Lowe’s – Smart Navigation via AI Assistant

Retail & E-commerce

Lowe’s – AI Virtual Assistant for In-Store Navigation

At Lowe’s, we introduced an AI-driven virtual assistant that redefined how customers shop in-store. Deployed through interactive kiosks, this assistant uses natural language processing to understand customer queries and guide them to the right product aisles in real time. It can answer DIY-related questions, provide product comparisons, and even scan barcodes using computer vision for instant info. The assistant is designed to operate without human intervention, drastically reducing wait times and easing staff workload. With this AI solution, Lowe’s enhanced the overall shopping experience, increased in-store conversions, and encouraged more self-service behavior among customers.

Project Overview

Client: Lowe’s (2,000+ US stores, $97B revenue)
Challenge:
- 68% of customers left stores frustrated due to difficulty finding products
- o Staff spent 30% of time on directional queries vs. sales
Aim: Design a voice-enabled AI assistant to perform tasks.
- Reduce product search time by ≥50%
- Handle 80% of routine requests independently.
- Integrate with Lowe’s mobile app & inventory systems
My Role: Lead Conversational AI Engineer (GenX Software Team)
Team: 6 (NLP Engineers, CV Specialists, UX Designers)
Timeline: 7 months (pilot → rollout of 500 stores)

“This isn’t just a wayfinder—it’s like embedding a veteran hardware associate in every customer’s pocket.”

Lowe’s Chief Digital Officer

The Challenge

Critical Pain Points:

Average 8-minute search time for specific items (e.g., "½" copper elbows")
15% staff turnover exacerbated knowledge gaps
40% of customers used competitors’ apps while in Lowe’s stores

Technical Hurdles:

Noisy store environments (accuracy <70% with off-the-shelf ASR)
Real-time aisle mapping across 20K+ SKU layouts
Process barcode scans in low-visibility conditions.

Tech Stack

Component	Technologies
Conversational AI	Rasa, Wav2Vec 2.0, BERT
Computer Vision	YOLOv7, OpenCV
Knowledge Management	Neo4j, AWS Neptune
Mobile Integration	React Native + ARCore
Analytics	Snowflake, Mixpanel

Key Innovations

The assistant employed NLP to direct customers in-store, answer DIY questions, and provide comparisons. It reduced staff burden by enabling autonomous, real-time product navigation. This self-serve kiosk improved conversions and user satisfaction.

“Follow-Me” AR Mode

Mobile app overlay guided users via arrow markers

Result: 62% faster navigation than store maps

DIY Troubleshooting

NLP parsed vague queries (“fix leaky pipe”) into part lists

Result: 28% increase in basket size for project-related purchases

Staff Assist Mode

Escalate sophisticated inquiries to specialists with full context

Result: 41% reduction in employee training time

Our AI/ML Architecture

Core Models

Conversational NLP Engine:
- Fine-tuned BERT with retail-specific intent classification
- Multi-lingual support (English, Spanish, French)
- Mapped user queries to 600K+ SKUs
Computer Vision Module:
- Barcode and product label scanner via camera kiosk.
- Visual similarity for product suggestions
Navigation & Recommendation System:
- DIY solution & product recommender database
- Knowledge graph for product comparisons & DIY guidance

Data Pipeline

Sources:
- Live Lowe’s POS inventory sync
- Store layout & planograms
- Barcode & SKU database
Processing:
- Google Cloud Dataflow + BigQuery (handling 100K+ queries/day)