AI Assistant for Easier Online Shopping
Query Answer Time
<1s
SapientPro built an AI-powered e-commerce assistant that helps users find the right products faster, reduces search complexity, and replicates the role of an in-store consultant – available anytime, across platforms.
Our Collaboration Story
Our collaboration with this client spans several years, during which we’ve contributed significantly to the development of their e-commerce platform. The client, with a strong technical background, occasionally implements new features independently. Their website is built on ShopWare, which offers both flexible admin-level configuration and opportunities for deeper customization through code. Given the client’s ongoing trust and the relevance of their use case, we proposed developing a proof of concept for an AI shopping assistant directly on their platform.
Challenges
Context-Aware Product Search
The assistant needed to interpret vague or loosely described requests and return relevant results, often from users unfamiliar with product specifications.
Overload of Choice
Customers frequently encountered too many similar options or unclear filters, making it difficult to narrow down selections with confidence.
Missing the Human Touch
There was no system in place to replicate how a knowledgeable salesperson would guide users through comparisons, clarify needs, or explain product differences.
Search Technology Limitations
Initial experiments with vector search engines lacked robust filtering and aggregation, making it hard to support real-time shopping logic.
Compatibility with Existing Stack
The assistant had to integrate smoothly into the ShopWare-based system without disrupting existing frontend and backend workflows.
AI Shop Assistant: Key Features
LLM-Powered Search Assistant
The solution combines a large language model (LLM) with hybrid search capabilities, integrating both classical keyword-based and vector-based search. This setup allows the assistant to interpret and respond to natural language queries with a high degree of contextual accuracy.
Dual-Mode Search Engine
We developed a custom search server that blends traditional search with semantic search, drawing from the online store’s product database. The LLM determines which method to apply, syntactic or semantic, based on the user’s request.
Parameter-Aware Query Handling
The assistant is capable of applying filters such as color, size, category, price range, and sort order. It can also handle pagination and calculate product availability (for example, when asked, “How many green T-shirts are available?” the assistant responds with: “5 options found”.).
Contextual Product Discovery
Users can describe items in abstract or imprecise terms. The assistant identifies which parameters are relevant, and whether clarifying questions are needed. This makes the search experience more intuitive and less rigid.
Semantic Flexibility with Real-World Accuracy
Semantic queries, such as a request for “a green T-shirt,” return related options (like light green, emerald, or turquoise – based on availability). To support this, we moved from Qdrant to PostgreSQL with pgvector, gaining better control over filtering and aggregation while keeping semantic flexibility intact.
SapientPro’s e-commerce assistant taps a large language model, so it feels like chatting with a knowledgeable store associate, not a scripted bot. It parses plain-language questions, suggests ideal products, and keeps the dialogue natural. Tight prompt design holds relevance, handles fuzzy queries, and guides shoppers to smarter buys.
Solutions
01
Hybrid AI Search Engine
Developed a search system that blends classical and vector search, giving the assistant the ability to understand both keywords and the meaning behind user input.
02
Adaptive Query Handling
The assistant can process incomplete or abstract product descriptions, recognize missing parameters, and prompt follow-up questions when needed.
03
Conversational Guidance with Real-Time Feedback
The chatbot mimics in-store support, guiding users through decisions and providing fast, clear answers such as “5 options found,” based on live product data.
04
Search Engine Refinement
Switched from Qdrant to PostgreSQL with pgvector to gain better control over filtering, aggregation, and query response accuracy.
05
Low-Friction Integration
Delivered the assistant as a flexible module ready for ShopWare integration, preserving the client’s ability to customize and extend features independently.
UNLEASH YOUR IDEA
Max Tatarchenko
CTO with 14 years of experience in solution architecture and engineering, specializing in blockchain and smart contracts. His broad expertise drives innovation across diverse technology projects.
CTO with 14 years of experience in solution architecture and engineering, specializing in blockchain and smart contracts. His broad expertise drives innovation across diverse technology projects.
UNLEASH YOUR IDEA
