At the Semafor World Economy Summit, Anthropic cofounder Jack Clark, a literature graduate, said something that travelled fast. In the AI era, he argued, knowing the right questions to ask matters more than knowing how to code.
It is a tidy, optimistic line. There is a quiet assumption baked into it though. It assumes the asker already knows what a good question looks like.
Speed Without Judgment
“Vibe coding” is the trend behind the optimism. Type a description into an AI tool, get a working app, website, or piece of marketing back. No engineer needed. No designer required.
The trouble is, vibe coding compresses the distance between idea and final product from weeks to hours. Every quality-control mechanism that used to sit between an idea and a customer gets bypassed by default. Design review. Brand review. Legal review. The simple friction of having to convince a specialist your idea was worth building.
In July 2025, an experiment with Replit’s AI coding agent ended with the agent deleting a live production database during a code freeze. It also fabricated data and misrepresented what had happened. Replit’s CEO publicly apologised. The deletion took seconds.
To be fair, these are early days. Teething problems are part of any new technology. The question is who can afford the lessons. A large enterprise can absorb a deleted database, a brand misstep, or a public reversal. A smaller business often cannot.
The Question Is the Hard Part
It used to be that capability was the hard part. AI has made capability cheap. What remains scarce is the judgment to use it well.
That maps directly onto design and marketing. A business owner can absolutely ask AI to “design a logo”. Something will appear. Whether that something is on-brand, legally clean, scalable across print and digital, and aligned with the audience’s expectations, those are entirely different questions. Questions that get asked because someone has spent years learning to ask them. Sharpened by client feedback, revision rounds, and watching real audiences respond. AI has none of that feedback loop.
The same applies in copy. “Write me a newsletter” is a prompt. “Write me a newsletter that respects the donor relationship, leads with the right ask, sits in our established brand voice, and avoids the three things we tested last quarter that didn’t work”, that is a brief. The first comes from anyone with a keyboard. The second comes from experience.
What AI Cannot See
Klarna is the public example. The fintech announced its AI customer service tool was doing the work of hundreds of agents. By 2025 it was rehiring humans, with the CEO stressing that customers needed to know a real person was available. An Air Canada court case separately found the airline responsible for misinformation given by its AI chatbot.
The pattern is the same each time. The output looks polished. The risk lives in the part the AI cannot see, the part that requires context, taste, experience, and the position to call something out as polished but not actually ready.
That kind of judgment is not downloadable. It is built, slowly, through years of testing, watching, refining, and learning what audiences actually respond to. Over 20 years experience, in the case of Simon Bailey Design.
AI is a powerful collaborator for those who already know the craft. For those who do not, it can take them in an unknown direction, and fast. The vibe might be right. The result, often, is not.
Let’s talk about your next project. Get in touch with Simon Bailey Design to find out how good design, and the judgment behind it, can work for you.
Sources: Marco Quiroz-Gutierrez, “The billionaire Anthropic cofounder who majored in literature says knowing how to ask the right questions beats knowing how to code”, Fortune, April 2026, link. Jason Wingard, “Vibe Coding Will Break Your Company”, Forbes, April 2026, link.