Cluse: Re-imagining design

Cluse, a large international watch brand selling their products in over forty countries worldwide, wanted to investigate whether they could use artificial intelligence algorithms to make their design process more efficient.

At the start of this project we took a deep dive into the current design processes of Cluse in order to learn their current design workflow. We had many conversations with designers, project managers and photographers to get a clear overview of the main pain points they were having and where they would see room for improvement. Based on these conversations and discussions we spotted many use cases in which AI models could be very helpful but we had to narrow these use cases down to one or two of the most valuable use cases.

These were the two use cases we proposed and which Cluse decided to proceed with:

1. Automatically discover and classify design features and combination of design features that are popular or are gaining popularity, in order to get real time insights of current and developing customer preferences. In this way Cluse could create and adapt their designs much faster to keep up with changing customer preferences and trends.

2. Creating new watch designs automatically in order for designers to scale up the design of new watches almost tenfold. By automatically generating designs of new watches, designers could pick out the interesting designs created by an AI algorithm and fine-tune these designs.

For the first use case we created a system that continuously extracts popularity features of watches from online sources which we combined with data from the CRM and ERP systems of Cluse. Thereafter we trained a machine learning model that could group the most popular watch features together, such as colours, textures, physical shapes and much more features. To visualize and present the results of the machine learning model we build a dashboard through wich Cluse can continously monitor ongoing and current customer preferences.

For the second use case we made use of a very exciting new brach of artificial intelligence models called generative artifical intelligence. With generative AI models you can train AI models in such a way that they learn to generate new samples of the input you the train the model on. Generative models consist out of competing discriminator and generator functions in order create a neural network that generates or designs new samples of the input you put into these models. We won't dive into the neaty details of these models over here since you can write a blog post alone about this subject. 

We trained a generative network based on a large dataset of visual images of Cluse watches, resulting in a generative artifical intelligence model that creates new designs of Cluse watches very quickly. The results of this model were integrated into the dashboard of the first use case. The designers of Cluse were at first a bit flabbergasted by the ability of the model to create new Cluse watch designs out of the blue but the more they played with the model the more they loved using it. They are now able to let the model design hundreds of new watch designs while they can focus on fine tuning the designs based on their creativity. 

Cluse was very happy with the results of this project. The two use cases and forthcoming models have a real impact on the design process and result in clear value for Cluse. We are currently working again with Cluse to see how we can integrate AI also in other parts of the company.