#43 | The 3 words that make you replaceable

TL;DR: “AI says, I/we …” is killing careers. Why your judgment beats your prompting skills.

👋 Hey there,

Recently, one of my friends was planning his first solo trip to Italy, so I asked him what he was most excited to see.

“Well, ChatGPT told me to visit the Colosseum and Vatican Museums, plus budget €100 per day and book trains in advance. Gemini suggested I see the Uffizi in Florence and learn basic Italian phrases. And Claude thinks I should do a day trip to Venice, pack light with just a carry-on, and stay in city centers.”

I waited for him to mention something—anything—that he actually wanted to experience.

Nothing came.

For some reason, he had outsourced his curiosity about the trip entirely. When I asked about his interests or what drew him to Italy specifically, he said he’d “ask AI to decide.”

He’d become a concierge for algorithmic wanderlust—because apparently asking machines where to find wonder is the peak of human evolution. (I guess that is the future of travelling.)

That conversation taught me something crucial about how we’re accidentally training ourselves out of independent thinking—and why those three specific words, “AI says, I/we…” have become the fastest way to signal you’re replaceable.

A recent newsletter from ​Greg & Taylor at Personal Math​ also hit on this exact issue from a workplace angle, and shows how common this pattern is.

The phrase that kills conversations

Walk into any discussion this week and listen for those three career-limiting words: “AI says, I/we…”

The energy shifts immediately.

People’s attention drifts because you’ve just announced you’re unnecessary—positioned yourself as ChatGPT’s messenger rather than someone with actual thoughts about the situation. (Psst! ​ChatGPT 5 is out.​)

Nobody needs a human to repeat what AI thinks, do we?

We usually acknowledge someone who can process information, add context, and form judgments that matter for their specific circumstances.

Ok! How to spot outsourced thinking (including your own)

Let’s imagine a typical business situation.

Emma needed to analyze her competitors before a client meeting.

In the past, she would’ve spent time studying their websites, comparing approaches, and forming hypotheses based on her market experience.

This time? Claude produced a thorough, well-organized report quickly, with excellent layout, formatting, and customized results for that industry.

The real problem showed up when her client asked, “Why do you think they’re launching now?”

Emma struggled because she hadn’t thought through the business dynamics of specific competitors that would’ve prepared her for that reasonable question.

AI had delivered what you might call the 80% solution—clean and presentable.

But that final 20%, where context meets intuition, was exactly where Emma’s irreplaceable value lived.

The moment your bluff gets called.

Here’s the brutal test for whether you’ve outsourced your thinking: someone asks you to defend your recommendation, and you freeze.

Liam faced this during a product review.

He’d used AI to prioritize features, presented the recommendations confidently, and felt great until someone asked about the trade-offs behind his choices.

He had no real answer because he’d skipped the messy work of actually thinking through the problem.

When pressed on why Feature A mattered more than Feature B, he couldn’t explain the reasoning beyond repeating those same three words: “AI says it’s…”

If you can’t defend your logic when challenged, you probably don’t have logic—just AI output you’ve agreed to repeat.

The process that keeps you in the driver’s seat

Here’s what works when you need AI’s help without losing your edge:

1. Start with your own thinking.

Before asking AI anything, sit with the problem. Write down your hypothesis, even if it’s rough.

What patterns do you notice? What constraints matter? What outcomes would actually move the needle?

This prevents intellectual laziness and gives you something real to build on.

2. Use AI to pressure-test and expand.

Ask it to challenge your assumptions: “What am I missing here?” or “What could go wrong with this approach?”

Request alternative frameworks or ask it to argue against your initial thinking.

But treat it like a research assistant who’s helping you think more thoroughly, never as the decision-maker.

3. Watch for the warning signs.

If you find yourself copying and pasting AI responses directly, or if you can’t explain the reasoning behind recommendations without starting sentences with “AI says,” you’ve crossed into dangerous territory.

Good collaboration leaves you more informed about the problem, not less involved in solving it.

4. Form your updated judgment.

Your final recommendation might shift based on what AI surfaced, but it should still feel distinctly yours.

You should be able to walk someone through your logic without mentioning AI at all, because the thinking process was yours, just enhanced by better information.

5. Own your conclusions completely.

Present ideas as your analysis. When someone asks follow-up questions, respond with confidence because you understand the reasoning.

If you get caught having over-relied on AI and can’t defend your work, own it quickly.

Don’t blame the machine. Just say you rushed the analysis and will do better next time. You get one free pass for this mistake—use it wisely.

Test your conviction. Before presenting anything important, ask yourself: “If my boss pushes back hard on this, can I engage in productive dialogue?” If the answer is no, you need to do more work.

Why this matters more than you think

AI literacy will be universal within two years, which means the advantage is shifting from “I can use ChatGPT effectively” to “I can think clearly with AI as my research partner.”

The people who thrive won’t be the fastest prompters.

They’ll be the ones who add context, form sound judgment, and stand behind their recommendations when it matters.

Your brain stays in the driver’s seat. AI just helps you see around corners and think more thoroughly about problems you’re already equipped to solve.

Jake eventually went to Italy with his AI-generated itinerary. He had a fine time visiting the standard tourist spots, but when I asked what surprised him most about the trip, he said going to cafè bars (fine by me).

Happy Friday,

Mark
The AI Learning Guy
👋⚡😎

Sources and books

1. ​OpenAI Study Mode Press Release​
2. ​OpenAI Study Mode – YouTube​
3. ​Open AI ChatGPT 5​

Note: No single website has all the answers. This list serves as a starting point for those who want to explore or satisfy their curiosity about AI. Links: Links with * are affiliate links. See disclosure below.