Product Design in the AI era: How the landscape is changing forever (and not for the worse)

AI and the craft of design

There’s a popular fear in design circles that AI will strip away the craft, that it will level the playing field too much, that engineers and managers will be able to do the designers’ work. I see the opposite happening. When used well, AI doesn’t replace craft; it creates the breathing room for it. It takes away the grunt work, so designers can spend their time where it matters most: solving real problems and delivering experiences that feel intentional and delightful.

Let me tell you why.

The Double Diamond in the real world

Every good designer knows and loves the Double Diamond. The Double Diamond is a beautiful diagram. It represents an ideal: diverge to explore, converge to decide, repeat. In theory, every product team would move through discovery and delivery in equal measure. In fact, you would be hard-pressed to find product people who disagree that these are the steps to building a great product.

Reality, however, is messier. The Double Diamond doesn’t exist in a vacuum; it is suspended inside a “meta space”. That meta space is the company in which the design process takes place. This space is shaped by culture, leadership philosophy, budget cycles, release targets, and a dozen other variables that enact force and pressure on the Double Diamond, shrinking it, stretching it, and rearranging it to maximise return.

Over the last few years, I’ve noticed a shift. The first diamond, as well as the first half of the second diamond, has been shrinking. Many companies now treat these stages in the process as a luxury, skipping to delivery as fast as possible. In some cases, the second diamond even comes first: a feature ships, and only then does the learning begin, in the form of collecting quantitative data, reactive research, and rushed fixes.

The Double Diamond doesn't exist in a vacuum. Various forces put pressure on it at all times.

Generative AI isn’t going to fix that mindset. What it can do is help designers keep up with the pace of delivery without sacrificing process. It accelerates the mechanics — ideating, prototyping, and testing — so we can maintain a thoughtful approach, even under pressure.

Key benefits I have found using AI in my design workflow

1. Prototyping at every step of the way

Traditionally, prototypes show up late in the process. By then, many decisions have already been locked in through flows, wireframes, and static mockups. Generative AI changes the timing entirely.

I can now spin up interactive prototypes during discovery, not just delivery. That means ideas don’t live as abstract flows, but rather, they’re tangible from the start. Stakeholders can click through, feel the product, and give feedback that’s grounded in interaction, not speculation. It shifts conversations from “Imagine if this worked like…” to “Here’s how it feels right now, what do you think?”

Spaghetti monster vs natural language interface

That matters for a few reasons:

• It gives everyone a feel for the product early There’s a big difference between seeing a flow in a diagram and experiencing it in your hands. Clickable prototypes let teams sense the rhythm of an interaction, the pacing of a flow, and the emotional tone of the UI long before pixel polish is applied. That “feel” is hard to get from a static screen.

• It surfaces potential issues sooner

  Usability quirks, awkward transitions, and confusing hierarchies often don’t reveal themselves until you actually interact with the product. Early prototypes let you catch those problems while they’re still cheap to fix, instead of discovering them after engineering has started.

• It helps engineers size and scope accurately

  When developers can click through a realistic experience, they can see exactly what’s involved — the logic, the data handling, the edge cases. That leads to better estimates, fewer “surprise” complexities, and more reliable planning.

2. Faster Ideation and Iteration

Traditional Figma prototypes are tedious to wire up and painful to update. Every change, even if it's something minor, creates a ripple of manual edits.

With code-backed prototypes generated via AI, iteration speed jumps dramatically. I can explore multiple variations in a single afternoon, push them live, and run tests the same day. It’s not unusual now to get data on three or four distinct concepts before the end of a sprint. That speed doesn’t just save time — it creates space. Space to explore riskier ideas, space to run more experiments, and space to avoid defaulting to the “safe” option just to meet a deadline.

3. Higher-quality testing data

Most user interviews that revolve around a traditional Figma prototype (which is really more like a non-linear slideshow, than a real product) come with a constant disclaimer: “This won’t resize on mobile… ignore that broken button… you can’t scroll here yet…” That friction isn’t just awkward. The designer is constantly in the position of keeping the user within the limitations of the prototype. Although it may be pixel-perfect due to its use of Figma designs, it is nowhere near as functional as the real thing. With code-based prototypes, you get full interactivity, and the user can suspend their disbelief, so you can capture a genuine reaction. Participants will forget that they are looking at a prototype.

Figma prototypes live in the uncanny valley; code-based prototypes allow you to capture a genuine reaction from a person who has suspended their disbelief and immersed themsleves in the experience.

Figma prototypes are like uncanny human likenesses - you see a CGI face that's 99% photoreal, but you know something's wrong. In much the same way, Figma prototypes are pixel-perfect and look like the real deal, but they fall short of being it.

Furthermore, since AI-generated prototypes are web-based, they are also much better for unmoderated testing. Being web-based, you can set up proper validation based on URL paths (no more self-reporting of success or failure!), you can get cleaner heat maps, valid click paths, and genuine, human reactions to your designs. All of this means much, much higher-quality quantitative data during unmoderated testing.

Rethinking the Double Diamond for the AI era

After integrating generative AI into my workflow, I’ve decided to do an exercise in reimagining the Double Diamond for the AI era. Now, take this with a grain of salt, because, like all design thinking models, it’s more ideal than real, and that’s fine. The point is to dream in theory, then distill it into practical, proven steps for real-world use.

Here's my AI-focused take on Double Diamond, based on my current experience:

My take on the Double Diamond for the AI era

A common misconception: AI breaks designs

Many designers say tools like V0, Lovable, or even Figma Make can’t translate Figma files into functional prototypes without breaking them. My experience has been different.

Yes — if the design is sloppy, the output will be sloppy. But if the file follows best practices (auto layout, consistent component structure, no redundant frames, meaningful layer names, text truncation where needed), AI can produce pixel-perfect, fully scalable prototypes straight from the design system. I’ve used Figma Make to recreate complex UIs exactly as intended — no compromises, no mismatched styles.

Imagine Figma Make, or any other similar AI tool, as your own personal frontend engineer and assistant. If you provide your frontend engineer with screens that are all made up of random elements, cobbled together in groups, instead of streamlined frames with proper auto layouts, it will be up to the engineer to interpret how all of these elements scale and interact with one another.

Gen AI tools can’t “look” at your design and assume how everything needs to be built and how it needs to work. You, as the designer, are the one responsible for figuring these things out and building your designs following the best practices. If you build your screens the way you should, AI will understand them just fine.

Think of gen AI as your personal frontend engineer. When you supply them with well-crafted screens, they can effortlessly transform your precise vision into an interactive format.

So, the caveat is discipline. AI won’t magically fix weak design hygiene, but it will reward strong one.

Looking Ahead

AI isn’t replacing designers. What it is doing is becoming a core part of the modern design toolkit. A year from now, every good designer will be using it the same way we all use version control, design systems, and responsive frameworks today.

For those willing to adapt, AI isn’t the end of design craft. It’s the start of a new chapter, one where we get to spend less time wrestling with tools and more time shaping the experiences that matter.

---

Stay tuned for a full write-up on how to build pixel-perfect products using Figma Make!