top of page

From the Workbench to the Algorithm: How ELLIE is Bringing Secure AI into Craft Workshops

7 days ago
3 min read

Imagine stepping into a historic Florentine leather workshop. Around you are sketches, physical archives spanning over 70 years, and the unmistakable scent of premium leather. Now, imagine artificial intelligence sitting at that very same workbench.


This isn't a distant sci-fi scenario; it is exactly what was discussed on 12 September 2026 during the Artigianato e Palazzo exhibition in Florence. Here, Artex, Fondazione per la Ricerca e l’Innovazione (FRI), the University of Florence, and Sapaf Atelier 1954 gathered to present the Tuscan pilot of the Horizon Europe ELLIE project.



Speaking the Same Language: The Glossary Challenge

How do you make AI useful in a craft workshop? For a craft business, adopting AI is rarely about buying off-the-shelf software. 


Elisa Guidi (Artex) highlighted that craft businesses have grown more open to digital tools since the pandemic—moving from basic marketing to archive digitisation. Her message: technology should serve the craftsperson as a practical tool, much like a hammer or a chisel.


But how do you teach an algorithm how a handmade handbag is constructed? Niccolò Calistri from Sapaf Atelier 1954 admitted that digitising even part of the company’s physical archive required a substantial investment of time and resources. Photos alone cannot convey everything: the feel of a material, the details of its construction or the knowledge behind a traditional technique. 


To bridge this gap, the team had to build a bespoke semantic glossary reflecting the workshop’s own language. Because a leather-working technique might be named differently in Florence compared to Bologna, teaching the AI the specific terminology of the workshop was a key breakthrough in making the tool effective. 


Beyond Text and Images: The Internet of Senses

The ELLIE project looks much further than standard text and imagery. One of its most revolutionary frontiers is the Internet of Senses. As Alessandro Monti (FRI) explained, the ultimate goal is to virtualise and digitise human perceptions. Imagine a digital twin of a luxury handbag that doesn't just show its shape, but allows a client in New York or Tokyo to virtually experience the texture of the fabric, the richness of the dye, or the smell of the leather!


The practical promise is just as compelling: digital representations can streamline client feedback. Instead of physically shipping multiple samples back and forth across the globe—incurring costs and delays—artisans can collaborate with international buyers in a matter of clicks, bringing Florentine cultural heritage to a worldwide audience.



Keeping Craft Knowledge in Safe Hands

Security and compliance were central themes of the panel. Where are confidential sketches processed? Who can access them? How is the company’s knowledge protected? The speakers stressed that adopting AI also means understanding how data is handled and the responsibilities involved.


Professor Paolo Nesi (University of Florence) explained that the ELLIE prototype uses customised agentic AI models hosted entirely on internal servers. Not a single bit of the information drawn from Sapaf’s 70-year creative heritage leaves this secure infrastructure. It is a private, custom-tailored system that progresses through school alongside the artisan, gradually learning the workshop’s language and the needs of the business. 


The Human Touch in a Digital Era

Ultimately, ELLIE shows how meaningful innovation relies on a careful blend of human expertise, structured data, and trusted collaboration. By keeping creative control firmly in human hands, advanced technology can help craftspeople draw on their heritage to explore new designs. 


The next handbag may begin with a conversation between generations of expertise and an algorithm—but, as stated by Prof. Nesi, while the prototype can draw inspiration from archived designs or turn a sketch into a visual mock-up, the artisan sets the direction, explores alternatives and decides which ideas deserve to leave the screen.



 
 
 

Comments


bottom of page