Designing the Operating Model for AI-First Product Teams.

Most design organizations are still optimized for a world where humans create and software executes. AI changes that assumption. At Kustomer, I rebuilt the design operating system around a new reality: designers, engineers, and models working as one system.

What became possible

BEFORE
  • Days to prototype

  • Weeks to validate

  • Engineering interpreted specs

AFTER
  • Same-day product exploration

  • Live AI interactions in design reviews

  • Working software instead of static artifacts

  • Direct testing with users earlier in the cycle

Prototype Creation.

3d → 3 hrs

Idea to test cycle time

↓80%

Spec creation reduced

80%

The Shift

Guiding Philosophy
  • Design behavior, not screens.

  • Work in live systems, prototypes over specs, always.

  • AI is part of the system, not a side tool.

  • Codify decisions so they scale (skills, patterns, rules).

  • If it can’t be tested, it’s not ready.

  • Evaluate outputs, not intent.

  • No handoffs, co-create with engineering from the start.

  • Default to small, fast loops over big, polished deliverables.

  • Use real data as early as possible.

  • Prioritize impact over completeness.

  • Align tightly so teams can move independently.

  • Share knowledge openly,no silos, no gatekeeping.

PERSONAL TRANSFORMATION

In the span of six months, I also rewired my own workflow completely—moving out of static design tools as the center of gravity and into direct interaction with models using:

  • Cursor

  • Claude Code, Claude Design

This wasn’t theoretical adoption. It was operational.

What I Built
AI-Native Design Workflow

Shifted design and engineering from static handoffs to AI-native co-creation, where teams prototype directly with LLMs, test outputs against real data, and iterate in hours instead of days.

Codified AI Design System

Introduced Markdown-based “skills” to standardize AI usage across the team, including prompt structures, component fidelity rules, output evaluation criteria, and human-in-the-loop guardrails.

This turned AI from individual experimentation into a shared, enforceable operating system.

Embedded Evaluation

Moved design critique from subjective review to observable output quality. AI-generated work is evaluated against real use cases, defined edge cases, and documented product intent before shipping.

Co-Authorship Model

Eliminated traditional handoff by embedding design and engineering in the same loop: co-defining behavior, prototyping together, and reviewing outputs collaboratively.

Impact

Compressed idea-to-working-output timelines from days or weeks to hours, reduced interpretation gaps and rework, and tested design quality against real conditions earlier.

Key Insights

Creative thinking did not disappear. It moved from static files into system behavior: prompts, rules, evaluation criteria, and the user’s voice in the process.