One Jewellery Brand, Two AI Trainings in Two Days: Design, Then Operations
In November 2025, a major Hong Kong jewellery retailer booked two AI training sessions on consecutive days. Day one: AI in jewellery design, for the creative team. Day two: AI in operations, for the operations staff. Different departments, different problems, different material built from scratch for each room.
The design session went the way design sessions at jewellery brands tend to go. The use case is immediate -- mood boards, concept iteration, trend research, rendering variations on a theme faster than a junior designer produces them by hand. The session was built around that interest: generate concepts from a brief, compare outputs against an existing collection, evaluate where human judgment is still needed and where the tool handles the production work. The room's energy ran on curiosity. Everyone wanted to see what the tool could generate.
Day two was different. Operations staff at a jewellery retailer don't care about image generation. Their work is documentation -- supplier communications, procurement workflows, quality-control reports, the written repetition that keeps a supply chain running. The tools changed. The exercises changed. The evaluation lens shifted from "does this look right?" to "does this read like something we'd actually send to a supplier?" The atmosphere was more practical, more cautious, less about possibility and more about whether the output was reliable enough to use.
I've run plenty of first sessions that went well and led to nothing -- a strong rating, a polite thank-you, and silence. What made this engagement different was the cross-department commitment. Two sessions for two departments with fundamentally different work is a company deciding that AI training is worth rebuilding from scratch for a second audience.
What cross-department demand signals
In corporate AI training, most engagements are one session for one audience. The company books it, participants attend, the satisfaction score comes back fine, and the engagement is complete. When a company extends training to a second department with entirely different problems, it signals something the satisfaction score can't: the first session changed someone's work visibly enough that interest spread to another part of the building. Whether that happens through a mandate or through someone mentioning over lunch that a tool actually helped -- the effect is the same. Demand that crosses departments is harder to manufacture and more reliable as evidence than any post-session survey.
What this means if you're planning training
If you're evaluating corporate AI training, the most useful thing to optimize for isn't the first session's satisfaction score. It's whether participants leave with something specific enough in their workflow that a different team notices. A well-designed workshop produces one tool or process each participant uses the following week -- something concrete, not a concept. When that happens, the next booking comes from a department you didn't pitch to. When it doesn't, even a company-wide mandate won't replicate the organic pull.
The two days at the jewellery brand required completely different tools, exercises, and evaluation criteria. They worked for the same reason: both were built around the participants' actual daily work, not around a general demonstration of what AI can do. That's probably the part that justified the second day.
