Written by John Maeda Design in Tech Report 2026: From UX to AX — presented at SXSW 2026

All text below is the original, unedited work of John Maeda, open-sourced as part of the E-P-I-A-S × SAE Framework for AI upskilling product designers.

Overview The E-P-I-A-S x SAE Framework L1 to L2 Transition L2 to L3 Transition L3 to L4 Transition

Working Within SAE 🚗➕ L1 (and Preparing to Move to L2)

AI suggests; the designer drives every step

This is the entry point for most product designers using AI today.

  • L1 = AI helps you think, draft, and explore
  • L2 = AI helps you build bounded production chunks

The shift is not automation — it’s clarity, judgment, and intent.

What “Good” Looks Like at SAE L1

By the top of L1 (Integrator → Architect), a designer can say:

  • “I know when AI helps — and when it hurts.”
  • “I can explain why I trusted or rejected an AI suggestion.”
  • “My prompts are assets, not one-off chats.”
  • “Quality is still my responsibility.”

L1 maturity is about judgment, not automation. AI augments thinking; the designer remains accountable for outcomes.

What L1 Looks Like in Practice

  • AI is used to brainstorm, summarize, draft, or remix
  • Most AI output is rewritten or refined
  • QA lives in taste, experience, and intuition
  • Speed improves, but quality still depends on the designer

Typical L1 Tools

  • General LLMs: ChatGPT, Claude, Gemini (chat)
  • Visual ideation: Midjourney, Firefly, Krea
  • Embedded helpers: Figma Make, Notion AI, Docs copilots

SAE L1 × E-P-I-A-S Growth Path

❶ Explorer → ❷ Practitioner

(From novelty to consistency)

Goal: Stop trying AI. Start using it deliberately.

What to learn

  • What AI is good at vs poor at
  • How framing affects results
  • When regeneration helps vs wastes time

Concrete upgrades

  • Reuse AI for the same task
  • Save prompts that work
  • Add light structure: context → task → output

Signal

“I know when AI will help before I ask it.”

❷ Practitioner → ❸ Integrator

(From isolated prompts to full-task use)

Goal: Embed AI across an entire task.

What to learn

  • Use AI across multiple steps (research → draft → refine)
  • Separate AI suggestions from design decisions
  • Maintain traceability and accountability

Concrete upgrades

  • Reuse prompts intentionally
  • Note where AI was used and reviewed
  • Explain why outputs were accepted or rejected

Signal

“I can clearly explain what AI contributed — and what I decided.”

❸ Integrator → ❹ Architect

(From personal workflow to reusable system)

Goal: Make AI-assisted work reusable by others.

What to learn

  • Turn prompts into patterns
  • Design review habits around AI output
  • Teach judgment, not tricks

Concrete upgrades

  • Prompt libraries organized by task
  • Simple review checklists (accuracy, tone, accessibility)
  • Examples of good vs risky outputs

Signal

“Others can use my prompts and get similar-quality results.”

❹ Architect → ❺ Steward

(From reuse to responsibility)

Goal: Set standards for responsible AI-assisted design.

What to learn

  • Where AI introduces risk (hallucination, bias, IP)
  • How governance can enable speed
  • How to mentor without prescribing

Concrete upgrades

  • Clear guidance on acceptable AI use
  • Review norms for AI-assisted work
  • Coaching on judgment and accountability

Signal

“AI use is trusted here because expectations are clear.”

The Transition: L1 → L2 (The First Production Shift)

This is the moment designers stop asking AI to help

and start asking it to build.

What changes at L2

  • Prompts become instructions
  • Outputs are expected to be usable
  • Quality moves from intuition to checklists

Mental shift

From “help me think about this” → “build this bounded thing”

L1 forges judgment.

L2 operationalizes it.

How Designers Move from L1 → L2

Step 1: Identify Safe-to-Automate Chunks

Designers automate bounded units:

  • UI components
  • Copy blocks
  • Small flows

If it has clear inputs and a clear “done,” it’s a candidate.

Step 2: Turn Prompts into Instructions

L2 requires less creativity, more clarity:

  • Context
  • Constraints
  • Output format
  • Acceptance criteria

This is where many designers stall.

Step 3: Add a Simple QA Loop

Before integration, run the same checks:

  • Design system fit
  • Accessibility basics
  • Tone and hierarchy

No eval frameworks yet — just repeatable review.

Step 4: Reuse, Don’t Re-Prompt

Reusing prompts intentionally is the moment you leave L1.

Signals

  • You copy/paste on purpose
  • You expect similar quality each run
  • You know how to fix failures

L1 → L2 in E-P-I-A-S Terms

  • L1 Steward → L2 Explorer

You stop optimizing prompts and ask:

Which parts of my work can AI reliably build?

  • Carry-over skill: judgment

Strong L1 maturity makes L2 safer and faster.

An S-Steward at L1 is better prepared for L2 than an E-Explorer chasing advanced tools.
← The E-P-I-A-S x SAE Framework L2 to L3 Transition →