AI News
01 Jul 2026
Read 10 min
Designing with AI as material: How to preserve intent
designing with AI as material helps designers preserve human intent by shaping adaptive, clear systems
Designing with AI as material means you shape it like wood or clay, not install it like software. Treat AI as a living, shifting medium. Set intent, rules, and feedback. Add healthy friction so people keep thinking. This approach protects brand voice, trust, and human judgment.
AI is changing not only how we design, but also what design is. Ioana Teleanu, founder of AI-R studio, argues that we should treat AI like a creative material. It can adapt, surprise, and make mistakes. Our job is to give it intent and guardrails. When we do, we avoid bland results and keep human purpose at the center.
From artifact to behavior space: Define what “good” looks like across many outcomes, not one screen.
From screens to protocols: Write intents, guardrails, and escalation rules that AI must follow.
From certainty to probability: Plan for variance, edge cases, and fallbacks.
From pixels to semantics: Capture the user’s meaning, not only their clicks.
From QA to governance: Monitor, measure, and refine outputs over time.
When designing with AI as material, name the levers you will tune (temperature, tools, context windows), the limits you will not cross (privacy, safety, brand), and how people can correct the system.
Intent taxonomies: Describe user goals, constraints, and trade-offs in plain language.
Guardrails: Set red lines, allowed tools, and handoff points to humans.
Protocols: Define how data flows, how the model reasons, and how it explains itself.
Evaluation: Create test sets, score outputs, and track drift across releases.
Recovery: Plan safe defaults and one-tap fixes when AI goes off-track.
High-stakes actions: Add a clear review step before money moves, data posts, or rights change.
Confidence and why: Show a confidence score and a short reason for any AI suggestion.
Reading rhythm: Use layouts that slow skimming and invite focus for long-form content.
Explain and edit: Let people inspect sources, edit prompts, and retry with one tap.
Escalation: Route hard or risky cases to a human by design, not as an afterthought.
Write a one-sentence problem statement and a list of “non-goals.”
Define users, their stakes, and what “right” means to them.
Create an intent taxonomy with examples and counterexamples.
Document brand voice with do/don’t phrases and tone sliders.
Specify guardrails: banned content, privacy needs, and approval gates.
Choose tools and data sources the model can use, and which it must not.
Design prompt patterns, context rules, and system messages that encode intent.
Plan fallbacks for low-confidence outputs and empty results.
Assemble a test set that reflects brand, culture, and edge cases.
Score outputs for accuracy, safety, and voice fit, not only speed.
Log user corrections and feed them back into the system.
Document how designing with AI as material changes your release and review process.
An appetite for experimentation and the patience to refine.
A clear editorial voice and a view on culture, not just trends.
Ability to explain trade-offs in plain words.
Comfort with metrics, logs, and evaluation datasets.
Ethical instincts and the courage to add friction when needed.
Preserve intent: Start from human goals, not model features.
Design the space: Write rules and bounds that shape many good outcomes.
Show your work: Let users see why AI decided, and let them challenge it.
Add friction with care: Slow down where judgment matters.
Protect voice: Encode culture, tone, and risk appetite into the system.
AI will keep moving fast. You do not need to chase every release. You do need to keep asking better questions and setting better rules. By designing with AI as material, you protect agency, keep your brand alive, and guide a flexible medium toward human ends.
