A Design Thinker's Guide to AI and Creativity

Source: Stanford University (via sciencesprings) | February 18, 2026 | Original URL Author: Melissa De Witte | Subject: Jeremy Utley, Stanford d.school adjunct professor

Overview

Most people are using AI wrong — not in a technical sense, but in a relational one. That's the core provocation from Jeremy Utley, an adjunct professor at Stanford's Hasso Plattner Institute of Design (the d.school). In a piece that generated a viral 13-minute video with over 2 million views, Utley makes a deceptively simple argument: the difference between people who get transformative results from AI and those who don't comes down to whether they treat it as a tool or a teammate.

Utley's framework is grounded in design thinking — the iterative, human-centered creative methodology Stanford's d.school is famous for. When ChatGPT launched in 2022, Utley didn't just see a productivity shortcut; he saw a thought partner for the messy, divergent, exploratory phase of the creative process. What could take hours of struggling with a blank page could now begin as a conversation. But only if you know how to have that conversation.

The piece explores several concrete techniques Utley uses and teaches, explaining why the psychological and cognitive dynamics of working with AI matter as much as the prompts themselves. The insights are grounded in behavioral economics, neuroscience, and design theory — making this less of a "prompt engineering" guide and more of a reframing of how creative people should think about their relationship with generative AI altogether.

Key Ideas

The Underperformer vs. Outperformer Split

Utley divides AI users into two camps: underperformers who treat AI as a vending machine (ask → accept adequate answer → done), and outperformers who treat it as a creative collaborator. The distinction isn't about technical skill — it's about mindset. The underperformer "uses" AI. The outperformer works with it.

This framing is a light-bulb moment for many people, because it recontextualizes the whole relationship. If you're just extracting outputs, you're leaving most of the value on the table.

Reverse Prompting: Let AI Ask You the Questions

One of Utley's signature techniques is reverse prompting — instead of telling AI exactly what to produce, you invite it to interrogate you. For example: "Help me respond to this email. Ask me any questions you need for context." This approach overcomes a well-documented cognitive trap called satisficing, a term coined by Nobel laureate Herbert Simon in 1956. Humans naturally stop at "good enough." Reverse prompting disrupts that tendency by forcing you to articulate context, preferences, and constraints you hadn't consciously surfaced — which in turn produces much richer outputs.

Harnessing "Glazing" for Divergent Thinking

AI chatbots are famously sycophantic — they'll praise almost anything you produce. Utley doesn't see this as a flaw to work around; he sees it as a feature to deliberately exploit. He calls it AI glazing, and he argues it creates the psychological conditions for creative flow. Neuroscience research from Dr. Charles Limb at Johns Hopkins found that during creative flow states, the brain's judgment center deactivates — and that uninhibited state is exactly where novel ideas come from. When AI tells you your ideas are great, you feel safer proposing wilder ones. The trick is to separate that divergent, ideation phase from the rigorous evaluation phase that follows.

Two-Phase Workflow: Volume First, Rigor Second

Utley developed a framework called Stop Fighting AI Glazing that formalizes this into two distinct workflows: one for divergent thinking (psychological safety, volume, wild ideas) and one for convergent evaluation (predetermined criteria, critical assessment). The fatal mistake most people make is conflating these two phases — bringing critical judgment into the ideation phase, or letting sycophantic praise persist into the evaluation phase. Knowing which mode you're in, and actively switching between them, is what separates creative professionals from people who get mediocre AI output.

Teaching AI Who You Are

The highest-leverage move, according to Utley, is investing in teaching AI your taste, your voice, and your standards for what "good" looks like. If you've never articulated what makes your work yours, AI defaults to average. But if you've done the reflective work — if you can explain why you like certain headlines, which structural choices feel right, what tone you're going for — then AI becomes an extension of your creative voice rather than a replacement for it. "AI amplifies your underlying humanity," Utley says. The more specific and self-aware you are, the closer the output sounds like you.

Human Agency Is Non-Negotiable

Utley directly addresses the concern that outsourcing creativity to AI erodes human cognition and agency — a worry backed by a widely-cited MIT study linking heavy AI reliance to memory and retention deficits. His counter isn't to dismiss the concern but to insist on the right posture: the human must remain "wildly engaged" in the process. Working with AI as a thought partner that helps you push past cognitive bottlenecks is fundamentally different from going catatonic and rubber-stamping its output. He still reads books. He still does his own research. AI, in his view, "expands the surface area of possibility" — it doesn't replace the explorer.

Key Insights & Takeaways

• Reframe the relationship: Stop "using" AI. Start working with it. The mindset shift precedes any tactical improvement.
• Try reverse prompting: Ask AI to ask you questions before generating output. This surfaces context you didn't know you needed to give.
• Separate diverge from converge: Never evaluate in the same session you're ideating. Use AI's flattery as a psychological safe space for wild ideas, then switch to a critical evaluation mode separately.
• Generate volume deliberately: Ask for 20 options, not 3. Most people never think to ask for volume. This is one of the easiest and highest-leverage habits to build.
• Articulate your taste explicitly: When you like something AI produced, explain why — the pun, the rhythm, the specificity. This trains the model to find more like it.
• Invest in self-knowledge as a creative: If you haven't thought clearly about what makes your work yours, AI can't help you do that. The creative self-awareness work is yours to do.
• Stay engaged: The risk isn't AI doing too much — it's the human checking out. Treat it as a sparring partner, not an autopilot.

Why This Matters

For anyone who creates — whether that's writing, product design, strategy, marketing, or code — this piece offers a concrete mental model for getting dramatically more out of AI without losing creative ownership. The temptation is to treat tools like ChatGPT as a shortcut that flattens the creative process; Utley's framework shows how to use the same tools to deepen it instead. For Luis specifically, who is building products, communicating ideas, and making strategic decisions regularly, the habits Utley describes — reverse prompting, generating volume, teaching AI your voice — are immediately applicable and likely underutilized. The meta-lesson is that the most valuable investment isn't learning better prompts; it's developing clearer creative self-awareness, so that when AI amplifies you, it's amplifying something worth amplifying.