"Should I be using AI?" is the wrong question. The right one is: where do I start if I'm still not sure?
I understand why it keeps coming up. There's a lot of noise. New tools every day. Conflicting opinions about which ones matter, which ones are hype, and whether any of it is worth the time it takes to learn. The landscape is genuinely confusing, even for people paying close attention.
But the answer isn't complicated. Yes. Start now. If you haven't, you're already behind.
That's not meant to be alarming. It's just true. And the good news is that "behind" is a recoverable position. Almost everyone using these tools right now is still figuring it out. The gap between someone who started six months ago and someone starting today is smaller than it feels. But the gap between someone who has started and someone who hasn't is real and it's growing.
The fear is the wrong question
Most of the hesitation I see isn't laziness. It's fear.
Fear that the tools are too technical. Fear of looking stupid in front of the screen. Fear that you'll invest time learning something that changes completely in six months. And underneath all of it, the bigger fear: if AI can do what I do, what happens to me?
These are all reasonable things to think about. None of them are a good reason to wait.
The tools are learnable. Looking stupid at the screen is part of learning anything. The tools will change, but the underlying skill of knowing how to work with AI doesn't disappear when a model updates. And the question of what happens to designers is answered much faster by getting into it than by watching from the outside.
Start. That's step one. Everything else comes from there.
Where to actually start
My recommendation is Cursor. Not because it's the only option, but because it's model agnostic.
That distinction matters more than it sounds. Cursor lets you switch between models from Anthropic, OpenAI, Google, and others. You're not locked into one company's approach. This is useful right now because different models genuinely excel at different things, and learning which is which is part of the education.
Here's the rough breakdown as it stands today:
For planning and thinking through architecture, Opus and Sonnet from Anthropic are the strongest. If you're starting a new project and need to think through structure and approach before writing any code, this is where to start. They reason well, ask good follow-up questions, and help you avoid the mistakes that come from building without a plan.
For building: Codex and GPT-5 from OpenAI are currently the best at execution. Particularly for longer running tasks, building out components, writing the actual code that does things. If Anthropic's models are the senior strategist, OpenAI's are the builder.
For copy, research, and marketing: Gemini from Google is consistently strong. If you need to write the product copy that goes into the UI you're building, or research a topic before you design for it, Gemini handles this well.
For repetitive, lower-stakes frontend tasks: Grok Code is cheap and surprisingly capable. Minor polish, small fixes, repetitive UI adjustments. You don't need to burn expensive model credits on these. Grok handles them efficiently at a fraction of the cost.
You don't need to memorise this on day one. But knowing it exists means you can reach for the right tool instead of using the same model for everything and wondering why the results are inconsistent.
The workflow that actually matters
The goal isn't to use AI to design. It's to build and deploy things.
This is the shift that matters. Not "use AI to make better mockups." Use AI to go from idea to working, deployed, shareable product. The loop is: build something, get it live, show it to real people, learn from what happens, build the next version.
If you're starting from Figma, the first exercise is to take something you'd normally design there and try to build it directly in Cursor instead. A landing page. A component. A small web app you've had in your head. The point isn't that the first version will be good. It won't be. The point is to feel what the loop is like.
Once you've built something, deploy it. Vercel is the easiest option for most web projects. Put it on a domain, share the link. Something about making a real URL exist focuses your thinking. It's not a prototype anymore. It's a thing.
Then iterate from there based on what you learn.
One honest observation
Everyone using these tools right now is a beginner in at least some sense. The people who started two years ago have more practice. But the tools themselves are new enough that the patterns are still forming. Nobody has fully figured out the optimal way to work with AI because the AI keeps changing.
This is unusual. Most skills have established paths. There are experts, there are beginners, and there's a clear direction of travel. AI workflows don't have that yet. The person who's been doing this for a year is figuring it out as they go just as much as the person who started last week. Just with more practice at figuring it out.
That's actually an advantage for anyone starting now. You're not late to a mature field. You're early to an evolving one.
The entry point is just: start. Download Cursor. Build something small. Break it. Fix it. Deploy it. See what it tells you.
The fear goes away when you're in the middle of actually doing it. That's true of most things, but it's especially true of this.
