Nobody Is an AI Expert

The phrase "AI expert" keeps showing up in places it doesn't belong. LinkedIn bios, conference panels, vendor pitches. I've started to cringe a little every time I see it.

What the Label Does to Conversations

The thing that bothers me about it is what it does in conversations. When someone walks in claiming to be an AI expert, they tend to lead the conversation, steer the framing, and set what's worth paying attention to. The people they're talking to, who often know real things about their own corner of the field, quietly stop sharing. The title becomes a reason to dominate the room, and everyone else's knowledge gets treated like it doesn't belong there.

That matters, because no individual can actually keep up with AI alone. It's moving too fast and branching in too many directions, and the only realistic way to stay anywhere close to current is to stay open to what other people are figuring out. Staying humble is how you keep learning from everyone around you, and in this field, that's what staying current looks like.

What makes the dynamic worse is that "AI expert" isn't really a coherent claim to begin with.

Why "AI Expert" Isn't a Coherent Claim

The word expert only does real work when it's pointed at something specific. If someone tells me they're an agent expert, I could ask them something hard about agents and trust the answer. Same with someone who specializes in evals, or long-context inference, or reinforcement learning. Those are specific claims, rooted in years of actually doing the thing and getting most of it wrong first.

"AI expert" is a different kind of claim, because "AI" isn't a specific thing.

It's computer vision, natural language, robotics, reinforcement learning, statistical learning theory, hardware, alignment research, product design for AI-native tools, evals, safety, and probably a dozen subfields I don't even know exist. Nobody is deeply expert in all of that at once. Calling yourself an AI expert after working in one corner of it is a little like a biologist saying they're qualified to be an astronaut. Both fall under the broader umbrella of science, but the day-to-day work is so different that the comparison barely holds. The label collapses too many things into one word.

And even within a single corner of the field, the shelf life of what you know is short. Last week I would have told you that serious agent orchestration needed a third-party tool to do it well. This week Claude launched Managed Agents, and a big chunk of what I thought I knew got rewritten in a single announcement. That kind of shift happens constantly. You don't really catch up in AI, you just keep learning. The whole field needs to lean harder into being a learn-it-all, not a know-it-all, because the ground keeps moving under everyone.

Put those together and you get the thing I keep bumping into. Breadth is impossible because the field is too wide. Depth decays faster than most people can maintain it. So what is someone actually claiming when they call themselves an AI expert?

Where Real Expertise Lives

I want to be careful here, because real experts in the narrow corners do exist. If someone has spent ten years on alignment research, or a decade writing CUDA kernels, or years shipping production ML at scale, they're experts in that thing. The label makes sense when it's specific. "Expert in long-context inference." "Expert in RLHF." That's coherent, and I'd take their advice seriously.

What doesn't hold up is the generic version. Someone spends a weekend learning a new API, reads a few papers, and the "AI expert" label starts showing up on their profile. Or a consultant's LinkedIn just says "AI expert" with no further specificity. In both cases, the title is doing work the underlying knowledge can't support yet.

Here's where I want to complicate my own argument, because the people claiming the title aren't always being dishonest. Something real is happening when you spend enough time across multiple AI subdomains. You start to develop pattern recognition. You can tell when a demo is a real capability versus a well-edited video. You have instincts about which technical bets are likely to pay off and which are hype. You can read a paper and quickly tell if it matters or not. That's a real skill, it takes years to develop, and it's genuinely useful.

But calling it expertise is still a mistake, because it ends up doing the same thing to conversations. The person wearing the "expert" title takes the lead, and the person who actually has a real ear for the field stops sharing what they're learning.

How I Think About It

I'd rather work with someone who is honest about what they know, able to lean into that, and eager to learn more in every conversation they have.

I work in product. I build with these tools every day, I read a lot, and I have opinions I'm willing to defend. But I wouldn't call myself an AI expert by any stretch. I think I'm getting really good at being fluent with agentic applications, applied customer-facing AI, building with agents, and more. But I go into every conversation knowing I'll learn something new, and working hard to make sure the person I'm talking to feels empowered to share the things they know.

AI is too big a field and moving too fast for anyone to realistically act like they've got it figured out on their own. The "expert" label is getting pasted onto too many profiles that can't really back it up, and it changes the way people show up when they wear it. What actually holds up is honesty about what you know, specificity about the corner you've actually worked in, and a willingness to learn from everyone you talk to about the rest.