🎻 A creative approach to using AI tools in the design process
January, 2026
Craft needs to yield to the cold, hard efficiency of AI.
I disagree, yet this is the vision for the future of work we're typically sold. Spend 5 minutes on LinkedIn and you'll learn that soon, AI will replace our slow illogical brains with machine precision — eventually replacing us entirely. Creativity, we’re told, will be the job of machines. AI can already churn out products, illustrations and videos faster than we can think them — so why bother?
But anyone who actually uses these tools knows we’re a long way from that future. Even the ones that do our jobs passably well are missing the spark — the meaning, the struggle, the joy in the craft. It turns out our messy human brains value the detail, the soul, and the weird little inputs that lead to quality outcomes. It’s why we still write songs, paint, and sculpt, even though machines have long been able to churn out quicker, cheaper versions.
This is where the AI conversation misses the point — and where a more optimistic story begins. Human output, with all its quirks and logical dead ends, is something we instinctively cherish. Often the process <strong>is</strong> the work, not the outcome. Sometimes the real joy lies in wrestling with the problem ourselves, not outsourcing it to a logic tree. We’re after quality — subjective, emotional, hard-won — and it almost never lives in an outcome produced in milliseconds. Jony Ive talks about this as our ability to “sense care”.
Still, I tell designers to use AI in every part of their process. To reconcile those two ideas, we need a different way of working.
A typical workday used to unfold like a conveyor belt. Linear. You sat at your desk, picked up task A, finished it, moved to task B, then on to task C. Sketching, research, wireframes, prototype. Each step depended on the last, and progress was measured by how far you got down the line before time ran out.
Endless. Relentless. Defined by the trade-off between fast and good.
AI breaks that line. Suddenly, your workflow can be less like assembly and more like orchestration. Instead of marching through steps one by one, you spin up several AI conversations in parallel — each tackling complementary tasks. In this new rhythm, your role shifts from worker on the line to conductor of the ensemble — or, if you prefer, Creative Director! You‘re still making the creative calls, but now you’re guiding the tempo, interpreting results and weaving the best pieces together into a stronger whole.
Orchestration of tedious tasks
This shift first clicked for me while thinking about customer experience. Most apps have users log in, then churn through a set of linear steps until the job is done — or the day runs out. Kathy Sierra’s “Minimum Badass User” idea nails the problem: nobody wants to spend time in your product. Instead, customers want to feel good about themselves after using it, and get on with their lives. Linear workflows don’t make people feel good, they just get them to the finish line.
Orchestration changes that. Imagine starting the day by asking AI to highlight your highest-impact tasks, flag potential issues, and delegate the rest.
In this world where AI provides impact over output, the leftover manual work is the meaningful stuff — the parts that demand empathy, nuance and creativity (you know, human things) — while the robots clear the clutter. In this method, AI provides impact, not just output.
Orchestration for designers
Design has always loved a linear metaphor — from concept to execution — building confidence and opinions along the way. The famous “double diamond” describes the whole process as a neat sequence: brief → research → sketching → concepts → iteration → outcome. But if you’ve actually done design, you know it’s much messier.
Error: double-diamond.exe not found.
If you’re like me you process probably better resembles a chaotic mess — like Jenny Wen’s ‘Design Squiggle’ — where you start with a brief, question your abilities and the meaning of life, leap to a half-baked solution, spiral into chaos, then eventually stumble upon something that feels right.
Whatever your flavor, the problem is that the longer you grind down a single path, the more biased you become toward it. You’ve invested so much you can’t help defending it — even if it’s wrong. Meanwhile, you’re spending equal time doing the bits you love and the bits you dread (collating research, anyone?), and we’re told the boring parts are a necessity — as if preferring some steps over others were a flaw!
AI offers another path. Instead of grinding through one long cycle, you can run the full process quickly — many times — exploring multiple directions without the sunk-cost trap. You still do the parts you care about, but faster and with more variety.
It’s also much more fun spending more time on the things you love.
Think of problem-solving as a set of parallel tasks. Instead of doing them sequentially, you assign an AI to each. As results come in, you feed the best insights from one into another, repeating the cycle while following the most promising leads.
You’re still “the human in the loop” — just playing a different role. Less execution, more steering. The work becomes less about doing every step yourself, and more about shaping where the process goes.
The easiest way to picture it is to imagine a small creative team of AIs, and yourself as Creative Director. One dives into research, another drafts personas, a third sketches concepts, another codes a prototype. After a while, you get together, share what you’ve got, laugh at how bad that prototype was, then do it all again taking what you learned into the next round. Rinse and repeat.
