The aisle just got a lot shorter. Meet Sagent.

There’s a running store on the edge of Vondelpark here in Amsterdam where the fitting is part of the experience. If you want, you can run on the treadmill while they film your stride. They look at how you land, whether you pronate, how much cushioning that calls for, then ask what distance you’re training for and what hurt last time. The whole place is set up to get your fit right.

You can walk in for a pair you already picked out online. You walk out with a different pair, and with the reason it suited how you actually run. That moment is the whole reason you would go back.

Online, that conversation goes missing. A shopper gets a search bar, 40 filters, and 2,000 results sorted by relevance to nobody in particular.

Sagent is our answer, and it’s the biggest bet Crobox has ever made. A year and change in the making, live today. Your shopper asks in their own words, and your catalog answers.

The bet we’re making

Everyone in ecommerce knows the problem by now. Endless choice stopped feeling like a feature a long time ago. Faceted search, filter panels, quizzes, a chatbot parked on the help page: we’ve all tried to patch it, and every patch still asks your shopper to translate what they want into language your catalog already speaks.

So here’s our bet. The next decade of product discovery happens in the shopper’s own words.

That’s a bigger claim than it sounds. Product data gets written for merchandisers, logistics, and compliance. It says 8mm drop, 265 grams. Your shopper is thinking “I’m training for my first marathon and my knees aren’t there yet.” An assistant that can’t bridge those two sentences is a search bar with better manners.

Bridging them takes more than a good model. It takes an assistant that understands your products the way your best associate does.

And the brands we talk to keep pointing at the same culprit.

What we keep hearing

A dozen conversations with major retailers and brands over the past few months. Two of them:

Dozens of markets, one product. One sells across dozens of markets. Product data lives in a different system per region, maintained by local teams who each describe the same product their own way. Any assistant dropped on top of that gives one answer in one market, a different answer next door, and is guessing in both.

Hundreds of brands, no shared language. Another carries hundreds of supplier brands. Every one of them sends its own feed, in its own format, with its own idea of what a spec sheet should say. Their shoppers compare products that were never described in the same terms.

In both cases the hard part sits in the data. Siloed product, brand, and category information is the single biggest reason conversational AI turns messy, because a fluent assistant working from fragmented data is confidently wrong in four languages at once.

So Data Enrichment sits underneath Sagent rather than beside it, unifying that information across territories and brands before a single conversation starts. Get that right and the promise holds: every shopper guided, every recommendation grounded in your catalog, every choice made with confidence.

What a year buys you

Sagent starts where most AI assistants stop. With your data.

Enriched product data underneath

Data Enrichment turns technical attributes into the language shoppers think in: usage contexts, style, sustainability, the drivers behind a category. It flags what’s missing and writes the enriched attributes back into your PIM. No model can personalize what it doesn’t understand, and this is the layer that makes everything above it work.

Behavioral psychology in the logic

The Fogg Behavior Model tells us when to ask and when to stay out of the way, because motivation, ability, and a prompt have to land together. The Spectrum of Needs Model is ours, built in-house over years of category research, and it works out why a shopper wants a product at all, connecting what they need to how your products meet it.

Answers grounded in your catalog

Sagent recommends from your enriched data, in your brand’s voice, localized per market. Your shopper gets advice they can act on.

That groundwork already runs Product Finder, our structured guided journey, where a shopper answers a few questions and gets one recommendation. A global sportswear brand saw 16x ROI and a 52% lift in conversion on its shoe finder. An outdoor brand hit 5.2x ROI on its pack finder. Sagent is built on the same foundation.

None of this was a weekend project. It’s why we took the long way around instead of bolting a chatbot onto a help page and calling it guided selling.

What this changes for your shopper

Your funnel assumes one direction of travel. Real shoppers arrive mid-journey, from an ad, a comparison article, a friend’s link, each one carrying a situation to solve.

A conversational layer changes three things.

1. The entry point stops mattering. Category page, product page, a screen in your store. Same question, same assistant, thread picked up wherever they landed.
2. The journey gets shorter and surer. Your shopper describes the situation and gets a short, explained set of options. Confidence at the point of choice lifts conversion and AOV, and it keeps product out of the returns pile.
3. You hear the question. Every conversation is a shopper telling you, unprompted, what they’re trying to solve and the words they use for it. That’s consumer insight most brands pay a research panel for, arriving as a by-product of selling.

Guided selling closes the gap between a need and the right product. Conversation is the shortest route there.

The why behind the buy

Here’s the part we’re most excited about. Ecommerce has spent fifteen years getting better at showing products. Sagent gets better at understanding people, and it sharpens every time a shopper talks to it. The brands that move first will know things about their customers that the rest of the category is still guessing at.

We’ve opened Sagent to a small group of pilot brands now. Three months, live in two to four weeks. If your catalog runs deep, your shoppers keep bouncing off the filter panel, and your returns tell a story you’d rather not read, let’s talk.