Guide · 6 min read
How can a jewellery manufacturer tag thousands of designs without manual tagging?
Let AI label each design from its catalogue photo, and keep the design code, weight and purity in your SKU master, where they already are. QuickLens's photo search matches the design itself, so nobody has to tag the catalogue first. With plain-words search, our models label the designs in your own catalogue, so nobody types tags against each SKU. Your team can then search by photo or in plain words, such as "letter P pendant with stones under 5 grams".
Pranshu Jain · founder · IIT Bombay
Last updated 31 October 2026
Part of the jewellery buyer's guide
Why tagging by hand breaks down
A manufacturer's catalogue often runs to thousands of designs, spread across ERP tables, Google Drive folders and salespeople's phones. To make those designs searchable by words, someone has to open each one and type tags against it. Every new collection adds to the pile.
The tags also depend on who typed them. One person writes 'kada', another 'broad bangle', a third leaves the field blank. A text search then finds a design when the buyer's words happen to match the typist's.
In practice, the catalogue gets searched by memory. Salespeople offer the designs they remember, and much of the catalogue is never shown.
What automatic labelling does
Automatic labelling reads each catalogue photo and describes the design in words a search can use. With plain-words search, our models label the designs in your catalogue, so nobody types tags against each SKU. New designs are labelled the same way. Each design is also indexed by its shape, motif and finish, which is what a photo search matches against. AI search for jewellery manufacturers: how it actually works explains that matching.
Together, the labels and the photo index let your team search in plain words. Take the search "letter P pendant with stones under 5 grams". Nobody tagged any design as 'letter P'. The words that describe the look ('letter P', 'pendant', 'with stones') are matched against the designs themselves, and 'under 5 grams' is filtered on the weight in your SKU master. The search returns the matching pendants from your catalogue.
The same search understands trade language, such as 'kada', and filters on category, weight band, carat and stone type at the same time.
What you still need
- Your catalogue images. The labels come from the photos, so each design needs one. A re-shoot is usually not needed. Search is built to run on catalogue photos as they are, and where photo quality genuinely limits matching, we tell you which parts of the catalogue are affected.
- A SKU master with weight and purity. A photo shows what a design looks like. Its weight, purity, size and stock come from your SKU master, which QuickLens reads from your ERP. How to add image search without replacing your ERP covers that connection.
- One person who owns the sales workflow. Everything else can be layered on later.
We start with a sample of your catalogue and expand once your team trusts the results.
Where automatic labelling falls short
- Labels describe what the photo shows. A detail the photo hides, such as the back of a pendant or the clasp of a chain, can't be labelled from that photo.
- Weight, purity and size can't be read from a photo. They come from your SKU master, so a gap in the master is a gap in the filters.
- A design that is badly photographed in your catalogue is harder to label and harder to find. The WhatsApp photo guide lists what makes a photo easy or hard to match.
- Your own collection names can't be read from a photo, so they stay in your SKU master.
Questions to ask before you choose a tagging tool
Different tools label product photos for different jobs. For a manufacturer's sales desk, these five questions sort them:
- Does it work on your own catalogue, with results drawn from your designs and nothing from the public web?
- Does it connect the labels to your SKU master, so a search can filter on weight and purity?
- Can your sales team search by photo and in plain words where enquiries arrive, such as WhatsApp?
- Does it run alongside your ERP, and what does it write back to it?
- Will the vendor show it working on a sample of your own designs before you commit?
QuickLens is built for all five. Results come from your SKU master and product images, and nothing from the public web. Search runs on WhatsApp, in the web app and through a REST API. Nothing is written back to your ERP until a person approves, and the audit below includes a live search on a sample of your designs.
AI tagging for jewellery designs: common questions
Is there an AI tool that tags jewellery designs automatically?
Yes. With plain-words search, QuickLens labels the designs in a jewellery manufacturer's own catalogue from their photos, so nobody types tags against each SKU. Photo search needs no tags at all, because it matches the design itself. Either way, each result's weight and purity come from your SKU master. It runs on WhatsApp, in the QuickLens web app and through a REST API, alongside the ERP you already use.
Do we need to re-shoot our catalogue before it can be labelled?
Usually not. Search is built to run on catalogue photos as they are, and where photo quality genuinely limits matching, we tell you which parts of the catalogue are affected. Each design does need a photo, because the labels come from it.
Do automatic labels replace our SKU master?
No. Labels describe how a design looks. Weight, purity, size and stock stay in your SKU master and your ERP, and QuickLens reads them from there. Nothing is written back to the ERP until a person approves.
What happens when we add new designs?
New designs reach the search through the QuickLens REST API, which keeps the search index in sync with your ERP or MIS. With plain-words search, our models label new designs the same way as the rest of the catalogue, so nobody tags them by hand.
How can we check automatic labelling works on our catalogue?
We start with a sample of your catalogue. Your team checks the search results on it, and we expand once they trust them. To judge the result, measure what your sales head already tracks, such as quote-to-order conversion on photo enquiries.
Get started
Search a sample of your own designs
Book a 15-minute AI audit and we will map your catalogue, show a live search on a sample of your designs, and give you the honest version of what will and will not work.
Book a 15-minute AI audit