Interviews and All That Kind of Crap

If you want to know the truth, I’ve been thinking about this whole AI-and-interviews thing for a while now. Not because I think AI is going to replace everybody or any of that corny science-fiction stuff. More like — interviews are such a phony communication surface to begin with, and maybe AI could actually make them less phony. I know. It sounds crazy. It sounded crazy to me too when I first started thinking about it. But the thing is, I think I might be right, and that depresses me a little, because I’m not usually right about anything.

The thing about interviews is they’re one of the hardest ways to communicate with another person that anybody ever invented. If you really want to know the truth, they compress about a million things into one stupid little conversation — memory, judgment, trust, social calibration, evaluation, your whole life basically — and you’re supposed to do all that while sitting there trying not to look like a nervous wreck. And there’s always this power thing going on. Some guy in a chair with all the cards, and you’re just sort of sitting there hoping you don’t sound like a phony, which is ironic if you think about it, because the whole situation is phony to begin with.

The problem isn’t always that you don’t know the answer. Sometimes you know the answer perfectly well. You know it backwards and forwards, you could write a book about it, you could teach a class. The problem is you have to retrieve it in real time while some person is staring at you, writing things down on a little notepad, and you’re supposed to act like it’s all completely natural. It is not natural. It’s about the most unnatural thing there is, unless you count funerals, which I don’t.

The Problem, If You Want to Know

A letter or an email, you get time. You can sit there, stare at the ceiling, draft something, delete the whole thing, start over, go get a Coke, come back, decide it sounds too emotional, rewrite it, leave it alone for a day, come back again, and eventually send something that doesn’t make you sound like a complete bastard or a complete idiot.

An interview does not give you that luxury. Not even a little bit.

Some guy asks you, “Can you walk me through a challenging project where you demonstrated technical leadership?”

And right away, about sixteen different things start happening in your head all at once, like a goddam pinball machine. You’re searching your memory. You’re trying to figure out which story makes you look good without sounding like you’re bragging. You’re trying to figure out what the guy actually wants to hear. You’re trying to figure out how much context is too much context. You’re trying to manage your tone so you don’t sound defensive or arrogant or like you’re reciting from a script. You’re trying not to mention the part where everybody hated each other for three months, or the part where you almost quit, or the part where the project was a complete disaster until the very last minute.

You’re trying to tell the truth without dragging in every depressing detail around the truth.

That is a lot of work. It really is.

For people with weird careers, or long experience, or any kind of history with conflict, or burnout, or rejection, or anybody who’s ever been evaluated by a system that didn’t understand them — the problem gets even harder. The answer may be there. The experience may be completely real. The technical judgment may be perfectly good.

But the output can still come out wrong.

Too long. Too caveated. Too defensive. Too emotionally exposed. Too much like you’re trying to explain your whole life in five minutes, which is impossible anyway.

That’s not a knowledge problem. It’s an interface problem. I hate that word, interface. It sounds like something a guy in a bad suit says in a meeting. But it fits here, so I’ll use it.

What Interviews Actually Want

Most people think an interview answer is supposed to be a confession, or a proof of worth, or some kind of legal testimony. It isn’t. It’s supposed to be a signal. A clean one. The guy on the other side is usually trying to answer a smaller set of questions, like:

• Can this person think clearly?
• Have they done work like this before?
• Can they lead without creating chaos?
• Do they understand tradeoffs?
• Can they collaborate without being a pain in the neck?
• Can they explain something technical without putting everybody to sleep?
• Are their claims grounded in anything real?
• Would I trust them with a difficult situation?

Meanwhile, the person sitting in the chair — the one being interviewed — is usually answering a much bigger, much more depressing invisible question. It goes something like:

Am I safe here? Am I being judged fairly? Will I be misunderstood again? Do I need to defend my entire existence as a professional starting from the beginning?

Those are completely different conversations happening at the same time. It’s a wonder anybody gets hired at all.

AI as Interview Regulator

So here’s where the AI comes in, and I know this sounds like I’m trying to sell you something. I’m not. I hate selling things. I hate people who sell things. I hate the whole business of selling things, especially when the thing being sold is some kind of software that’s supposed to fix your life.

But the useful version of AI in interviews — the version I can almost get behind — is not some agent whispering perfect answers into your ear while you sit there sweating. That’s just cheating with extra steps, and it misses the whole point anyway.

The better version is preparation. Structure. Regulation. Before the interview, AI can help you turn your messy, depressing, complicated work history into something that actually sounds like a story.

It can take all your crazy project experience and ask you some simple questions. Not to make you sound like somebody you’re not, but to make you sound like yourself, only cleaner. Questions like:

• What was the actual problem, without all the nonsense?
• What was your role, exactly?
• What made it hard, specifically?
• What tradeoffs did you consider?
• Who did you influence, and how?
• What changed because of your involvement?
• What was the outcome?
• What would you do differently, if you had to do it again?

That is not cheating. If you ask me, that’s structured reflection. It’s like having a friend who’s good at asking questions instead of a friend who tells you what to say.

Old AI and the Entry Point

The thing is, experienced people usually have the opposite problem from junior people. They don’t lack examples. They have too many. Everything connects to everything else. Every answer has caveats. Every technical decision has fifteen years of context behind it. Every story has a prequel and a sequel and a spin-off.

But an interview usually needs one clean entry point. One story. One example. One answer that doesn’t require a bibliography.

