One Designer + AI: Scaling Design for 7,000-Person Conference

This included overseeing assets for more than 140 sponsors, 300 speakers, and 600 sessions. The sheer scale of these requirements highlighted the need for automation to manage the workload effectively.

This broad scope included all track mascots and signage, demonstrating the extensive reach of their design responsibilities.

Managing a high volume of design outputs carries specific risks that can significantly impact event success. Key concerns included the potential for missing sponsor logos, which could cause substantial issues. Additionally, incorrect speaker schedules posed another major challenge, highlighting the need for meticulous accuracy.

This composition signifies a shift where traditional tools are no longer the primary hurdle; instead, the key advantage lies in leveraging these resources to solve scale-related challenges.

When using Large Language Models (LLMs) for design assets, it is crucial to think creatively to overcome their inherent limitations. LLMs often struggle with specific vector tasks, such as accurately rendering complex images like a pelican on a bicycle. The workaround involves generating a basic PNG image through the LLM, then using Figma for subsequent vectorization, which enables the creation of detailed vector graphics. Success in this hybrid approach depends on adapting traditional creative processes to align with and maximize AI capabilities.

Overcoming scale issues in design relies on a five-step framework. This framework includes establishing a strong foundation, creating reusable designs, implementing automated workflows, diligently validating all outputs, and actively working to remove friction points in the process.

Establishing a robust design system is essential as the foundational element for any scalable design operation. This system encompasses critical elements such as typography, colors, and components, which are crucial for maintaining consistency. The consistency begins with the development of core assets during the initial stages of website creation.

To enhance design scalability, atomic design principles are employed to create modular, reusable components. This approach involves building design foundations much like Lego pieces, where the smallest parts are created and then combined. This method allows for efficient and consistent delivery of various design assets.

This includes defining specific typography for both desktop and mobile applications, which is essential because LLMs often generate random font sizes if not explicitly guided. Enforcing consistent colors and fonts is also critical to maintaining overall brand integrity.

This means marketing can generate email designs, flyers, and documents directly based on the already defined website branding. The consistent use of mascots and other design elements ensures a unified visual identity across all communication materials.

This automation allows for the direct pulling of the latest data and exporting it accurately to PNG files for event displays. This eliminates significant friction and reduces feedback loops between the design and engineering teams, streamlining the entire scheduling process.

To manage the creation of assets for over 300 speakers, an automated generator was developed to scale speaker announcements. This tool provides customizable templates, allowing for easy name changes and the integration of speaker headshots. The system automatically exports pixel-perfect graphics, including popular trading card formats, significantly streamlining the production process.

The product designer's workflow has evolved significantly, moving away from traditional design-thinking cycles that emphasize research and extensive feedback loops, which are now considered outdated for high-velocity tasks. The current process integrates AI tools like Devin and MCP directly into Slack, enabling real-time asset delivery. Figma plugins and spec sheets are utilized to ensure pixel-perfect standards are maintained throughout the streamlined workflow.

Large Language Models (LLMs) demonstrate the capability to interpret unorganized design files effectively, even without neatly arranged layer names. The LLM automatically defines crucial elements such as spacing, font sizes, and colors within a design. This functionality significantly aids in achieving a pixel-perfect product, regardless of the initial organization level of the input files.

This capability allowed for the verification of over 140 sponsor logos using AI's visual detection. The collaboration between human input and AI effectively creates a robust internal QA team, eliminating friction and ensuring high-quality output, including application to merchandise like t-shirts.

Effective design necessitates thinking from a user's perspective, recognizing that every user has specific needs. Designers must simulate user experiences, such as navigating registration, wayfinding, and QR code usage, to ensure a cohesive and connected journey. The ultimate goal is to prevent attendees from getting lost and to ensure they can easily find their designated rooms and other necessary information.

Real design work frequently involves managing unforeseen operational exceptions, such as last-minute schedule changes. The designer leveraged AI to add an 'edit button' to static files that initially lacked one, enhancing flexibility. This workflow allows for rapid updates, re-exporting files to PNG, and quickly replacing them on screens, thereby enabling scalability by automating the handling of these exceptions.

The core philosophy for solving scale problems in design involves thinking small and anticipating every potential failure point. Designers should consider all the smallest possible details and everything that could go wrong to address issues proactively. Viewing problems as assets can push teams to ship better products by finding innovative solutions before issues escalate.