Designing with AI as material: from pixels to behavior
Most digital design used to be static. A button was a button for every user. AI makes the “material” of design probabilistic. It shifts by input, context, and data. You no longer ship a fixed artifact. You shape a space where the system can act with purpose. At this point, the designer translates human goals into machine behavior. That work includes intent, language, and context. For example, instead of only copying pixels between apps, an AI can read the meaning of the data and place it in the right fields. The value comes from the translation, not the motion.Key shifts you should expect
Non-deterministic systems call for new design work
Titles like “UX Designer” do not cover the job anymore. Teams now define behavior spaces, write intent taxonomies, and set rules for when AI should act or ask. The workload shifts from drawing final screens to writing the logic that shapes many possible screens and replies.Core activities for modern design teams
Intentional friction builds authenticity and trust
For years, we chased “frictionless” flows. That trained people to act on autopilot. With AI, this can lead to “cognitive capitulation,” where users accept outputs without thinking. Good design now adds gentle speed bumps that keep people present and in charge. This also helps brand voice. Average, instant outputs feel generic. Real brands have tension, quirks, and point of view. They take aesthetic risks. They speak from a culture. Friction leaves room for that voice to show up and for people to choose with care.Where to add useful friction
A practical framework to preserve intent
Use this starter playbook to guide AI work without losing your values.Set the purpose
Shape the behavior space
Build the protocol
Measure and govern
Hiring and culture for the AI era
Polished screens matter less now. Taste, judgment, and curiosity matter more. Look for people who can hold ambiguity, test fast, and still care about craft. They should protect analog time, read widely, and ask “should we?” before “can we?”Signals to seek in candidates
Case notes: extending habits, not replacing them
Strong AI products often respect old gestures and make them smarter. One example expands copy-and-paste so the system reads meaning and moves data into the right places across apps. It succeeds because it keeps a familiar habit, adds semantics, and protects user control.Principles to keep your design human
(Source: https://www.designboom.com/technology/ioana-teleanu-ai-artificial-intelligence-tool-interview/)
For more news: Click Here
FAQ
Q: What does “designing with AI as material” mean?
A: Designing with AI as material means treating AI like a medium you shape rather than software you simply install. It frames AI as a living, probabilistic substance that adapts and can make mistakes, so designers set intent, rules, and feedback loops. Designers also add healthy friction to keep people thinking and to protect brand voice and human judgment.
Q: How does AI change the nature of digital design artifacts?
A: AI turns static artifacts into probabilistic behavior spaces where outputs shift with input, context, and data. Designers now define protocols, intent taxonomies, and the translation between human goals and machine behavior instead of shipping a single fixed screen.
Q: What practical steps should teams take to preserve human intent when working with AI?
A: Start by writing a one-sentence problem statement, listing non-goals, and defining users and their stakes. Shape the behavior space with an intent taxonomy, documented brand voice, guardrails, prompt patterns, and fallbacks. Measure with test sets, score outputs for accuracy, safety, and voice fit, and log corrections as part of designing with AI as material.
Q: How can designers add intentional friction without harming usability?
A: Add friction selectively for high-stakes actions by requiring clear review steps, showing confidence scores with brief explanations, slowing reading rhythm for long-form content, and letting users inspect and edit suggestions. When applied thoughtfully, these speed bumps preserve agency and brand voice without needlessly degrading everyday usability.
Q: How should teams measure and govern AI outputs over time?
A: Assemble a test set that reflects brand, culture, and edge cases, score outputs for accuracy, safety, and voice fit, and log user corrections to feed back into the system. Monitor for drift, plan safe defaults and recovery flows, and treat governance as ongoing rather than a one-off QA pass. Document release and review processes to reflect how designing with AI as material changes iteration cycles.
Q: What skills and signals should hiring focus on in the AI era?
A: Prioritize taste, judgment, curiosity, and the ability to hold ambiguity over purely polished screens. Look for candidates with an appetite for experimentation, a clear editorial voice, comfort with metrics and evaluation datasets, ethical instincts, and the patience to refine ideas.
Q: What is an intent taxonomy and why is it important?
A: An intent taxonomy describes user goals, constraints, trade-offs, and provides examples and counterexamples to clarify correct behavior across many outcomes. It guides guardrails, prompt patterns, and evaluation so the AI acts in alignment with human purpose.
Q: Will AI make brand output generic, and how can brands remain distinctive?
A: Systems optimized for statistical averages often produce clean but generic results lacking tension, attitude, and quirks. Brands stay distinctive by encoding voice, cultural perspective, and aesthetic risk into guardrails and by adding intentional friction that invites judgement and choice. Designing with AI as material means shaping those boundaries so outputs reflect a specific brand rather than a neutral average.
Contents