AI can help you find that entry point. That’s all it’s really doing — helping you find the cleanest way into your own experience. Everything after that still has to come from you.

Raw experience sounds like this:

I worked on a large mobile app with a lot of legacy complexity and security findings and product pressure and cross-team dependencies and some modernization work and some authentication problems and release confidence issues and a general need to make everything more maintainable without stopping delivery.

That may all be true. It probably is true. But it’s not an interview answer yet. It’s a data dump.

A better version sounds like this:

One challenging project was modernizing a large customer-facing mobile platform while the team was still under active delivery pressure. The technical challenge was balancing feature work with security remediation and maintainability improvements. My role was to break the problem into concrete risk areas and introduce better patterns incrementally.

That version has a shape. It tells the other person where to listen.

Influence Without Phoniness

One of the hardest interview questions — if you want to know the truth — is some version of “tell me about a time you influenced a team to adopt a new technology.” That question sounds technical, but it’s really about leadership, which is one of those words that makes me want to leave the room.

Technical leadership is usually not about having the most correct opinion. That’s what the confident bastards think, anyway. It’s about making change adoptable. Teams don’t resist change because they hate good ideas. They resist because the migration path is unclear, or the cost is hidden, or the risk is unevenly distributed, or they’ve seen modernization become an endless rewrite before. And you can’t blame them, really. I wouldn’t trust some guy coming in with a big plan and a lot of hand gestures either.

AI can help you find the leadership signal inside the technical story.

Not:

I wanted to use a better architecture.

But:

I connected the architecture change to specific problems the team already worried about.

Not:

I convinced everyone.

But:

I used small examples and code reviews and incremental adoption to reduce the risk of change.

That’s the difference between sounding opinionated and sounding like somebody people actually trust. It’s a small difference, but it’s the whole difference.

The Ethical Part

There’s a real boundary here, and I think about it more than I want to. AI can help you prepare. It can help you organize your stories. It can help you identify claims you can’t actually support. It can run mock interviews. It can suggest clearer phrasing. It can debrief after the fact and tell you where you rambled, which is probably everywhere.

But you have to own the answer. You really do. If the AI is secretly answering live questions while you sit there nodding, that’s just a more complicated version of lying. If it’s inventing experience you don’t have, that’s lying plain and simple. If it’s making you sound like you have judgment you don’t actually have, that’s probably going to come out anyway, and then you look like a phony and a fool, which is worse than just being honest.

The trustworthy version is simple. AI can prepare, coach, organize, and debrief. The human has to answer, decide, and own the claims. That’s it. That’s the whole thing. Everything else is just a more expensive way to be dishonest.

The Real Value, If You’re Interested

The real value of AI in interviews — the thing that actually matters — is not making people sound impressive. That’s what the phonies think it is. The real value is helping people be legible. That means helping them be understood, clearly and accurately, without all the noise.

A lot of capable people interview poorly. Not because they don’t know their stuff, but because the format rewards a very specific kind of real-time narrative control. If your experience is messy, or broad, or senior, or emotionally loaded, or difficult to compress into a neat little box, the interview is going to under-measure you. It’s just going to. It’s a bad system.

AI can help by creating a better interface between lived experience and evaluation. It can help you remember that the guy across the table asked for one story, not your entire relationship with work. It can help you answer the question directly before adding all the nuance. It can help you stop before the useful answer becomes an over-explanation that makes you sound insecure.

It can help you say:

That approach can work. The tradeoff I’d manage is ownership and failure modes.

Instead of:

That approach has a lot of problems.

Same judgment. Different signal. The first one sounds like a guy who’s been there. The second one sounds like a guy who’s complaining. I’ve been both. I prefer the first one.

What I’d Want

If I were going to use an AI for interviews — and I’m still not sure I would, because the whole thing makes me uncomfortable — the one I’d want is not a teleprompter. It’s more like a preparation and reflection system. A tool. A thing you use and then put down.

Before the interview, it helps me choose the right stories for the role.

During practice, it pushes me to be concise and grounded and specific.

After the interview, it helps me figure out what happened, where I rambled or undersold myself, draft the follow-up, and update my memory for the next time.

It doesn’t replace me. It doesn’t pretend to be me. It doesn’t make me sound like somebody who went to a fancy school and plays golf. It helps me show up as myself, but with a cleaner interface. I hate that word, but I’m sticking with it.

Closing, More or Less

Interviews are supposed to evaluate ability, and judgment, and communication, and fit. But they also evaluate state management, and social fluency, and memory retrieval, and how comfortable you look while being watched. Those are real skills — kind of — but they are not the whole person. They’re not even the most important part.

AI can make interviews worse. If it becomes covert impersonation, or synthetic polish, or another way to game a broken system, it’s going to make everything more phony, and everything is already plenty phony.

But AI can make interviews better if it helps people organize truthful experience, communicate at the right level, and stay present under pressure. Not by inventing anything. By helping people find the signal they already have.

That’s the version I care about, if I care about any version at all.

Not AI as a fake candidate.

AI as a regulator. AI as a mirror. AI as a system that helps a person communicate what they actually know, without sending every answer through fear first.

It’s a nice idea anyway. I don’t know if it’ll work. It probably won’t work. Most things don’t. But it’s a nice idea, and there aren’t many of those floating around, so I figured I’d write it down before somebody turns it into something phony.