Product data enrichment
New products arrive half empty. Your team fills them by hand.
Cernel's agents fill every field and write each product in your brand voice, by rules you set once.
Sound familiar?
If every new product starts from a blank page, this is what changes.
How it works
One agent per field, set up once, run on every product.
An agent is one saved rule for one field: what to write or pick, in which tone, from which data. Describe it in everyday words, test it on your own products, and from then on it fills that field on every product in the category. Pick an agent to see it.
Upper materialMulti selectShoes › Sneakers
What should your agent do?
Pick every material the upper is made of, from our material list. Read the supplier data and the product photos.
UsesSupplier dataProduct imagesMaterial list
Run test5 of your products
- Court sneaker, Cream
- Runner, Sand
- Canvas low, Ecru
- Court sneaker, White
- Trail runner, Moss
Apply feedback
What agents can do
- Write
- Titles, meta titles, descriptions, search keywords and care instructions.
- Write with layout
- Descriptions with headings, paragraphs, bold and lists.
- Pick one
- One value from your own list, such as skin type or fit.
- Pick every one that fits
- Every value that applies, such as materials or occasions.
- Research
- Facts and product photos found on the web, on the sites you allow.
- Make images
- New images, such as styled or lifestyle shots, with alt text and file names.
- Choose the best
- The best of several candidates, such as the cleanest packshot.
- Working together
- One agent per field and category. An agent can build on what another wrote, and Cernel runs them in the right order.
- 1
Describe the rule
Say what the field needs in everyday words. Point it at the product data, the images, other fields and your brand voice.
- 2
Test on your products
Run it on a few of your own products. See each value with the reasoning behind it, and keep refining. Every change is a version you can go back to.
- 3
Run it
On any selection of products, only on empty fields if you like, and in every language you sell in.
- 4
Keep it running
An automation runs the agents on every new product as it arrives, or when the attributes it depends on change.
Reviewed before it goes live
One invented claim is worse than an empty field.
You choose which values go live automatically and which get a review step; review steps check every value and only ask a person when something fails or needs a decision. Here, a step written in plain words holds a highlight that promises a waterproof upper the product data does not support.
Leather Court Sneaker
- Highlight 1Cream leather with suede overlays
- Highlight 2Waterproof leather upper
- Highlight 3Stitched rubber cupsole
Not live yet
Review step
Fail if a highlight claims something the product data does not include.
Product data
- Upper: leather, suede
- Lining: textile
- Treatment: none
Held for a person
Highlight 2 says waterproof. The product data has no treatment.
RejectEditApprove
What you get
Complete products you can trace, in every language, back where you sell.
Every value traceable
Each value shows whether an agent wrote it, it came in with the product or a person typed it, when it changed and its earlier versions, which you can restore. Every AI-written value shows what it was based on and the reasoning. You see how complete each field is, per category and market.
Agentic reviewEvery language you sell in
Per language, each value is written fresh or translated from a language you pick, by that language's own rules, and translated values go through the same checks.
Translation and localizationBack where you sell
Approved values go back to your shop or PIM, typically within a minute, for the systems we integrate with. Anything else gets a file in its own columns, or reads the changes through the API.
AI PIM
Tiffany Lighting Direct
Tiffany Lighting Direct gets a new supplier range live in days, not months
- to get a new supplier range live
- Days
- products from 20+ brands
- 13,000+
- extra hires needed
- 0
Firtal Group
Firtal Group added a thousand extra hands without adding headcount
- products enriched
- +25K
- organic traffic to product pages
- +20%
- faster time to market
- +80%
Molly & My
Molly & My builds every market on one foundation
- products enriched across 8 markets
- 20,000
- faster time to market
- 4x
- organic traffic to product pages
- +29.3%
Questions
Product data enrichment, answered.
What is product data enrichment?
Product data enrichment turns a thin supplier record into a complete product: the attributes shoppers filter on, the copy they read, the images and the fields each channel needs. In Cernel, AI agents do that work attribute by attribute, from your own data and rules, and you decide what goes live.
Where does Cernel get the values from?
From the product's own data and images, the fields other agents already filled, and the brand voice and product rules you add as text or images. Research agents also look on the web, on the sites you allow, matched by an identifier such as GTIN, EAN or SKU. Agents that pick a value only choose from the options you define.
Which AI models write the values?
For writing and picking values you choose Google, OpenAI or Anthropic, and Cernel picks the model. Images are made with a Google or an OpenAI image model.
Will enrichment overwrite values my team already wrote?
Only if you want it to. You can run agents on empty fields only, and with review on, every new value waits for approval with the reasoning behind it. Every earlier version of a value can be restored.
Can Cernel write alt text for product images?
Yes, for the images Cernel makes. You describe how alt text and file names should read once, and each new image gets its own, sent to your shop with the image.
How does Cernel work with our PIM?
Cernel is the AI layer that fills your product data: agents fill the attributes and write the content, your checks and your team decide what goes live, and the finished values go back to your PIM or shop, or out as a file or through the API.
Where to next.
- External data and images
Some facts are not in your data at all.
An agent finds them on the sites you allow, matched by EAN or name.
- Translation and localization
Each new market means rewriting everything.
Every market gets its own wording, not a word-for-word copy.
- Market insights
Which words do shoppers actually search for?
The terms people use, per category and market, so your copy uses them.