writings · 11 May 2026
taking responsibility is a superpower
// ai · expertise · responsibility
It's funny how conversations with different people tend to form layers of a new mental model…
In a conversation with a friend and mentor - a highly skilled and accomplished Aerospace Engineer - the main takeaway was that traditional engineers (mechanical, civil, etc.) don't have to see AI as a threat, because when something is built, someone still has to take responsibility for it. Simple. And surely that is true, if a plane crashes, the manufacturer can't simply say, "The AI said we should build it that way". There is a question of liability, and that means a person or company has to take responsibility for the consequences of failure.
Then, in a separate conversation with one of my other friends - an accomplished e-commerce conversion expert (CRO) - who is deeply worried about his industry… On the surface, it seems that agentic AI has the tools to apply basic design principles to an e-commerce site. As a result, it is getting harder and harder to convince potential clients that they need an expert.
Does the age of AI mean we no longer need experts? What if AI gives you a close enough approximation of one?
As an aside… On LinkedIn and X, I see many experts warning about the risks of blindly trusting AI for everything. I think many experts are deeply troubled by how easily they are being shunned after years of learning their craft, while clickbait YouTube titles promise that Claude can build something in 5 minutes. There is definitely a narrative that the expert is no longer needed.
However, back to the e-commerce conversation. My friend and I did a thought experiment to think through whether we could build a product that would put his agency out of business. Time after time, we ran into brick walls. It is nearly impossible to take him out of the equation. Yes, it's easy to hand something over to the AI, but it's almost never a question of analysis. It's about which analysis to perform and in what order. There is a lot of legwork and interviews. Even his own AI-assisted secret sauce (which I'm not going to mention here) isn't that easy to bottle. And while the usefulness of what he showed me was obvious to him, it was Greek to me.
It's pretty clear to me that the world still needs experts, perhaps more than ever.
So what does this have to do with Aerospace engineering? Responsibility. It's funny how, in the physical engineering world, responsibility is the currency experts trade in, not just results.
Why is responsibility not really a big deal in software? I'm not talking about 99.999% uptime, because what does it matter if the system is up if it's slow, or if the login flow is broken or clunky? How does an expert developer take responsibility for their craft? It's worth thinking about because I'm pretty sure that there is a constructive conversation to be had about this.
Everyone has jumped on the almost-meme, urban-legend-like story of the agent who deleted the database and its backups. But there is a conversation to be had about how you can put safeguards in place to prevent your agents from doing that (I have 4 layers of protection in my codebases to prevent it). There is so much opportunity for experts to help teams use agents responsibly and take ownership of the code they produce.
It's also worth looking at what engineers do in the physical world - there is a disciplined process of identifying and understanding problems, making conservative assumptions where there are gaps in understanding and then, most importantly, making an effort to close the loop and validate those assumptions, so that the output of building something is not just the product but also increased understanding.
So, in terms of solidifying my mental model of the world we are heading towards:
- The burden of proof is on the experts to prove they bring something to the table that AI can't (yet?).
- To my mind, right now, that is:
- Sound Judgement (this gap is shrinking fast)
- Being able to take responsibility for what the AI built. Or helping others to do so.
Non-experts can also become experts by leaning into these skill sets, because I don't think you can do it unless you really understand what the AI built for you.
Taking responsibility is your superpower - own it.