Calvin's Updates

The Latchkey Club Daily Draft — 2026-07-30

Teleprompter / Blog Script

I used AI to finish a work task recently, and it did exactly what I wanted.

It organized the information, produced a clean first version, and probably saved me a couple of hours.

Then I had a less comfortable thought.

That task was also the kind of task somebody younger might have learned from.

Welcome back to the channel, guys.

Today I wanted to talk about one of the stranger problems AI may be creating at work.

It can make experienced people more productive. But if it removes too much beginner work, how does the beginner ever become experienced?

This is not an anti-AI video. I use AI constantly. At 57, it has helped me turn ideas and work experience into tools I could not have built nearly as quickly on my own.

I can describe a process, give an agent some examples, let it draft the documentation or code, and then use my judgment to correct what it missed.

That feels like leverage because I already know what wrong looks like.

But that last part matters.

I know what wrong looks like because I spent years doing things slowly, making mistakes, sitting through difficult conversations, watching projects fail, and learning which small detail becomes a large problem three weeks later.

AI did not give me that judgment. It arrived after the judgment was already there.

A Dallas Fed analysis from earlier this year made a useful distinction between codified knowledge and tacit knowledge.

Codified knowledge is the part you can put in a book, procedure, or training course. Tacit knowledge is what you pick up through experience. It is knowing when the normal rule does not fit this situation. It is hearing a customer ask one question and recognizing that the real problem is somewhere else.

The analysis found early evidence that AI may replace some codified tasks while making experienced judgment more valuable.

But the report also points to the problem underneath it. Entry-level workers used to do those codified tasks while slowly gaining the tacit knowledge.

The beginner work was not always exciting. It may have involved building the first draft, checking the records, sitting in the meeting, making the routine change, or following the checklist while somebody experienced watched.

But that was the entrance ramp.

If AI does all of it now, the company may get faster output today and fewer experienced people later.

I think this matters to Gen X because a lot of us are close enough to retirement to see both sides.

We want the tool to reduce the load. We have spent decades doing the repetitive part. If software can build the first draft in five minutes, I am not going to insist on doing it manually just to preserve the historical suffering.

At the same time, many of us are now responsible for people who still need the repetitions we already had.

I have been thinking about what I need to hand off before I eventually leave work. At first, knowledge transfer can sound like a documentation project.

Write down the process. Record a video. Put the files in the right folder. Make sure somebody has the passwords and knows which system holds what.

All of that is necessary.

It is also not enough.

A document can explain the normal process. It cannot fully explain why I ignored the normal process last Tuesday.

It may say, “Ask these discovery questions.” It does not automatically teach somebody how to notice that the answer does not match the physical situation.

It can list the customer contacts. It cannot transfer a relationship.

That only happens when another person gets into the work early enough to watch, ask questions, make a decision, and live with the result.

A PBS report on knowledge leaving with retiring workers described companies using job shadowing, mentoring, phased retirement, and bringing successors into customer and vendor meetings long before the retirement date.

That makes sense to me. The handoff has to happen while the experienced person is still around and while the work is still real.

Not in a two-hour download on the Friday before the retirement lunch.

AI can help with this, but I think we need to give it the right job.

It can help capture what happened after a project. What decision did we make? What alternatives did we reject? Which assumption turned out to be wrong? What would we check sooner next time?

It can turn a messy conversation into a first draft of a case study. It can compare several projects and notice recurring patterns. It can build a searchable library so the next person does not have to remember which folder contains the one useful document from 2021.

It can also act like a practice partner.

Give a younger employee a real scenario with some details removed. Ask what they would do next. Let AI challenge the answer with questions, but have an experienced person review the reasoning.

The goal is not for the machine to announce the correct answer.

The goal is to make the person explain the decision.

Why this option? What could go wrong? What information is still missing? Who will be affected if the assumption is wrong?

Those questions are where experience starts growing.

I also need to be careful not to use AI in a way that makes me look indispensable.

That can happen quietly. I become the person with the powerful tools. I finish more work. Everybody is impressed with the output, but nobody else understands how the system works or why the decisions were made.

