Ten Pages in a Day — AI-Assisted Workflows in Production

The brief was aggressive: ten pages for Erdos Logistics, fully designed, fully built, ready for staging. The client expected it in two weeks. We did it in a day.

The Pipeline

The trick was not speed. It was architecture. Before touching a single design, I built a system: a Python pipeline that takes structured content inputs (headlines, body copy, CTAs, feature lists) and generates Elementor-compatible JSON templates. Each template follows the same design system — dark SaaS aesthetic, consistent spacing, unified typography.

Claude handled the content generation. Not as a replacement for strategy — the positioning, messaging hierarchy, and page structure were all human decisions. But once the skeleton existed, Claude filled in the muscle: feature descriptions, technical explanations, benefit statements. Each output was edited, but the first draft was 80% there.

What This Changes

This is not about replacing designers or writers. It is about compressing the commodity work so humans can focus on the judgment calls. Which pages need to exist. What story they tell in sequence. Where the conversion points live. Those decisions require taste, experience, and strategic thinking that no model provides.

But the execution layer — translating decisions into production assets — that can be dramatically accelerated. Ten pages in a day is not a flex. It is a new baseline.