Challenge
A US based clothing manufacturer were experiencing significant costs processing artwork to be printed on clothing. Before clothing could be sold digitally, it had to be tagged (described) with metadata for each vendor it was sent to.
The organisation was operating a team of 6 employees to manually describe artwork files with extensive sets of metadata. The descriptions were done using Microsoft Excel, a slow and error prone approach. No technology integrations were in place to automate the process.
Costs to operate the team were in the hundreds of thousands per year and more importantly the tagging process was a bottleneck for their entire production and revenue generation.
Solution
An Azure automation platform was produced using multi-modal generative AI to:
automatically collect artwork from source systems
collect and structure metadata, including applying mappings to enrich the base metadata
classify images with remaining tags
export the data into the target system ready for human review and sending to vendors
Project Details
Project Duration - 2 week proof of concept, 2 month implementation
Project Team - 1 delivery lead, 1 data engineer, 1 data scientist
Benefits
Artwork throughput was increased from 150 images per day to the total produced of 3000, unlocking huge amounts of potential revenue.
FTE required in the tagging team could be reduced from 6 down to 1 saving huge costs.
The automation platform was one of the first major efforts in Azure, enabling the opportunity to investigate integration of CRM data to further increase revenue generation via AI.