Technology

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.

Running Shoes: Nike Pegasus 41

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.

Running Shoes: Nike Pegasus 41

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.

Running Shoes: Nike Pegasus 41

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.

Customer Query

I need something for my sore back after sitting all day

Keyword Search
  • Back scratcher
  • Backpack
  • Chair back cushion

Matches 'back' literally

With Essences
  • 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

Customer Query

Something to make my apartment feel more cozy

Keyword Search
  • Cozy brand pajamas
  • "Cozy Nights" candle

Matches 'cozy' in product names only

With Essences
  • 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.

01

Catalog Ingestion

Your product feed (names, descriptions, attributes, images) gets imported via CSV, API, or direct database sync.

02

Essence Generation

An LLM analyzes each product and generates three-layer essences: functional outcomes, emotional resonance, and sensory experience.

03

Vector Embedding

Essences get embedded as 4096-dim vectors alongside the original product data. Both semantic meaning and exact attributes become searchable.

04

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.

Agent reasons about intent, not just attributes
Products ranked by outcome relevance, not popularity alone
Cross-category discovery: BBQ query returns meat, sides, and drinks
Without Essences

Customer: “I want to feel energized in the morning”
Search: “energized” → 0 results (no product named “energized”)

With Essences

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.

50 free product essencesResults in 48 hoursNo commitment
Product Intelligence | IsarTech