The Case for Refusing to Build the AI Feature Everyone Is Asking For
A thought experiment on when saying no to an AI feature is the more disciplined engineering decision than finding a clever way to build it.
This one's speculative — a thought experiment about design principles, not a report on something I've built.
Capability is not the same as "should"
Most requested AI features are technically buildable — the current state of the tools makes an enormous range of things possible that weren't a few years ago. Buildable and worth building are different questions, and the second one gets skipped far too often in the excitement of discovering the first answer is yes.
The request is usually a symptom, not the actual need
A request for a specific AI feature is often a proposed solution to an underlying problem that hasn't been named explicitly yet, and building exactly what was asked for can solve the stated request while leaving the actual underlying problem untouched — or, worse, papering over it in a way that makes the real problem harder to see and address later.
Some problems get worse with an AI layer on top
A process with an unclear underlying policy, or a genuine disagreement about what the right outcome even is, doesn't get resolved by automating it — it gets automated inconsistently, at scale, with an AI system now confidently enforcing whichever interpretation happened to be baked into its training or its prompt, which can be worse than the original ambiguity, because it now looks authoritative.
The discipline is asking before building, not after
The valuable move isn't refusing every AI feature request — it's asking, before starting, what underlying need the request is actually a proxy for, and whether an AI feature is the right answer to that underlying need or just the most exciting-sounding answer to the surface-level ask. Sometimes the right deliverable is a clearer policy, not a smarter system enforcing an unclear one faster.
I'm Jesse Myers — Marine veteran, 32 years in enterprise IT, now building production AI systems. This site is where I write about what I've actually built, and occasionally about ideas I haven't built yet but think are worth taking seriously.