The New Frontier of Craft: Why AI Won't Commoditize Everything

Many people assume AI will eventually turn design into a sea of generic, beige interfaces. However, a conversation with John Maeda and Paul Bakaus points to a different reality. While AI raises the floor by handling mechanical tasks, it creates a new opening for high-level, opinionated human craft. The real advantage does not go to those who use AI to churn out more content, but to those who use it to raise the ceiling. By using agentic workflows to handle the final 10 to 20 percent of a project, builders can apply genuine taste where it matters most. For leaders, the path is clear: stop chasing the quick wins of speed and start betting on the long-term value of conviction.

The Hidden Cost of Fast Solutions

Most teams treat AI design as a volume game. They use Large Language Models to generate interfaces in hopes of saving time. As Bakaus points out, this leads to a problem with slop. When everyone relies on the same models, the output drifts toward a common denominator, resulting in the repetitive gradients and backgrounds that define the current AI aesthetic.

The trap here is that teams prioritize speed while ignoring the long-term cost: brand dilution. When your product looks like every other AI-generated site, you lose your ability to stand out.

The hard truth is that in rooms where decisions actually happen, taste as a whisper and velocity a megaphone, it is almost impossible to defend a design nuance or instinct against a looming deadline.

-- John Maeda

The danger is cognitive as much as it is aesthetic. Bakaus warns of cognitive surrender, where builders stop reading the plans or questioning the results, letting the model dictate the product direction. This creates a feedback loop: the more we rely on the model defaults, the more the model reinforces those defaults as the correct output, which narrows the creative landscape.

Why Design Engineers Hold the Advantage

The most interesting insight from this discussion is that the wall between design and engineering is fading. Bakaus found that when designers use the same tools as engineers, they get better results. This happens because their vocabulary becomes more precise. Designers use terms like vertical rhythm and negative space, which serve as higher-fidelity instructions for the model.

This suggests that the most valuable people in an AI-native world will be design engineers, who can translate high-level taste into the substrate of code.

Designers are closer to programmers, but programmers of interaction. And I think Paul probably he subconsciously is working in this way when we talk about E-VALs, Paul. E-VALs live in latent space construction as literal meta design.

-- John Maeda

By building tools like Impeccable, Bakaus is creating a PostScript moment for design. Just as PostScript provided the building blocks that allowed engineers to create Photoshop, these new agentic tools provide the vocabulary that allows designers to steer AI away from the generic and toward the bespoke.

The 18-Month Payoff: Betting on Conviction

Conventional wisdom says AI will make design cheaper and faster. But the real competitive advantage lies in the last 10 percent. As the floor for good enough software rises, the market will stop rewarding basic functionality and start rewarding human trust and accountability.

This requires a change in leadership. Maeda notes that the winning leaders are not those with the most ideas, but those with the conviction to recognize and bet on the right ones. This is a long-term investment. It requires the patience to ignore the quick release in favor of a global maximum that actually improves the user experience.

Key Action Items

• Audit your AI-defaults: Over the next quarter, review your AI-generated assets. Are they indistinguishable from your competitors? If so, you are suffering from model-drift.
• Adopt a High-Fidelity Vocabulary: Stop prompting for a nice website. Start prompting with specific design primitives, such as use a 12-column grid, increase vertical rhythm, or apply negative space. This pays off immediately in output quality.
• Prioritize Agentic Experience: Begin shifting design focus from visual interfaces to agentic affordances, such as designing for how agents interact with your system through robust command-line tools, clean API structures, and clear error messaging. This is a 12 to 18 month investment in future-proofing.
• Insist on Human-in-the-Loop E-VALs: Do not let your team perform cognitive surrender. Implement a review process where the human must explicitly sign off on the intent behind the model output, not just the result.
• Hire for Design-Engineering hybrids: Look for talent that bridges the gap between aesthetic judgment and technical implementation. These individuals can steer the AI toward the last 10 percent of unique value.
• Cultivate Conviction in Reviews: In the next design review, challenge your team to defend their choices against what the model suggested. Reward those who push back, even if it adds time to the sprint. This builds the organizational muscle needed for high-taste output.