Prompt It

Lately I’ve noticed something about myself: before asking someone who actually knows about a topic, I ask AI first. Things I would have messaged a friend about a few years ago now go to ChatGPT. Sometimes the answer is good enough, and I never ask the person at all.

Google caused a similar shift. At some point, “Google it” became the default response to simple questions. Instead of asking someone, you were expected to look it up yourself. But Google still required some effort. You had to search, open a few links, compare answers, and decide what made sense. With AI, you just ask. If you don’t understand, you ask again. Most of the time, a few minutes is enough to get an answer you’re satisfied with.

That’s where I think the real difference starts. AI isn’t only taking questions away from Google. It’s also taking questions away from people.

This gets more interesting at work. In an office, if you didn’t understand something, you might ask the person sitting next to you. In remote work, you’d post it in Slack. Sometimes that small question would turn into a longer conversation, and you would learn something you weren’t even asking about. Now it’s much easier to paste the code, error, or concept into a model and keep working.

That’s not necessarily bad. Most of the time it’s useful. You interrupt people less, you can learn the basics on your own, and when you do ask someone, your question is usually better. But there’s also a side effect: some conversations simply never happen because the AI answer is good enough.

Remote work probably makes this even stronger. You already have to make a conscious decision to open Slack and message someone. AI is sitting inside your browser or editor. The difference is only a few seconds, but over hundreds of small questions, that probably matters.

Now imagine the next step. The company AI knows the codebase, documentation, old Slack threads, meeting notes, and pull requests. Today, for a question like “why did we build this this way?”, you may still need the person who remembers the history. In the future, the model might know that too.

“Google it” was the habit of one era: if you don’t know something, search the internet first.

Maybe the next one is “prompt it.”

If you don’t know something, ask AI first. Ask a person only if that isn’t enough.

What I’m curious about is how often that second step will still happen.