Click through a real guided sale.
Not a mockup — two working stores you can use right now. Pick the catalog closest to yours, answer a few questions the way your buyer would, and watch the recommendation, the quantity and the cart resolve.
ACME Coatings — product finder
Sealers and coatings across six surface types. Answer substrate, area, traffic and condition; the finder returns a complete system — primer, sealer, applicator — with the coverage math shown and the kit count worked out.
Try the finder → A different catalogACME Tours — guided booking
The same decision layer on a catalog that has nothing to do with chemistry. Useful if you want to see the mechanism separated from the coatings vocabulary it is usually demonstrated in.
Try the booking flow →Both run in your browser. No signup, no data collected, no sales call attached.
Five layers, each doing only what it is qualified to do.
The demo looks like a conversation. Underneath, the work is split so that no single component is asked to be right about everything. The language model handles ambiguity; approved knowledge, business rules, calculations and your systems of record stay authoritative where correctness matters.
From a customer’s problem to a placed order
Understanding is probabilistic. Knowledge is structured. Inference is deterministic. Calculations are mathematical. Commerce is transactional.
Start with the smallest decision worth trusting.
The failure mode is starting broad: a general assistant across the whole catalog, with undefined authority and nothing to check it against. The sequence that works is the opposite — one consequential product line, made trustworthy end to end, then widened.
Build the smallest trustworthy decision
A focused, auditable decision is a stronger starting point than a broad assistant with undefined authority.
1 · Choose. One product line where getting it wrong actually costs something — a ruined floor, a failed bond, a voided warranty. Consequence is what makes the decision worth governing.
2 · Normalize. Your product master becomes a shared vocabulary and a case schema: what each product is for, what it is not for, what it requires alongside it, and how quantity is derived.
3 · Govern. The rules that carry commercial or safety weight are written down and made checkable — compatibility, exclusions, thresholds — each one attached to the evidence behind it.
4 · Connect. Coverage math, pricing and inventory come from the systems that already own them. The decision layer orchestrates; it does not invent commercial truth.
5 · Operate. Run it, watch where buyers hesitate or drop, correct the rules, then widen to the next product line. Expansion is earned, not scheduled.
How Selrite guided selling works
Selrite replaces the century-old search-and-grid storefront with a guided buying layer. Instead of returning a wall of SKUs and leaving the shopper to guess, it asks a few plain-language questions about the job, computes the quantity actually needed, and recommends the right product variant plus the companion items the job requires.
What your customer feels is a short sequence — substrate, area, traffic, condition — that resolves to an exact recommendation with the consumption math shown, then drops the full kit into the cart. Showing the working is what makes the buyer trust the number.
What makes it possible is the ingestion behind it: Selrite reads your catalog, enriches each product with FAQs, documents, video transcripts and approved peer specs, then normalizes everything into decision-grade JSON — suitable-for, not-suitable-for, required companions, and quantity logic — with provenance and a confidence score on every fact. Low-confidence or safety-sensitive claims route to human review before they ever reach a buyer.
Selrite Finder runs on WooCommerce, Shopify and BigCommerce, and a companion mobile app gives the store owner a real-time view of buyers moving through the finder.
This page shows the mechanism. The case for it — why better search never closes the gap between a buyer’s job and a SKU — runs as a short visual explainer in eight scenes.
Common questions
Do I need to sign up to try the demo?
No. The demos at demo.selrite.ai run in your browser with no signup, no data collection and no sales call. Pick the catalog closest to yours and click through a complete guided sale.
How does guided selling reduce returns?
Most wrong-fit orders come from a store that makes the shopper do the expert's job. By asking the right questions and recommending the correct product, quantity and companions up front, Selrite catches mismatches before checkout — where they're free to fix.
How long does it take to get one running?
That depends on the state of your product data, which is usually the honest bottleneck rather than the AI. The sequence above is deliberately narrow at the start: one consequential product line, made trustworthy end to end, before anything widens.
Which platforms does Selrite support?
WooCommerce, Shopify and BigCommerce.
That's the whole idea.
Decision-grade product intelligence, surfaced as a guided sale — and as a phone-in-pocket view of how buyers respond.