How AI sparked creativity at Hassell

In a bold step toward integrating emerging technologies into creative workflows, Hassell launched an internal competition to explore the role of AI in design—using the imaginative brief of a treehouse at Kew Royal Botanic Gardens, London.

Architecture & Design sits down with Xavier De Kestelier, Principal and Head of Design, to uncover the thinking behind the initiative, the unexpected results, and what it all means for the future of landscape architecture and beyond.
 

Image: The mycelium treehouse / Yue Rao

Architecture & Design: What inspired you to create this internal competition, and why did you choose the treehouse concept at Kew Gardens as the design brief?

Xavier De Kestelier: We ran this competition just over a year ago. At the time, a variety of AI tools were becoming available, particularly those for generating imagery. I wanted our architects and designers to engage with these tools and explore what they could do with them. 

Rather than simply offering a training session—which felt a bit dull—I thought a competition would be a more engaging way for people to self-train and experiment.

We chose the treehouse brief because it was based on a previous Royal Institute of British Architects (RIBA) competition, so there was already a well-defined framework. More importantly, the concept appealed across disciplines—architecture, landscape architecture, and interior design—all of which are core to Hassell.

How did you see AI playing a role in helping teams explore creative directions in landscape architecture through this initiative?

The competition wasn’t exclusive to landscape architects, but it did attract a strong response from them. We had around 60 participants overall. One reason for the enthusiasm among landscape architects is that it’s traditionally been harder to create high-quality visualisations in this field.

AI helps generate photorealistic imagery more easily and quickly, which can be a real asset.

Image: Daisugi Treehouse / Razvan Ghilic Micu

Were there any surprising or unconventional approaches that emerged from the entries, particularly in how AI was used?

Yes. We asked entrants to include the prompts they used along with a short (100-word) description and their final image. Interestingly, in the top submissions, almost none of them actually used the word "treehouse" in their prompts. Including it often led to cliched, fantasy-like designs because the AI pulled from stereotypical online imagery. 

The most successful entries developed original ideas first and then crafted prompts around them—highlighting how crucial creative prompt engineering is.

What criteria did you and the judging team use to evaluate the submissions?

We had five criteria. Key among them were sustainability and regenerative design, visual quality of the image, originality, and how well the concept was executed and explained. The alignment between the concept description and the visual output was particularly important—we judged this competition just as we would a real-world one.

How do you see AI shifting the traditional design process in landscape architecture over the next few years?

AI won’t replace creativity—it enhances it. Interestingly, this tool wasn’t just picked up by younger team members. We had senior associates and principals getting involved, which is unusual for new tech. 

AI is accessible because it’s based on language rather than complex software skills. It’s reinvigorating the value of hand sketches and written concept descriptions in the design process.

Image: Blooming Ruin / August Sjolin

Were there any recurring themes or ideas across the entries that you found particularly exciting or telling about the future of design?

Sustainability and regenerative design came through consistently. That’s central to our work at Hassell, and it was evident in nearly every submission.

How did this competition influence or inform broader conversations within Hassell around the integration of technology and design thinking?

We intentionally kept it informal—no structured training. The real aim wasn’t to design treehouses, but to get people experimenting with AI. One of the winners only started using the tool the weekend before the deadline! The competition created a playful, low-pressure space for learning, and that’s been very valuable.

What support or tools did you provide to participants to help them explore AI as part of their design workflow?

We didn’t prescribe specific tools. Instead, we offered a budget for participants to purchase subscriptions to tools like Midjourney, DALL·E, and Stable Diffusion. These platforms weren’t yet tailored for corporate environments, so this flexible approach let us see which tools people naturally gravitated towards.

Image: The Birch Halo / Joe Loughnan

What’s your favourite aspect of the winning entry—and were there any ‘near misses’ you personally loved that didn’t make the final cut?

There were some fantastical entries—floating pavilions and the like—that, while not technically feasible, were visually stunning. The winning entry was simple and thoughtful: it proposed reusing discarded festival tents, like those left behind at Glastonbury, as hammocks suspended in the treetops.

A small music event would form the centrepiece. It used minimal material, had a clear concept, and the image captured the idea beautifully. Again, they didn’t use the word “treehouse” in the prompt.

Do you foresee similar competitions or experimental platforms becoming a regular part of Hassell’s internal culture?

I hope so. I’m already thinking about a follow-up. With developments in AI for video, our innovation team is exploring new ways to integrate emerging technologies into the design process. Future competitions might look very different—but they’re definitely on the horizon.

Hero image: Airpark / Bryn Jones and Rachel Clements