AI / RAG System
Power AI answers with your real product knowledge.
Plain chatbots make up specs and prices. Scouty AI looks at your real catalog and documents first, then asks the model to answer using only what was retrieved. Every claim is tied to a source the user can click.
Ungrounded AI answers don't belong on a storefront.
A chatbot that answers from training data alone is a liability. Shoppers, support agents, and sales reps need answers tied to your real product knowledge.
- Plain LLMs hallucinate specs, prices, and compatibility
- Without citations, no answer is trustworthy enough to act on
- Stale or partial product data produces confident wrong answers
- Costs balloon when generation is unbounded
Retrieval first, generation second, citations always.
The system prompt tells the model to answer using only the retrieved sources, and to say 'I don't know' when the sources don't cover the question. That single rule kills most of the things shoppers and support agents complain about.
- Hybrid keyword + semantic + visual retrieval
- Inline citations linked to products, document pages, and assets
- Permissions enforced at retrieval time (public, partner, internal)
- Generation costs capped or passed through
- Multiple assistants for shoppers, support, and sales
- Evaluation hooks for your own quality bar
Every claim ties back to a source.
The assistant retrieves relevant products and document pages first. Answers cite each source so shoppers, agents, and reps can verify.
Question
Which boots are best for wet trails over multiple days?
Answer
The Trailmaster GTX is the closest match for multi-day wet routes. It uses a GORE-TEX membrane for waterproofing and a Vibram Megagrip outsole for slick rock and mud. The care guide recommends 30 minutes of break-in walking before a multi-day trip.
Trailmaster GTX Boot
Product · SKU TM-GTX-42
care_guide.pdf · page 4
Document
spec_sheet.pdf · page 2
Document
collection · trail-shoes
Asset taxonomy
What ships with Scouty AI.
Grounded retrieval
Hybrid retrieval over products, documents, images, and assets.
Inline citations
Sources cited inline and as a list, with linked targets.
Permissioned
Apply audience-level and document-level permissions at retrieval time.
Tunable assistants
Multiple assistants. Product Q&A, guided buying, support, sales. With separate prompts and eval sets.
Citation metrics
Click-through to cited sources is a first-class metric, not an afterthought.
LLM flexibility
Bring your own model or use Scouty's. Cap or pass through generation cost.
Where it matters most.
Product Q&A
Shoppers ask 'is this washable?' and the assistant cites the care guide.
Guided buying
'Recommend a hiking boot for wet trails' returns 1–3 SKUs with reasons and citations.
Support copilot
Agents search products and manuals from one bar; AI summarizes with sources.
Sales enablement
Reps generate sourced summaries that pass legal review.
What Scouty AI retrieves from.
Get started
Ready to put one search bar across your products, documents, and assets?
Start with a free expert-led Search Audit. A Scouty specialist will review your storefront and recommend a clear next step.