🌐Diese Seite ist noch nicht auf Deutsch verfügbar.
NeuLernen Sie Sagent kennen, unseren KI-Shopping-AssistentenLernen Sie Sagent kennenMehr erfahren

Your product is better. Your website should say why.

Shoppers describe a situation. A dishwasher safe frother. Dog food for senior dogs. Shoes for first marathon. In your stores, someone turns that into the right product in about three questions. This runs the same conversation across your full catalog, in every market you sell in.
Book a Demo

What is an AI shopping assistant?

An AI shopping assistant is a conversational layer on an ecommerce site that helps a shopper work out what to buy, rather than helping them find something they can already name. It reads intent from plain language, asks a handful of clarifying questions, and narrows a catalog of thousands down to the two or three products that actually fit.

That's a different job from a support chatbot, which handles order status and returns, and a different job from site search, which needs the shopper to already know the vocabulary. An AI shopping assistant sits where the decision happens, and it's the part of the journey where choice overload quietly costs the most revenue.

Crobox has been solving that decision moment since 2014, first with guided product finders and now with Sagent. Both run on the same foundation: your product data, enriched into the language shoppers actually use, plus more than a decade of behavioral science about how people choose.

From the problem they have to the product they need
16x
return on investment
41.5x
return on investment
5.2x
return on investment

AI shopping assistant vs. chatbot vs. site search

These three get lumped together in RFPs and they solve different problems. Most ecommerce teams need all three; the mistake is expecting one to cover another's job.

CapabilityAI shopping assistantSupport chatbotSite search
Understands a need described in the shopper’s own wordsYesNoNo
Asks follow-up questions before recommending anythingYesNoNo
Explains why a product fits, not just that it matchedYesNoNo
Works from enriched product data, not raw spec fieldsYesNoNo
Answers order status and returns questionsNoYesNo
Returns results instantly for a known product nameNoNoYes

Three questions, one right answer

Every answer comes from your live catalog, enriched so it knows what your products are actually good for, not just what they are made of.

Understands the real need

It asks enough to understand the situation before recommending anything, then works out which product that points to.

Remembers and confirms

It holds what a shopper already said and checks it back, instead of asking twice.

Explains the why

Every recommendation comes with the reason. Good match, great match, and what it is good for. The engineering finally reaches the shopper.

Why most AI shopping assistants fail

Pointing a language model at a product feed is a weekend project. Getting it to recommend well is not, and the reason is almost always the data underneath it.

A feed says Gore-Tex, 4mm drop, 280g. A shopper says my feet get soaked on the commute. Nothing in the feed connects those two sentences, so the assistant guesses, and a confident wrong answer erodes trust faster than no answer at all.

Crobox Data Enrichment translates technical attributes into the benefits and situations shoppers describe, and keeps doing it as your catalog turns over. That's the part most teams underestimate when they scope an assistant in-house, and it's why Sagent starts from product understanding instead of a guess.

Ten years of knowing why people buy

A better product does not sell itself. Recommendations are framed the way a decision actually gets made, so the reason your product is better lands as a reason to buy it. Your team picks which principles apply, and how strongly.

See every conversation

Know what happened, and what it was worth.

Monitoring

Every conversation is scored and flagged before it becomes a problem, on a dashboard you build with the metrics you already report to the board.

Ask Analytics

Ask which questions came before a purchase, or where sessions dropped, by market or by category. Get a chart back, instantly.

Sagent's Ask Analytics dashboard, showing conversation topics, flagged conversations, and event charts

Set the rules once

You define the guardrails, tone and business rules. No engineering ticket, no release cycle, no roadmap slot.

