How do you keep institutional knowledge when senior techs retire

Your best commercial tech gives notice on a Tuesday. Thirty years in the trade, most of them with you. Good notice, no drama, nothing to fix.

Then you start doing arithmetic that has nothing to do with headcount. There is a campus across town, three buildings under one management company, and a rooftop unit on the middle building that has come back from a hard freeze twice, both times because that tech knew the sequence. They also know something written down nowhere: the failing equipment on that roof belongs to the second building, not the third, whatever the account record says. Billing has gotten it wrong before.

None of that is in your system. It has never been in your system. It has been in one person, and soon it drives away.

This isn't anecdotal. More than one in five construction workers is now over 55, and industry researchers describe a skilled-trades workforce aging out faster than new people are replacing it.

What actually leaves when a senior tech retires?

Not "experience." That word is too soft to act on. Say what it actually is.

The freeze sequence on that rooftop. Which building really owns the unit. That the property manager at the industrial park will approve anything in October and nothing after the budget closes, and why. That the compressor on the north side has been drawing high for two seasons and everyone has been quietly watching it. That a customer was told years ago they would need a full changeout eventually, and has been setting money aside ever since.

Every one of those is a fact about a property or a customer. Specific, checkable, worth money. Not one of them is written down anywhere.

Why doesn't the knowledge transfer before they go?

Because every attempt happens in the wrong window.

The documentation project asks them to write down what they know. You get a page or two of generalities and then it stops, because the request is impossible as posed. Nobody can inventory their own expertise on demand.

The exit interview gets war stories. The rooftop that caught fire, the customer who screamed. War stories are the least useful thing they carry, and they surface first precisely because they were unusual. The valuable knowledge is ordinary, which is exactly why it does not come to mind in a conference room.

The last-month ride-along transfers whatever the dispatch board happens to send them to. If the campus does not call that month, the campus knowledge does not move.

All of it fails the same way. Each asks for recall outside the context that produces it. Your senior tech does not know the freeze sequence sitting at a table. They know it on that roof, with a hand on the unit, looking at what is in front of them. Take the context away and the knowledge does not come when called.

Why does it live in one head to begin with?

Because nothing in the working day ever asked for it.

Follow one visit through. They arrive, diagnose, fix it, and close the ticket. "Replaced blower motor, unit operating." That is a record of what was done. There is no field for what was understood. Which building owns the equipment, why October matters, what they have been watching for two seasons: none of it has anywhere to go, so it goes where everything without a home goes, which is into the person.

Then that repeats for a career. You do not end up with an archive. You end up with an expert, and an expert is a single point of failure with a retirement date attached.

What kind of knowledge can actually be captured, and what can't?

Split what they know in two, because the two halves behave completely differently.

The first half is property-specific history. Which unit sits on which roof, owned by which building. What was found last spring. What the customer said, and when. What was recommended and never done. This half is factual, attached to an address rather than to a person, and it can outlive whoever observed it. It is a record, and records are keepable.

The second half is craft judgment. Hearing a compressor and knowing it has one season left. Knowing which manufacturer's control boards fail first. Reading a system and understanding what the last person to touch it did wrong. This half is pattern recognition assembled over a career, and it does not survive being written down. You cannot hand someone a document that makes them able to hear it.

Be direct about that, because the market is not. Any vendor who tells you they can bottle the second half is selling you something. What they deliver is a document nobody opens and a junior tech who is no better on the roof than they were the week before.

Naming that limit is what makes the first half credible. Property history genuinely can be kept. Craft judgment genuinely cannot. The moment you stop trying to solve both with one purchase, the first turns into a problem you can actually solve.

Why does the record have to be a byproduct of the work?

Because you never know which visit mattered until years after it happened.

Any deliberate record requires somebody to decide, today, that a thing is worth recording. Nobody makes that call well. The freeze sequence did not look important the first time. It became important the third time, on a Sunday, when the person who knew it was out of town.

A record built on purpose is a record of what someone thought was significant at the time. A record that accrues from ordinary work is a record of what happened. Only the second kind is reliably there when you finally need it, because nobody had to be wise enough to choose it.

That is the whole requirement, and it is worth being strict about. Whatever you use has to leave the record behind as a side effect of work that was already going to happen. Anything needing a separate step, a separate decision, or a separate hour loses to the calendar, and you find out during the exit interview.

What do you do about the retirement you can already see coming?

For the one already in motion, honestly, not much.

If your senior tech leaves next quarter, no purchase fixes that. Nothing captures a career during a notice period. Buy something this week and you will still lose most of what they know, because the record you need is one that had to be accumulating while they worked. Anyone telling you otherwise is selling to your panic, and you should hear it that way.

So that is not the decision in front of you. The decision is about the next one.

You have other techs. Some have been at this a while. Some are learning your accounts right now, and in time someone will sit where you are sitting and do this same arithmetic about them. The real question is whether everything your people learn between now and then accumulates anywhere, or whether it collects in heads again and leaves the same way it is leaving now.

That is a harder question than the one you walked in with. It is also the only one you can still act on.

Where Mentat fits

Mentat is field service revenue recovery software. It rides on top of the field service system a company already runs, and it is never the system of record: not a field service management system, not a CRM, not a replacement for anything you have. A technician records a short voice memo about what they saw. Mentat structures that memo, prices the work, and drafts the estimate straight back into that same system, where a person on your team reviews it and sends it. Every opportunity carries the name of the technician who spotted it. Today the loop runs against Jobber.

The part that matters for this question is a side effect. Every memo is dated and attributed, and it accrues to the property, so what one tech observed becomes history the next one inherits. That happens as a byproduct of capturing work worth quoting, not as a documentation program somebody has to run and keep running.

And the limit, plainly. Mentat is not a knowledge management system and it does not train anyone. It will not teach a junior tech to hear a failing compressor. It keeps the record. The craft is still yours to pass on.

FAQ

How do you keep institutional knowledge when senior techs retire? You cannot keep all of it, and the part you can keep has to have been accumulating before anyone gave notice. Property-specific history, meaning what was found at an address, when, and by whom, can live in a record. The craft judgment a long career produces cannot. The practical move is making sure ordinary visits leave a record behind, starting well before a retirement is in sight.

Can you document what a senior tech knows before they leave? Not usefully. Asking someone to inventory their own expertise on demand produces generalities, and an exit interview surfaces the memorable stories rather than the useful details. Most of what they know only comes back to them in context, standing in front of the equipment.

What kind of knowledge can actually be captured? The factual, property-attached kind. Which unit is on which roof, what was found last spring, what the customer said, what is still open. Pattern recognition built over a career is a different thing, and it does not survive being written down.

Does a shadowing or mentorship program fix this? It helps with craft judgment and barely touches property history. A ride-along only covers the calls that come in while it is running, so whether the knowledge about a particular account transfers comes down to whether that account happens to call that month.

Is this a knowledge management system? No. Mentat is field service revenue recovery software. The record it builds is a byproduct of capturing work worth quoting rather than a documentation product, and it does not train technicians.