Nad Chishtie: Lovable's Design System For Agents

Summary

Nad Chishtie, Lovable’s Head of Design, describes how AI-assisted software creation is changing design work and team structures. A self-described generalist, he says his strengths are inventing products from scratch, balancing craft with speed, and working closely with engineering on technically ambitious ideas.

Designing and Building with AI

• End-to-end ownership matters: Chishtie sees people who can take an idea from user insight through prototype and into production as especially effective in AI-native organizations. He calls the needed initiative “gumption.”
• Prototyping is no longer limited to designers: Marketers, data specialists, and others can now test ideas that previously required learning tools such as Figma or front-end coding. Lovable uses prototypes and reversible, controlled rollouts to learn from real usage.
• More creation also means more risk of fragmentation: Prototypes can become informal commitments, and independently built features can make a product feel inconsistent. Lovable’s design function is therefore shifting from gatekeeping to enabling people while maintaining a coherent experience.
• Experimentation includes deleting work: Teams are encouraged to prototype quickly, test promising ideas in limited ways, and remove weak ideas rather than letting them accumulate. Internal hackathons create room for experiments that may never ship.

Lovable’s Organization and Enterprise Use

Lovable’s core product organization includes teams focused on its AI engine, the user experience, and growth. Designers and engineers are embedded in teams, which are expected to own problems and decisions end to end.

The company uses beta groups, Discord, and an enterprise customer advisory board to gather feedback. As adoption grew, enterprise readiness—including features such as SSO, role-based access, service-level agreements, and security practices—became a priority.

In enterprise settings, teams may disagree about who should own “vibe coding.” Chishtie says security and design quickly become key stakeholders: security because of data and compliance concerns, and design because widespread creation can produce work that is off-brand or inconsistent.

Design Systems for Agents

• Lovable initially went without a design system, but Chishtie says it later became important for keeping the product consistent as more people began making changes.
• The design system is now roughly half interface components and half “skill banks” written for agents. These skills guide agents toward existing components and consistent patterns—for example, applying keyboard-focus behavior across an application.
• Chishtie compares these skills to linters: they can run automatically or be invoked for a specific task. Lovable also uses automated checks, visual diffs, and related processes in pull-request workflows.
• A two-week experiment in intensive agent use found that background agents often created artificial productivity: reviewing many separate results and switching contexts could take more effort than doing the work directly. More useful were skills that check or improve work during its creation and as part of development workflows.

Internal Tools and Future Direction

Lovable has built custom internal software, including an agent-based analytics tool that can answer questions and visualize data, and an HR planning tool for representing people’s work across multiple teams. Chishtie says smaller companies may be more willing to build custom tools, while larger organizations weigh build-versus-buy decisions against compliance and risk.

He says internal tools and agent skills often spread informally: people share successful work, and other teams adopt it or build their own.

His broader thesis is that software-creation tools should become interoperable, allowing people to choose different environments and workflows rather than being confined to one design or coding tool.

Chishtie says Apple’s restrictions on mobile apps that enable vibe coding are an unresolved issue for Lovable; the company is discussing how to respond.

Hiring and Advice

Chishtie says Lovable expects high AI fluency, including an understanding of how models work, and views AI as a material designers need to learn to shape. He also emphasizes self-learning and intrinsic drive when evaluating candidates. His advice to people who want to build AI skills is to start with small projects, follow an itch, and treat the work as play.

Reviews, guides, or tutorials: None are presented; the video is an interview and analysis.

Main speakers: Tommy Geoco (host/interviewer) and Nad Chishtie (Head of Design at Lovable).