar equipment, and surface ‘this has happened before’ at the moment it’s needed.
- Treat knowledge as an asset. Give it an owner, a review cycle and a life cycle, just like the physical assets it describes.
One honest word of warning on AI: it can only learn from what you’ve captured. Rubbish in, rubbish out.
Or, as one practitioner put it rather more bluntly:
AI amplifies everything – including your data quality problems… Garbage in, garbage out applies even more to AI than traditional systems.
It’s no surprise, then, that plenty of organisations are trying AI but far fewer are relying on it. Closer to home:
58% of engineering employers currently use AI, but only 18% use it regularly.
Very often, the gap between trying and relying on AI is the quality of the data underneath.
Get the foundations right, though, and AI and experienced people make a powerful partnership. Summing up a mining case study where an AI platform flagged a developing pump failure on a haul truck, the MAINSTREAM report put it neatly: ‘the AI detected the deviation, but the expert diagnosed it.’
That’s the point. Technology doesn’t replace Mick. It makes sure what Mick knows is still there when he isn’t.
Monday Night, 2am. Take Two.
Same plant. Same fault. Same technician.
This time he opens the asset on his tablet. The failure history shows it straight away: reset valve sticking; apply WD-40 to the stem first, with a twenty second video. There’s a link to the RCA from a few years back explaining why it happens, and a note that a permanent fix is planned for the next shutdown.
Line 3 is back up in twenty minutes.
Nobody’s phone rings.
Mick is still on the cruise. But his knowledge stayed behind.
Where to Start
Don’t bite off more than you can chew.
Pick your ten most critical assets. Work out who holds the knowledge about them. Get them in a room with a facilitator and a structured method.
And book it before the leaving do.
At ProAIM, we build reliability and asset performance software – criticality analysis, FMECA, root cause analysis, maintenance strategy and AI-assisted tools – designed to capture exactly this kind of knowledge and put it to work.
If brain drain is keeping you up at night (or your phone ringing at 2am), let’s talk.