That is productivity, but it is not succession.

If the process only works while I am standing next to it, I have built a dependency with better software.

The real test is whether another person can carry the responsibility without copying my personality.

I do not need to make three versions of myself. That would be unfair to them, and probably exhausting for everybody else.

They may approach customers differently. They may use newer tools. They may notice things I miss. The point is not to preserve my preferred method forever.

The point is to give them enough context, practice, and authority to make sound decisions when I am not in the room.

That means letting them do some work I could complete faster myself.

That may be the hardest part.

When you have experience and AI, it is very easy to take the task back. I can fix it in twenty minutes. Explaining it may take an hour. Letting somebody work through it may take longer and include a mistake I can already see coming.

But if I remove every struggle, I may remove the lesson too.

So maybe the better question is not, “What can AI do for me?”

Maybe it is, “What should AI remove, and what does a person still need to practice?”

Let AI handle the formatting, the search, the first-pass summary, and the repetitive comparison.

Let the person interview the customer, inspect the situation, defend the recommendation, and explain what they learned afterward.

Then increase the responsibility.

Watch one. Do one together. Lead one while I observe. Teach the next person.

That is slower than pressing a button. It is also how a team survives after the experienced people leave.

At 57, I am starting to think that the value of experience is not just what it lets me finish today.

It is whether that experience can keep helping people when I am gone.

AI can preserve information. It can organize examples. It can help us see patterns.

But it cannot make me invest in another person. It cannot decide that their growth matters more than my speed or my need to feel necessary.

That part is still mine.

Anyway, that’s what I’ve been thinking about.

If AI is doing more of the beginner work where you are, how are younger people supposed to gain judgment? And if you are getting closer to retirement, what do you know that still lives mostly in your head?

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: What did AI finish quickly that made you wonder whether somebody else had lost a learning opportunity?
   • Follow-up: Why did that feel different from simply saving time?
2. The experience gap: What do you know now only because you did slow, repetitive, or uncomfortable work earlier in your career?
   • Follow-up: What does “knowing what wrong looks like” mean in your work?
3. Two kinds of knowledge: What is the difference between information that fits in a procedure and judgment learned through consequences?
   • Follow-up: Give a safe example of when the normal rule did not fit the real situation.
4. The entry ramp: Which beginner tasks quietly helped people gain context before AI could do them?
   • Follow-up: What happens if companies remove those tasks without replacing the learning path?
5. Your handoff: What do you want your team to be able to do without you before you retire?
   • Follow-up: Why are documents, recordings, and organized folders necessary but insufficient?
6. A better role for AI: How can AI capture decisions, create case studies, organize lessons, and help people practice scenarios?
   • Follow-up: Which questions should force the human to explain the reasoning?
7. The leadership tension: When is doing the task faster yourself actually a failure to develop somebody else?
   • Follow-up: How much struggle should remain so the lesson remains?
8. Success without clones: Why should successors become independent leaders rather than copies of the retiring person?
   • Follow-up: What responsibility can you hand over next?
9. Close: What knowledge still lives mainly in your head, and who needs the chance to use it while you are still available?

Title Options

1. If AI Does the Beginner Work, How Does Anybody Become Experienced?
2. Before Gen X Retires, We Need to Teach More Than the Procedure
3. AI Can Save the Work—and Break the Apprenticeship

Thumbnail / Onscreen Text Options

• WHO TRAINS THE NEXT EXPERT?
• AI REMOVED THE ENTRY RAMP
• DON’T JUST SAVE THE FILES

Shorts / Reels Cutdowns

• “AI arrived after the judgment.” Contrast AI’s speed with the years of mistakes, consequences, and customer conversations that taught an experienced worker what wrong looks like.
• “A document cannot transfer a relationship.” Use the section on procedures versus tacit knowledge, then land on shadowing and real responsibility.
• “Productivity is not succession.” Cut the warning about becoming the only person with powerful tools and building a dependency with better software.

Viewer Question

If AI is now doing the entry-level tasks in your field, what experience should employers deliberately create so the next generation can still develop judgment?