Executive Brief · 2026

AI workflows for digital product design

The case, the operating model, and the playbook for designing in code with AI agents: from brand file to shipped PR without a handoff layer. Summarized from the full playbook in 13 slides.

Kyle Cooney, Head of Product Design · Read the full playbook →
Executive summary02

Three messages: the translation layer is the cost, agents removed the barrier, and judgment is now the differentiator

01
The handoff is the most expensive artifact in product development

Design in Figma, spec for engineering, rebuild from scratch: every stage loses fidelity, and two sources of truth drift permanently. Designing in code eliminates the translation layer entirely. The screen the designer builds is the screen engineering reviews and ships.

02
Agent harnesses removed the technical barrier for designers

With an agent harness (Claude Code, Codex, Hermes), a designer goes from idea to live URL in an hour and a complete flow in a day. No HTML/CSS prerequisite: the designer directs in design language; the agent implements against a tokenized system.

03
Production is cheap; evaluation is the craft

Generating five directions costs minutes. Knowing which is right is the designer's moat. The workflow instruments judgment: audits, critique loops, decision logs, and quality gates keep taste, not typing speed, as the limiting factor.

Source: AI Workflows for Digital Product Design, sections "Why this approach" and "What you get"
The problem03

The traditional pipeline rebuilds every screen twice and keeps two sources of truth in permanent drift

Traditional Figma pipeline

Design in Figma

A mockup that looks like the product but does not behave like it

Spec for engineering

Interactions, states, and edge cases documented secondhand

Engineer rebuilds

Every decision re-solved from scratch in a different medium

Mismatch & revision

QA against the mockup; designer re-enters after the fact

This playbook

Design in code

Working screens against real tokens and components from day one

Every state designed

Hover, loading, error, empty: built, not deferred to QA

PR as handoff

Engineering reviews a diff it can run, not a picture

Ship

One source of truth in git; nothing drifts because nothing is duplicated

Source: playbook section "Why this approach"
The approach04

Five shifts define the model: the codebase becomes the design file, and handoff becomes continuous

DimensionTraditionalThis model
ArtifactFigma mockup that must be rebuiltRunning screens in a browser; the prototype is the product
Source of truthFigma file and codebase, drifting apartOne codebase: tokens, components, decisions, screens in git
HandoffA moment; designer involvement drops offContinuous; designer submits PRs, engineering reviews and merges
Design systemLibrary in a tool engineering cannot importtokens.css + shadcn components + skill files; imported directly
States & edge casesDiscovered during QADesigned at build time, verified by the agent in a browser
Source: playbook section "Why this approach"
Operating model05

The workflow runs in three temporal modes: set up once, build in a loop, maintain weekly

1
Setup (once per project)
  • Brand input & skill files
  • Instruction file (CLAUDE.md / AGENTS.md)
  • Design system scaffold
  • Token export for engineering
  • Accessibility audit on tokens
  • Governance rules
2
Design & build (every sprint)
  • Build flows end-to-end, all states
  • Canvas & prototype review views
  • Extract & enforce patterns
  • Log decisions as they happen
  • Copy & content audit
  • Handoff via PR
3
Maintain (weekly, background)
  • Git sync & system branch merge
  • A11y, consistency & performance audits
  • Token export refresh
  • Changelog generated from the diff
  • Scheduled design critique

The middle column is where designer time lives. Setup is behind you after week one; maintenance is largely agent-run. The first week is slower than Figma; every week after compounds.

Source: playbook section "The full progression"
Tool stack06

Four tools run the entire workflow, and the most important one is deliberately replaceable

Agent harness

Claude Code (the pick) · Codex · Hermes · Cursor

  • Holds project context, builds everything
  • Terminal CLIs preferred; desktop apps improving fast
  • Swappable: context lives in files, not the vendor
Vercel
  • Every branch push auto-deploys a preview URL
  • The shareable artifact for testing & review
  • Hosts the living design system site
GitHub
  • Single source of truth for design and engineering
  • PRs as the handoff mechanism
  • CODEOWNERS as the review gate
v0 (occasional)
  • Throwaway exploration outside the system
  • Lanes: v0 for React, Lovable full-stack, Figma Make in-Figma
  • Explore there, build here
Source: playbook section "Tools overview"
Foundation07

Output quality is a context problem: what the agent has loaded matters more than how you phrase the request

Reference skills

One markdown file per domain (brand, a11y, copy, performance, components): the rules, the exceptions, and what good looks like. Plain markdown, readable by any agent, versioned in git.

The instruction file

CLAUDE.md / AGENTS.md: stack rules, folder structure, absolute constraints ("never hardcode values, never push to main"), and current status. Read automatically at every session start.

Designer language

The agent responds to visual, spatial, felt vocabulary: contrast, density, hierarchy, tone. Interaction patterns are complete instructions: "apply progressive disclosure" needs no elaboration.

The industry now calls this context engineering. The setup is the heavy lifting; prompting is steering. Without design context loaded, every agent produces the same generic UI.

