Skip to content
Aurene.

Retail & E-commerce

An AI assistant that lifted conversion

Shipped an LLM-powered shopping assistant with RAG over the product catalog and live inventory.

+23%
Conversion lift
10 wk
End to end
A/B
Measured rollout

The challenge

Shoppers struggled to find the right product in a catalog of tens of thousands of SKUs. Search was keyword-based, and support tickets for 'which one should I buy' kept climbing.

  • Keyword search failed across tens of thousands of SKUs
  • Shoppers couldn't find the right product; carts stalled
  • 'Which one should I buy' support tickets kept climbing

Our approach

We built a retrieval-augmented assistant grounded in the live product catalog and inventory, with evaluation and guardrails baked in from the start. It shipped behind an experiment so impact was measured, not assumed.

  • Built a RAG assistant grounded in the live catalog and inventory
  • Baked in evaluation and safety guardrails from day one
  • Shipped behind an A/B experiment to measure real impact

The outcome

The assistant lifted conversion by 23% in the test cohort and cut pre-sales support volume, while staying accurate because retrieval stayed fresh with inventory.

  • +23% conversion in the test cohort
  • Lower pre-sales support volume
  • Answers stay accurate as inventory changes

How it works

The path a request takes through the system.

  1. 1User query
  2. 2RAG retrieval
  3. 3LLM
  4. 4Guardrails
  5. 5Response

Results

  • +23% conversion in the test cohort
  • Lower pre-sales support volume
  • Answers stay accurate as inventory changes
  • Shipped behind a measured A/B experiment

What we delivered

RAG pipeline grounded in the live catalog
Evaluation harness and safety guardrails
A/B experiment framework for measured rollout
Admin tooling for content teams

Have a project in mind?

Tell us where you’re headed. We’ll tell you the fastest, soundest way to get there.