Stop Prompting: Build an AI "Design App" Instead (Demo)

What It Covers

Igor Lore, CEO of Weebly (acquired by Figma), demonstrates how to build systematic AI creative workflows instead of endless prompting. The episode covers node-based workflow construction, multi-model testing, brand consistency systems, and YouTube thumbnail A/B testing automation. Lore argues creative teams must shift from pixel editing to system building to scale AI capabilities across organizations.

Key Questions Answered

• •Multi-Model Testing Framework: Weebly enables simultaneous testing of Google Imagen, Runway, Mystic, Minimax, and GPT image models from one prompt, revealing drastically different outputs. This node-based approach prevents vendor lock-in and allows teams to swap models as technology improves without rebuilding workflows, ensuring future-ready creative systems that adapt automatically to better AI capabilities.
• •Brand Style System Architecture: Teams can create four reference illustrations in their brand style, then generate unlimited new assets matching that aesthetic by feeding examples through an LLM analyzer that extracts visual patterns. This workflow took one week to build but enables non-designers to produce on-brand content instantly, transforming creative from bottleneck to enabler across 300-person marketing teams.
• •YouTube Thumbnail A/B Testing Automation: The workflow accepts one baseline thumbnail, analyzes it through visual design extraction prompts, then generates five variations testing specific hypotheses (visual hook, headline, pose changes). Building this system required one week of iteration to achieve consistent quality, but eliminates the manual barrier preventing most creators from leveraging YouTube's native A/B testing features.
• •Creative R&D Investment Mindset: Successful AI-native teams allocate dedicated time for workflow experimentation, accepting that some efforts yield no results. Leaders must recognize that simple ChatGPT prompting cannot compete with proprietary systems built through months of refinement. Teams treating creative AI as research and development gain competitive advantages through capabilities competitors cannot replicate with consumer tools.
• •System Building Over Asset Creation: Professional creatives must evolve from making individual pixels to constructing reusable processes that encode brand guidelines, quality standards, and creative intuition. One specialist builds the system incorporating expert knowledge about composition, color theory, and brand constraints, then exposes simplified interfaces allowing entire marketing teams to generate production-ready assets without creative bottlenecks or inconsistent outputs.