Collaboration Between Design and AI
That’s it—AI tools are now part of our daily routine at our e-health company. We designers are starting to get the hang of them, as are people in other roles, and that’s a very good thing! With tools like Lovable and Claude Design, generating a web interface takes just a few prompts, which really opens up a world of possibilities for everyone.
But let me say this right off the bat: an AI-generated mockup isn’t a final product—it’s first and foremost a starting point for discussion. And I’m not saying this because I’m afraid of being bypassed or because I think product managers shouldn’t get involved.
On the contrary. I say this because we’ve been experimenting with this at the Design Center for several months now, and the conclusion is simple: no matter how attractive and seamless a prototype may be, if the right questions aren’t asked early on, it’s meaningless for both the product and the user.
That’s why I wrote this manifesto—to try to lay the groundwork, together, for the future of product design in our e-health context.
AI: A Tool That Lets You Skip Steps?
Not quite. As we see in our company today, anyone spearheading a project or an idea would be making a big mistake not to take advantage of these tools. For a project manager who wants to quickly bring an idea to life, AI is unbeatable: it can turn something that’s been sitting in a document for weeks into a tangible reality in just a few minutes.
But there’s also everything it doesn’t do, and that’s where it gets important: verifying that the interface is consistent with our Design System, ensuring that it’s accessible, and asking ourselves whether what we’re building actually solves the right problem or fits properly into the overall user journey.
These aren't just finishing touches: they're truly at the heart of the product design process, and no tool can do this for us.
“Taste without accountability is mere decoration. Taste with accountability is leadership.” — Muhammad Dani Asyrofi, UX Designer
Explanation of the AI Product Design Process / Design by Martin Pontegnier Using Claude and Figma
AI tools like Lovable or Claude are really meant for exploration, first drafts, and prototypes that you put together before a meeting to win people over. Figma remains the tool for precision design, maintaining the design system, final UX validation, and accessibility reviews. And developers—whether working with Claude Code or the codebase—only begin work once a deliverable has been approved by the design team; they never start from raw AI output.
If someone asks, “Should I use an AI tool for this?”, my answer is yes when the goal is rapid exploration or a first draft, but not when it comes to precise interaction design or implementation.
Our Experience at Santexpo
In March, Clément and I prototyped the instant messaging feature on Lovable right before Santexpo. In just a few hours, we had something truly interactive. It wasn’t a series of clickable Figma screens, nor was it a deliverable that perfectly aligned with our Design System—it was a real, living experience that we could put directly into the hands of customers right there on the floor. And it completely changed the way we presented and sold the product.
Clément, PM:
For the biggest health expo of the year, we wanted to showcase this new solution we’re currently working on to our clients. But preparing a demo for that event would have slowed down and prolonged our developers’ work.
AI tools gave us the agility we needed and allowed us to produce a demo we could present to a client. We did all of this in just a few hours, using Martin’s mockups to ensure compliance with the style guide and Design System.
The feedback from our customers has been extremely positive, and we've been able to gather a lot of feedback and use cases to refine our vision.
What I take away from this is that AI tools and Figma really serve different purposes: one persuades, the other builds. Both are necessary, but not at the same time.
Just a word of caution before you get started: these tools use tokens that can be quite expensive, especially on Lovable. The cost can add up quickly, so feel free to ask the Product Design Team for best practices.
If we had to summarize
Before a screen goes to development, it must be approved by the design team—not to redo all the work, but to ask the right questions. We need to make sure that what’s being sent off is solid, that the screen meets a real need, that it fits properly into the user journey, and that it truly meets users’ expectations.
The goal is to avoid delivering 40% while thinking we've done 100%.