Your AI understands why people buy
Product essences give every item in your catalog three layers of meaning, so the agent matches customers to outcomes, not keywords.
Three layers of product understanding
Every product gets an LLM-generated essence that captures what it does, how it makes someone feel, and what using it is actually like.
Functional
What it enables, problems it solves, outcomes it produces.
“Enables consistent daily training on paved surfaces. Solves heel impact pain on long runs. Produces stable mileage accumulation with reduced injury risk. Fits wide feet without break-in period.”
Emotional
Identity transformation, tribal belonging, values alignment.
“Signals committed runner identity, not weekend jogger. Belongs to the serious-but-approachable running community. Aligns with discipline and self-improvement values without elitism.”
Experiential
Sensory feel, emotional experience of use.
“Cushioned landing that absorbs without feeling mushy. Light enough to forget mid-stride. Breathable mesh keeps feet cool through kilometer 15. Satisfying lace lock that stays put.”
Keywords miss intent. Essences capture it.
Side-by-side: the same query through keyword search vs. essence-powered search.
“I need something for my sore back after sitting all day”
- Back scratcher
- Backpack
- Chair back cushion
Matches 'back' literally
- Foam roller: relieves tension from prolonged sitting
- Ergonomic lumbar pillow: corrects desk posture
- Heating pad: soothes muscle stiffness
- Yoga mat: enables stretching routines for desk workers
Matches the outcome: pain relief from sedentary work
“Something to make my apartment feel more cozy”
- Cozy brand pajamas
- "Cozy Nights" candle
Matches 'cozy' in product names only
- Wool throw blanket: warmth and tactile comfort
- Soy candle trio: ambient light and calming scent
- String lights: soft warm glow for any room
- Hot chocolate mix: the ritual of a warm evening drink
Matches the feeling: warmth, comfort, atmosphere
From catalog to intelligence
Essences are generated automatically when you import your catalog. No manual tagging, no taxonomy design.
Catalog Ingestion
Your product feed (names, descriptions, attributes, images) gets imported via CSV, API, or direct database sync.
Essence Generation
An LLM analyzes each product and generates three-layer essences: functional outcomes, emotional resonance, and sensory experience.
Vector Embedding
Essences get embedded as 4096-dim vectors alongside the original product data. Both semantic meaning and exact attributes become searchable.
Live Matching
When a customer describes what they want, the agent matches against essences in real-time, surfacing products by outcome, not just attributes.
Recommendations that feel human
Traditional recommendation engines match purchase history and collaborative filtering. Essences match what the customer actually wants to achieve: the outcome, the feeling, the experience.
Customer: “I want to feel energized in the morning”
Search: “energized” → 0 results (no product named “energized”)
Customer: “I want to feel energized in the morning”
Agent matches functional essence “provides sustained energy” →
Matcha powder, cold brew concentrate, acai bowl mix, protein granola
See how your products look with essences
Upload a sample of your catalog and we'll generate essences for 50 products, free. See the difference in search quality yourself.