Build your AI Design practice. Design better human-AI teams. Lead intentional adoption. — frameworks, practices, and prompts tested in d.school workshops and live product work.
Unlike previous materials, it introduces three distinct relationship layers that designers must hold simultaneously:
Using tools, prompting, conversing with the material
Creating products and experiences around AI capabilities
Shaping model behaviors, values, and interaction defaults
Most practitioners are working in the first layer — it's where the tools are, and where the learning curve is steepest right now. The second is where product teams are placing their bets. The third is active at frontier companies, but design's voice in it is still finding its footing.
Start where you have agency. A practitioner who goes deep on working with AI changes what their team believes is possible. A team with strong shared norms changes what their organization can actually sustain. Each layer you develop creates real pull toward the others.
Designers need a more proactive stance across all three. Someone has to hold the system actors in view — the practitioner, the team, the org, the model. That’s system design work. And it’s exactly the work we’re trained for as designers.
Where this comes fromResearch, teaching, and building — across d.school workshops, enterprise product work, and too many failed AI rollouts to count.
Articulate why the work exists before you choose any tool. Purpose tells you which steps should be AI-assisted and which must remain human, and it survives every model swap that's coming (and a lot is coming).
The smell testIf the project becomes pointless without one specific model, the project was always about the model.In practice → Start-with-Why Workflow: ask why the workflow exists before adding AI. JTBD framing reveals which steps to touch — and which to leave alone.Users don't need a slicker interface for AI. They need grounds to trust it — transparency, role clarity, consistency, and the courage to say "I don't know" out loud.
Adoption metrics measure compliance. Trust is what makes adoption stick when the novelty wears off (about week six).
In practice → Accountability by Design: name a human owner for every AI-mediated decision class. Publish accountability maps. Build escalation paths into the product itself.AI surfaces possibilities; humans decide. Make this distinction explicit in every product, every workflow, every interface. Ambiguity here is corrosive — and it tastes like the future for about a quarter, until it tastes like a lawsuit.
The cleanest AI products are the ones where you always know whose call it is.
Shaw and Nave (2025) call it cognitive surrender — adopting AI outputs with minimal scrutiny, overriding both intuition and deliberation. In their experiments, accuracy fell when AI erred, but confidence rose regardless. People felt more certain while being more wrong. That's a design problem: the systems feel trustworthy while quietly eroding the judgment they were meant to support.
In practice → Human-in-the-Loop Moment: design explicit points where AI presents options and humans decide — earlier than feels natural, with a recorded human choice at each one.When AI automates practice, humans unlearn. Design for skill retention, not just task completion. The best AI systems make people more capable over time, not less.
Underreported riskWhat you can no longer do without help is your team's quietest liability.In practice → Skill Preservation: list 3 tasks where slow practice grows your craft. Block time where AI is off-limits. Track what you can no longer do unaided as a leading indicator.Industrial designers converse with physical materials. We need to converse with AI through making, not just prompting — by sketching with it, not at it.
The point of vibe-coding isn't speed. It's that we can finally sketch behavior, the way we always could sketch form.
In practice → Vibe Coding as Sketching: prompt for behavior, not finished code. Iterate fast, throw most away, annotate prompts like sketch captions. Hand the strongest to engineering as a brief.At org scale → AI Integration Rituals + Capability Metrics: sustainable integration needs cultural design — new norms, rituals, and measures that go beyond throughput to track what humans can still do on their own.Why three?Because AI integration leaks if you patch only one layer. Individual practice, team collaboration, org architecture — three altitudes, all required, none optional.
How you work with AI one-on-one — the choreography of a single craftsperson and a tool that thinks (sometimes).
If your craft has a "this is how I think" inner monologue, this is the scale where it stays alive or quietly dies.
Choreograph who leads, who critiques, when handoff happens.
Natural language as design material. Try the embedded prompt.
Live demo insideProtect practice time before deskilling happens to your craft.
How a group collaborates with AI as a member — not a tool, not a colleague exactly, but a presence that needs a charter.
If nobody on your team knows whether AI is a teammate or a microwave, this is your scale.
Re-form purpose, role boundaries, trust when AI joins.
JBTD framing reveals which steps AI should — and shouldn't — touch.
Explicit decision points where humans hold the call.
How culture, metrics, and structure decide whether the patterns above can sustain — or whether they quietly die in Q4.
The scale where good intentions go to be re-orged.
Explicit ownership for AI-generated outcomes at every layer.
New norms and rituals replace one-time rollouts.
Measure judgment, mastery, retention — not just throughput.
A working bibliography. Each entry has earned its place by sharpening one of the patterns above. Submissions invited; defensiveness is not.
| Scale | Title | Source | Type |
|---|---|---|---|
| Individual | Five Designerly Ways to Prototype Human–AI Collaboration | Design Meets AI | Essay |
| Team | People + AI Guidebook | Google PAIR | Guidebook |
| Individual | GenAI UX Design Patterns | Agentic Path | Pattern Library |
| Org | Humane by Design | Jon Yablonski | Principles |
| Team | Shape of AI | Emily Campbell | Pattern Library |
| Team | Your AI Teammate — Designing Work in 2030 | Stanford d.school | Workshop |