Guidance Prompts

Reusable rules for brand style and specific dos and don’ts, set by your team without touching the roadmap.
Guidance Prompts
Reusable rules for brand style and specific dos and don’ts, set by your team without touching the roadmap.
Guidance prompts configuration screen in the Sagent admin panel
AI Guardrails
Every response is filtered before it reaches a shopper. On topic, on brand, every time, in every locale.
AI guardrails configuration screen in the Sagent admin panel
Tone & Response Length Presets
Friendly, professional, consultative or direct. Concise or thorough. Set the voice once and it holds.
Tone and response length presets screen in the Sagent admin panel
Business Rules
Exclude the outlet range. Push the categories carrying your margin. Always surface sustainable options. Your merchandising priorities hold on every response.
Business rules configuration screen in the Sagent admin panel
Guidance prompts configuration screen in the Sagent admin panel

Set it up once, use it everywhere

One integration. Every place a shopper meets your catalog, in every market you sell in.

Always-on widget

A consistent way in from anywhere on the site.

Ask about this product

Scoped to one product, for the shopper one answer away from adding to cart.

QR Code activation

A code by the product turns a phone already in a shopper's hand into the same guidance your site gives.

Conversation, with visuals

Text and product cards together. Browsing stays visual, the guidance sits on top.

More capabilities, built in

What else ships, at no cost to your roadmap.

Guided Setup Flow

A guided process and light implementation. Our team at every step, not your engineers.

Live Preview & Test Mode

See and test the exact configured experience before it reaches a shopper.

Dynamic Visitor Personalization

Recommendations adjust in real time to visitor data you already collect.

Adaptive Question Ranking

Questions are ranked by how much they narrow things down. Fewer steps to a product, even on a catalog in the hundreds of thousands.

Brand Theming

No-code controls, or a full CSS override when you need precision.

Translations & Locale Support

Every market, every language, including the small print. A new locale is a setting, not a project.

Custom Tools & Extensibility

Wire in sizing engines, loyalty systems, or the PIM you already run.

Built to pass your security review

EU AI Act disclosure built in
GDPR compliance
Never used to train outside models
See our security program →

Frequently asked questions

We already have on-site search and merchandising. Where does this fit?

Alongside, not instead. Search answers a shopper who can already name what they want. Sagent handles the one who can only describe the situation, a first marathon, a small kitchen, a dog that's getting older, and turns that into a product. Your search rules and merchandising logic keep running exactly as they do now. Sagent works from the same catalog, inside the same business rules you already set.

Can we start with a pilot? What does it cost, and what does it take from us?

Yes. One category, one market, live in two to four weeks depending on how fast your side can turn things around. Depending on scope, typical pilots start from €7,500 over three months. What we need from you: a product feed, a decision on the starting category, and one person who knows the catalog well enough to check the enrichment output. Usually a few hours a week during setup.

How do we measure whether it worked?

We track conversation through to click and add to cart, so you can compare shoppers who used Sagent against those who didn't, in the same category, over the same window. Agree the metric before we start and read it at the end of the pilot. If the number isn't there after three months, you'll know, and so will we.

What does this cost my engineering team, and what does it do to page speed?

A site-wide snippet, pageview events, and access to your product feed. No replatform, no PDP rebuild. The widget loads without blocking your page render. After that, the questions, the tone, the guardrails and the business rules are configuration your team sets in our app, not tickets in someone's sprint.

We have a product feed and access to LLMs. Why can't we build this ourselves?

You can point a model at your feed today and it will answer spec questions well. Shoppers don't ask in specs. They ask for something for wide feet that survives a wet winter. Turning your technical attributes into that language is Data Enrichment, and keeping it current as your catalog turns over is most of the build. The chat window is the easy part.

Once it's live, who owns it?

Ecommerce, with a light review cadence. Enrichment runs continuously and flags missing or incomplete attributes automatically, so gaps surface instead of sitting there. The enriched data exports back into your PIM, which improves your search and filtering at the same time. No new headcount, no dedicated maintainer.
Sagent recommending a product with reasoning

Ready to sell the difference you built?

Book a Demo
Sagent helping a shopper find the right product