Onboarding Claude: What our recent hires are teaching us about the future of design

Bernadette has been prototyping some new team shapes, and we’ve recently brought in a few new hires. Alongside Google Stitch, and Figma Make, our newest recruit is Claude. They all had a strong portfolio, experience delivering complex experiences for unique brands and importantly had boundless energy to go above and beyond. But there’s something particularly exciting about Claude.

Most designers would baulk at the dreaded words ‘have fun with it’ – but by combining Claude Code, MCP servers and Claude Design – our new colleague has revelled in the challenge. When given even a throwaway prompt they surprised and seriously impressed us. Full interactive experiences spun up in seconds – vivid visuals and surprising interactions delivered in design and code created near instantly.

But there was a hitch. The reality is that these results didn’t always connect the dots coherently. They feel more like a party trick than an effective experience. There were inspired moments among the generated outputs, but Claude’s wider vision often missed the nuance required to get a brand expression spot on. In their eagerness to impress us they confidently hallucinate entirely new brand assets (on occasion this is something that should be encouraged) – but this eagerness could be their biggest weakness.

Like in human teams collaboration is key, and we’ve found augmenting tools like Gemini, Figma Make and Claude together have delivered the most promising returns – each have their own personality quirks which can work with or against you. What has become clear is that a collaboration is essential, even between machines.

Gemini’s canvas is highly effective but is ultimately single player. Figma Make promises seamless design and prototyping, but is let down by its attention to detail. Claude’s design suite (particularly when connected to Figma’s MCP) feels like magic but overshoots and builds in too much complexity. What is clear is these tools will become more effective as they become more collaborative and more reliable – the same qualities we often look for in our human colleagues.

The transformational unlock (right now at least) is in focussed, functional tasks. Like anyone else in the team, the more guardrails and support we give them, the more likely the outcome will resonate. Designers have been rightfully proud of the manual craft behind beautiful documentation and systems, but Claude can eat this for breakfast. This unlocks precious hours by handing off the manual drudgery of spacing and system maintenance. The tools can also act as a superpowered sketchbook – letting us communicate interactive snapshots of an experience instantly. Ultimately this unshackles human teams, giving them time to be genuinely inventive and expressive – and to think bigger than tokens and corner radiuses.

This is where the real value lives; in finding the creative tension between human, market and brand truths. In making connections between categories, concepts and data sets. Some of which are not public, some of which require an inspired leap between disparate ideas. This is where you find effective and surprising solutions. This is where you find an idea.

Here is the reality: the tools are radically changing how we get there, but the why stays exactly the same. Customers don’t care what built or powers their experience; they want an experience that works, they want to feel connection, inclusion and trust. But the sheer velocity of this toolkit should be a massive wake-up call for businesses. We can now operate at a pace unthinkable a few years ago.

It’s worth noting that Generative AI isn’t limitless magic. The true cost of compute is becoming clear. The brief (and inevitably short-lived) infinite compute cheat has been patched. Finite resources require the same clear direction as anyone else in our team. But running out of credits (just like burning out a team) is a real and increasing risk when working with these tools. We’ve seen companies like Uber and MIcrosoft drive adoption of these tools at scale, and due to the cost now find themselves making pragmatic decisions about where and how they are being used. What is worrying is that when these anthropomorphised tools decide to go off-brief, they take no responsibility. Worse still they completely down tools when out of credits. This stifles creativity and creates very real blockers to delivery – something the humans we work with do not.

At Bernadette we’re anticipating how roles and processes are changing. Horizons have shrunk. Brands and products can no longer plan in years, they need to be responding in days and weeks. – not just with prototypes, but with final delivery. So we’re challenging brands to be truly responsive, enabled by these new AI tools, and working together with us.

Because whether your teammate is human or generative, the mandate is identical: move faster, adapt instantly, and never lose sight of why you’re building it in the first place.