Calvin's Updates

The Latchkey Club Daily Draft — August 10, 2026

Teleprompter / Blog Script

I was looking at a set of instructions recently, and on paper they were correct.

Every step was there. The order made sense. Somebody could follow the document and probably complete the job most of the time.

But I also knew there was one situation where following those instructions exactly would create a mess.

That exception wasn’t written down anywhere. It was just living in somebody’s head.

Welcome back to the channel, guys.

Today I wanted to talk about something I think matters for a lot of Gen X people who are somewhere between experienced at work and thinking seriously about retirement.

The checklist isn’t the job.

The checklist is the part of the job that was easy enough to explain.

The real job also includes knowing when the checklist no longer fits, which warning can be ignored, which one cannot, and who needs a phone call before somebody makes a change that looks harmless on a screen.

That kind of knowledge accumulates slowly. Usually after a few mistakes.

You remember the customer who had an unusual setup. You know why an old rule still exists even though the original system is gone. You know that two things with different names are actually connected, and changing one will affect the other three days later.

None of that looks impressive in a procedure.

It looks like a sentence that says, “Before doing this, check with somebody.”

But knowing who that somebody is, and what they are checking for, may be the difference between a smooth day and a very long evening.

I have been thinking about this because companies are trying to put AI into more work, while a lot of experienced people are getting closer to the exit.

A recent report from the Center for Retirement Research looked at workers 55 and older. It found that people our age are just as exposed to generative AI in their occupations as mid-career workers. It also found an increase in job exits among older workers in highly exposed jobs after ChatGPT arrived.

The researchers were careful. That does not prove AI caused each person to leave. Some jobs were already changing. Some people may have retired by choice. Some may have been displaced. And for other workers, AI could make the job easier and help them stay longer.

But it raises a question I do not hear enough.

When an experienced person leaves during an AI transition, what leaves with them?

The obvious answer is years of knowledge. But I think the more specific answer is exceptions.

An AI system can summarize a manual. It can turn notes into a cleaner procedure. It can draft a checklist, compare records, and find patterns across a pile of old documents much faster than I can.

That is useful.

But if the documents only describe the normal path, the system becomes very good at repeating the normal path.

It does not automatically know about the time the normal path failed.

It may not know that a customer once lost service because two databases disagreed. It may not know that a field instruction is technically accurate but impossible to perform safely in the actual location. It may not know that the number on the dashboard looks fine because the dashboard is averaging away the problem.

A person who has been around for twenty or thirty years may see that immediately.

Not because older people are always wiser. We are capable of being confidently wrong with a lot of experience behind us.

But consequences build pattern recognition. You have seen enough ordinary days to notice when today is not ordinary.

That may be one of our most useful contributions during this stage of work.

Not proving that we can click every new button faster than somebody who is 28.

Not guarding information so the company cannot function without us.

And not standing back and waiting for the new system to fail so we can say we knew it would.

The useful move is to help put the exceptions into the handoff.

I think that starts with a different kind of documentation.

Most procedures ask, “What are the steps when this works?”

We also need to ask, “How do I know when these steps do not apply?”

What are the warning signs?

Which assumptions must be true before we begin?

What mistake looks small at first but becomes expensive later?

Who has enough context to approve the exception?

And after the work is done, what should we verify instead of assuming the green check mark means everything is fine?

Those questions get closer to the job.

AI can actually help with this if we use it carefully. Instead of asking it to replace the experienced person, we can use it to interview the experienced person.

Give it an old incident report and ask what condition made this case different.

Compare five jobs that went well with two that went badly.

Turn a conversation into a draft decision tree. Then have the team challenge it with real examples.

Ask the newest person what is still unclear, because experts are often terrible at noticing the steps they skipped in their own explanation.

That last part matters. I have done work long enough to think I explained something from the beginning, only to realize I started at chapter six.

The goal is not to preserve every habit just because an experienced person developed it. Some workarounds are only workarounds. Some old rules need to disappear. Some exceptions are no longer exceptions because the system has changed.

