How designers use AI without becoming dependent on any one tool, or mistaking output for finished work.
I've stopped being surprised when a new AI tool makes last month's setup feel outdated. What still catches me off guard is how many designers are chasing the tool itself, as if fluency in one interface is the differentiator.
It isn't. It never really was.
With design tooling, we used to get clear market winners over time; Adobe, then Sketch, and InVision, then mostly Figma. Each one settled in, and we governed our process around what it could offer us.
AI tooling doesn't behave that way. We were deep in Figma Make most evenings until the tokenisation limits hit, scrambled into Claude, then started eyeing Claude Code, with GPT and Gemini running consistently in parallel. We ran weekly design jams just to keep pace, and half the time nothing stuck, not because the tools were bad, but because something better always showed up before we'd finished learning the last one. Chasing "the" tool is a losing game right now.
It's calmer today. We've settled into individual choice over team mandate. People use what suits them, and I stay unattached to any one platform on their behalf.
The AI tell
It feels like you can spot AI-generated design the way you can spot AI-influenced writing. There's a “sameness” to it with the use of patterns that models keep reaching for, and once you've seen enough of it, you recognise the fingerprints.
In that way, generated doesn't mean designed.
AI-produced UI can look entirely convincing while being strategically wrong, technically unbuildable, or completely off-brand. Spotting that requires taste, judgment, accessibility awareness, and product context, none of which the model inherently has, and all of which the designer is still accountable for. If we ship it, we own it. AI doesn't get to sit in the room when something goes wrong, or answer to the commercial and customer backlash. We do.
The tool feels less important than the designer's ability to understand the medium.
The transferable skill isn't fluency in one interface. It's knowing how to work with models at all - how to prompt with intent, structure context, manage constraints, and evaluate what comes back critically instead of gratefully. Once a designer has that foundation, moving between tools is just syntax.
Used well, AI is genuinely useful for stretching designers into adjacent territory. It can help someone who isn't a data analyst interpret a raw dataset. It can help someone who isn't a business analyst unpack a requirements document into journey implications. It can help someone who isn't a front-end developer understand where their design system will hold or break. That's real value, and it closes gaps faster than most upskilling plans can.
Where it gets shakier is high-fidelity, production-grade work, especially inside archaic ecosystems and large organisations. Early concepting, synthesis, low-fidelity exploration - that's where AI shines for designers. Inside a legacy system, without a mature design system, or within a regulated environment with a hundred edge cases, it gets much harder to lean on meaningfully, and the risk changes shape.
What we still don’t know
That's the real question underneath this whole conversation, and I don't have a tidy answer to it. I'm not sure anyone does, yet.
Design has always run on reference: looking at what exists, and reshaping it into something new. AI has put that process on a much faster loop, and it's forcing a harder look at what we actually own in the creation process.
On the flip side, something similar is happening in how our work gets used. We've spent our careers designing for people to use what we make. That's no longer a guaranteed truth, because agents are starting to do some of that usage on a person's behalf. We don't know yet how far that goes, but it's a real enough possibility that it's worth paying attention to now so we can meet it head-on.
What is the definition of design in this new normal? I don't think we get to skip this question because it's uncomfortable, or because it doesn't resolve neatly in a single viewpoint. It's one the industry needs to sit with over time, and reshape our understanding around.
Concluding thoughts:
None of this comes down to which tool you've mastered, or how fast you adopted it. That's the version of relevance I've stopped chasing, and I don't think it was ever the real one.
What's actually proven itself is judgment: knowing when generated work is genuinely useful, and when it's wrong. Knowing where AI opens a door into disciplines a designer wouldn't otherwise have easy access to, and where it should stay in ideation, not production. And being honest enough to sit with the harder question this raises about what design even is now, instead of skipping past it because it's inconvenient.
That's still uniquely human. Better individual use of AI is only the beginning. The larger opportunity is what happens when we connect that capability to the systems an entire organisation uses to create. That's a thread I'll pick up in the next piece in this series.
#Design #AI #ProductDesign #DesignLeadership #CreativeTechnology