Source: playbook sections "Skills & instruction files" and "Prompt like a Designer"
Execution08

Seventeen steps take a team from brand file to production rhythm

Design system setup (1–5)
  • 1. Prepare brand input
  • 2. Scaffold the system (shadcn + Tailwind v4)
  • 3. Export DTCG tokens for engineering
  • 4. Accessibility audit (WCAG 2.2, token level first)
  • 5. Publish the living design system site
Design & build (6–12)
  • 6. Build major flows, all states
  • 7. Canvas & prototype views
  • 8. Localization as content tokens
  • 9. Extract & enforce patterns
  • 10. Decision log & changelog
  • 11. Content & copy audits
  • 12. App-specific skills
Eng collaboration (13–17)
  • 13. Performance budget
  • 14. Handoff via PR, not spec
  • 15. Design on real production code
  • 16. Deploy & environments
  • 17. Git sync & weekly design rhythm
Source: playbook, steps 1–17
Quality gates09

Quality is enforced by layered gates, and the agent verifies its own work in a browser before you see it

Standing budgets & standards
  • WCAG 2.2 AA minimum; contrast fixed at the token level
  • Performance: <200kb JS, LCP <2.5s, CLS <0.1
  • 44px touch targets; reduced-motion on every animation
  • Budget breaches block the weekly merge
Agent-run verification
  • Browser-in-the-loop: the agent walks flows, screenshots states, flags breakage
  • Deterministic scanners in CI; agent judgment on top
  • Pre-PR self-review against the engineering codebase
  • Weekly consistency, copy, and a11y sweeps
Engineering review gateApplies to
Blocking approvalShared components, new dependencies, auth/payments/data, API shape changes
Awareness onlyNew screens on existing components, copy changes, token-scale adjustments
Designer merges directlyDesign-repo tooling, skill files, decision log, documentation
Source: playbook steps 4, 13, 15
Portability & risk10

Portability is the operating model: any designer works from any harness and resumes in any other, with no seams

AGENTS.md

The cross-tool instruction standard. Write project rules once; every harness reads them. Claude Code bridges via a one-line import.

MCP

Vendor-neutral tool connections (Figma, shadcn registries, browser, analytics) supported across Claude, OpenAI, and Google. Config travels in the repo.

DTCG tokens

The W3C community token format, stable since 2025. Consumed by Style Dictionary, Tokens Studio, and Figma variables; legible to any agent.

Markdown skills

All accumulated expertise is plain text in git. Works pasted into any model or read by any CLI. No export, no lock-in.

The payoff is daily, not hypothetical: harness choice is per-task and per-designer, and the repo is the session state (branch, current status, decision log, skills), so work started in one harness resumes in another with no seam. Vendor risk falls out for free: if a harness disappeared tomorrow (one major consumer CLI did in 2026), the team loses a driver, not the system.

Source: playbook section "Keep it portable"
Scale path11

The system scales from solo to a five-person team because the documentation is the onboarding

Team sizeGovernance model
Solo / 1–2One owner. Second designer contributes to skill files via PR. Instruction file doubles as the onboarding doc.
3Ownership split by domain: brand + tokens, components, governance. One reviewer on shared files.
4–5CODEOWNERS on /skills and system files; monthly consistency audit; quarterly ownership rotation.
Ready-to-scale signals
  • Instruction file stable across sprints
  • Skill files produce consistent output without correction
  • Decision log has real entries, not templates
  • At least one full flow shipped through PR review
Proof point
70–80%
drop in design-system support questions measured by Miro after making its system agent-readable. The system starts answering for itself.
Source: playbook sections "When you're ready to grow" and "Team & governance"
Frontier12

The same foundations unlock the frontier: parallel agents, measured iteration, and agentic products

Multi-agent parallelism

Build one flow while a second agent audits, documents, or localizes. Defined specialists, git worktrees for isolation, and agent teams with debate-and-vote councils for critique.

Metrics-driven refinement

The autoresearch pattern applied to flows: agents propose variants inside design system constraints, previews serve them, analytics scores them. ~700 experiments found ~20 real improvements in the reference run.

Designing agentic products

A separate competency clients now ask for: generative UI assembled at runtime, plus agentic UX patterns (planning visibility, tool disclosure, streaming states, recovery, memory surfacing).

Source: playbook appendix
Next steps13

Start Monday: five moves take you from zero to the loop

01

Write brand.md. Convert your brand guidelines to one plain-text file: colors, type scale, iconography, motion, voice.

02

Scaffold the system. One prompt: shadcn + Tailwind v4 themed from brand.md, with a living style guide at /design-system.

03

Write the instruction file and skills. Absolute rules, folder structure, a11y and copy standards. This is the onboarding doc for every future agent and designer.

04

Build one flow end-to-end. Every state, agent-verified in the browser, deployed to a preview URL, shared as a link.

05

Open the first PR. Let engineering review a diff instead of a spec. Then set the weekly rhythm: sync, audit, merge, changelog.

Full detail, prompts, and worked examples: kylecooney.com/ai-design-playbook
Source: AI Workflows for Digital Product Design, 2026 edition
1 / 13