So the handoff has to include a little humility.

Why do I do it this way?

Is the risk still real?

Can I show an example?

Would the person taking over reach the same conclusion with the information available now?

If I cannot explain the reasoning, I may be passing down a superstition instead of judgment.

There is also a boundary here. AI can help preserve context, but responsibility still has to land with people.

A decision tree does not own the outcome. A chatbot does not mentor the new employee. A database does not notice that somebody is afraid to ask a basic question.

The point of capturing knowledge is not to build a digital version of the person who is retiring.

It is to give the next person a better starting point, then stay around long enough to let them use it, question it, and improve it.

I think that changes the way we look at our final years at work.

Maybe the measure is not how many problems still have to come through me.

Maybe it is how many problems can be handled well when I am not there, without pretending that judgment can be reduced to a hundred-page manual nobody reads.

That is a strange adjustment because being needed can feel a lot like being valuable.

If everybody knows how to do the thing, part of me may wonder what I am still there for.

But holding the answer is not the same as building people.

And leaving behind a pile of files is not the same as making a handoff.

So I am trying to pay more attention to the moments when I say, “Normally we do this, but in this case…”

That sentence is probably worth capturing.

Not because every exception belongs in an AI system. Some information is sensitive, some decisions need direct oversight, and some situations are too unusual to automate.

But the reasoning should not disappear simply because the person carrying it finally takes a day off, changes jobs, or retires.

Anyway, that is what I have been thinking about.

What is one thing you know at work that is not in the manual, and who are you teaching it to?

Leave me a note in the comments. Thanks for listening.

Video Prompt Script — Questions to Answer Without Reading

Use these as prompts. Don’t read them on camera; answer them naturally.

1. Opening: When have you seen instructions that were technically correct but wrong for one important situation?
   • Follow-up: Where was the missing exception stored?
2. The distinction: What is the difference between the checklist and the actual job?
   • Follow-up: What kinds of judgment accumulate only after seeing consequences?
3. The retirement and AI tension: What did the Center for Retirement Research find about AI exposure and job exits among workers 55+?
   • Follow-up: Why should you avoid claiming that AI caused every exit?
4. What leaves: When an experienced person retires during a technology change, why are the exceptions often the most important loss?
   • Follow-up: What can an AI learn from a manual, and what may be absent from the manual?
5. Document the failure conditions: How do you know when the normal procedure does not apply?
   • Follow-up: What warning signs, assumptions, downstream risks, approvals, and verification steps should be captured?
6. Use AI as an interviewer: How could AI compare incident reports, organize a conversation, or draft a decision tree without being given final authority?
   • Follow-up: Why should a newer employee challenge the draft with real cases?
7. Avoid preserving superstition: How do you test whether an old rule still protects against a real risk?
   • Follow-up: Can you explain the reason and show an example?
8. People still own the handoff: Why can a chatbot organize knowledge but not replace mentoring, accountability, or psychological safety?
9. Close: Is your value measured by how many problems require you, or by how well people can handle them after you leave?
   • Follow-up: What “normally we do this, but…” sentence should you capture this week?

Title Options

1. The Checklist Isn’t the Job
2. What Retires When an Experienced Worker Leaves?
3. Before Gen X Leaves Work, We Need to Pass On the Exceptions

Thumbnail / Onscreen Text Options

• THE CHECKLIST ISN’T THE JOB
• WHO KNOWS THE EXCEPTIONS?
• DON’T LET THE KNOWLEDGE RETIRE

Shorts / Reels Cutdowns

• “The exception lives in somebody’s head” — use the opening and explain why technically correct instructions can still create a mess.
• “Use AI to interview experience” — show how old incident reports and real conversations can become a tested decision tree without giving AI final authority.
• “Being needed versus building people” — cut the section on measuring value by how many problems still require you, then land on a real handoff.

Viewer Question

What is one thing you know at work that is not in the manual, and who are you teaching it to?