Asking a Person Used to Be a Win-Win. Asking AI Only Wins Once.

What's your first move when you get stuck at work these days?

I noticed my own answer had changed. It used to be walking over to a coworker's desk, or firing off a message: "have you run into this before?" Now it's opening AI and dropping the question in, and a few seconds later there's an answer that looks decent enough. I never consciously chose this, it just crept in, until one day I realized it had been a while since I'd actually worked something out together with a coworker.

This is happening in two places at once

Here's a number that stopped me. Engineers used to post a stuck problem to a public technical forum and wait for someone to answer, and those forums were once the most important place to get help. Today the daily volume of new questions on the largest ones is down to about one percent of what it was a decade ago, and similar knowledge forums have gone nearly silent. It's not that engineers got so good they stopped needing to ask. It's that "asking" moved to a place where you don't have to wait on anyone, and nobody sees you doing it.

Companies are making the same call inside their own walls. Word recently came out that Google has dissolved its entire internal Engineering Education team, the group that used to turn veteran engineers' hands-on experience into training material so new hires didn't have to hit every wall themselves from scratch. The reasoning is blunt: people just ask AI now. A department built over twenty years, specifically to package knowledge and pass it to other people, has been judged unnecessary as a dedicated team.

One of these is happening in an office, the other across an entire technical community. But it's actually the same thing: problems that used to require asking a person are now all handed to AI instead.

Asking a person used to be a two-for-one

Here's the part that's easy to miss. Whether you asked a coworker or posted to a forum, that act was never just "getting an answer" for yourself. It usually left something behind for someone else too.

A question posted to a forum, along with the discussion underneath it, stays there. Six months later, a total stranger stuck at the exact same point can search and find it, saving them the trouble. This is also exactly why forum traffic maps so directly onto forum survival: every question used to be a small deposit of knowledge for future strangers. Once everyone routes those questions to AI instead, the forum simply empties out.

Asking a coworker works the same way, just leaves behind a different kind of residue. When you ask, they find out what you're working on, and you find out what they've been dealing with too. Someone sitting nearby who overhears might learn something, or remember they hit something similar. A small request for help quietly leaves a few people knowing each other a little better, with a bit more shared understanding than before.

Asking AI only solves the one problem in front of you

AI gives you a fast, accurate answer, but the whole process happens somewhere only you can see. Once this one question is answered, nothing gets left on a forum for the next person to search, and nobody in the office learns anything new. The next person to hit the exact same problem, whether it's a coworker or some stranger on the other side of the world, still has to ask AI all over again, because the last person who solved it left no trace at all.

Whatever side effect used to be built into the act of "asking," whether it was helping the next stranger or letting coworkers understand each other a little better, AI simply can't include it for free. What it gives you is always a single answer, for you alone, this one time.

Losing the side effect means losing something else too

Someone once ran a small experiment: she had a tricky work conversation she didn't know how to open, so she sent the same question to four different AI systems at once, hoping to find the right way to phrase it. All four gave genuinely different, useful advice. But her own conclusion afterward was this: AI can hand you ten different ways to say something, but only you know which one the specific person you're about to face actually needs to hear. The one step that mattered most, none of the four AI systems could help with.

That points to another layer of difference. No matter how detailed AI's advice is, it's only responding to the text you typed in. It has no idea about the unspoken history between you and that coworker, or the understanding you two share without needing to say it out loud. But a coworker who actually knows your situation can land the point with a single sentence, because they're standing in the same room you are. AI gives you breadth. A person gives you fit. And when a real person answers, they often surface an angle you'd never have thought of yourself, simply because they're standing somewhere different and seeing something you can't. That jolt of being hit by someone else's perspective is hard for AI to give you, because it was never actually standing in the room with you.

The one-line takeaway

AI really is useful, and the speed is real, this isn't a call to stop asking it things. But it's worth remembering that some questions are fastest solved by AI, while "asking a person" is worth deliberately preserving, whether that's a coworker or some public place strangers can find later, not for efficiency, but for the thing efficiency can't buy: a shortcut left behind for the next person, an angle you'd never have found alone, and the bit of shared understanding that's left over once the asking is done.

Whether AI eventually starts stitching everyone's questions together into some new kind of shared public knowledge is a separate question, one this piece isn't going to get into.