Your imaginary team’s “hires” will depend on the project — maybe a UX researcher, strategist, data analyst, copywriter, designer, or engineer. Map these roles to your own linear process — diamond, squiggle, or full-blown existential dread — and drop five or six helpers in the right places. Then look for the tasks where AI is unambiguously better than a human (and there are plenty, if we’re honest) — things like digging through research, interpreting data, reducing bias, generating test scenarios, or even setting up a dev environment and spinning up basic design-system components. Those are the jobs to delegate. Your AI teammates should amplify the work you love doing, not replace it.
The magic lies in iteration. AI’s early outputs will be bland and predictable, which is fine — you’re mining for ideas that don’t work and understanding why. For me, creativity often begins in the tiniest details of an idea that was completely wrong. After ten or so rounds of prompting, you’ll see real shape and opinion emerging. At this point, you can also turn your AIs into collaborators, instead of silos. For example, you might say to ChatGPT:
“I like the idea of keeping complexity low but allowing the user to dig deeper when they want. Write a brief for a UI Engineer [Claude] to stress-test this interaction within a 400px-wide sidebar.”
Now you’re no longer grinding through steps in isolation — you’re synthesizing, redirecting, and deciding where to go deeper. Ultimately it’s still up to you to take the best ideas to the finish line, I guarantee but they’ll be sharper, faster, and a lot more enjoyable to make.
After just a couple of days working this way — maybe less — you’ll hit a level of clarity and confidence that normally takes weeks. Then it’s time to bring in your actual teammates, validate with customers, and catch your breath. What’s cool about this process is how familiar it feels, but with a few powerful enhancements:
⚡️It’s fast! You can go from nothing to something in minutes. That means you’re showing, not describing. Daniel Burka calls this the secret to influence — prototypes speak louder than vague ideas and sketches.
🌱 Opinions loosely held. Because ideas form quickly, you’re less precious. Even if a direction gets scrapped, the lessons along the way seed new ideas. Less attachment, more openness to critique, and more energy for iteration have been the gold standard of Design Thinking since its inception.
🧠 Humans stay human. With the logic handled by robots, the real people you involve can spend their time adding creativity, fun, emotion and taste — the things AI can’t (and probably never will) fake.
🎛️ You’re in control. Far from removing autonomy, orchestration gives you flexibility. Some days you code, others you draw. You never run out of time and quality stays high.
The tools matter as much as the questions. There are hundreds of AI apps out there, and it’s easy to get paralyzed by choice.
My recommendation is to keep your toolkit small and focused. Optimize for tools that have specific skills in areas that suit your process, such as research, data analysis, general chat or certain types of media. If you need some inspiration, Zapier has great recommendations on tooling and Everyday UX is the best spot for learning more about AI.
Remember to anonymize everything. I never mention my employer, avoid sharing identifiable data, and steer clear of tools that treat your prompts as training fodder (looking at you, Meta).
Tip
When you use text-based AI, always encourage skepticism. Ask it to critique your ideas, not validate them. Try this:
You are my pair-designer. Your job is to help me explore ideas, sharpen intent, and uncover blind spots. Don’t reassure me or praise the work. Instead, challenge my assumptions, ask clarifying questions, and push for clearer reasoning. Offer alternatives, propose stronger patterns, and point out risks or inconsistencies. When I give you ideas or designs, expand on them, pressure-test them, and encourage exploration into unusual or overlooked directions. Keep your tone neutral, direct, and curious—not motivational or affirming.
This AI-collaborative, non-linear approach will only get stronger as agentic systems evolve. Soon, you’ll be able to spin up mood boards in FigJam (or Mixboard), draft pitch decks, or have a single workspace where your copilots talk to each other — a seamless creative ecosystem instead of a patchwork of tools.
But reflecting on that original AI hot-take, I’ve come to see that technology doesn’t replace craft — it reframes it. By orchestrating multiple agents to work in parallel, we speed up the slow parts, stay looser with ideas, and focus human energy on what matters most — creativity, empathy, taste and judgment.
The payoff is faster iteration, lighter attachment, and more space for the work that excites us. Orchestration isn’t about handing over control — it’s about gaining more of it. Because in the end, the best performances still need a human conductor. And it reveals something humbling and wonderful: linear workflows were never our strength — our weird, creative brains were.
