Enhancing a Product Availability Voice Experience with Generative AI
As part of Lowe's Conversational AI team, I led the design and implementation of a major enhancement to our product availability IVR experience. The goal was to bridge the gap between structured item number queries and unstructured product descriptions using generative AI, ultimately improving customer containment and satisfaction while increasing business value and revenue.
Timeline
4 Months - Initial Project Scope
Ongoing - Analyzation & Optimization
Project Type
IVR - Interactive Voice Response
Roles
AI Product Designer & Prompt Experience Designer
The existing product availability IVR experience provided structured responses when customers gave an item number. In this case, the system returned:
Inventory quantity
In-store location
SMS offer to view the product online
However, when customers gave a product description instead of an item number, the system offered only vague responses such as "I found a few items that meet your description," followed by an SMS to a generic product landing page. This limited interaction created a poor customer experience and failed to drive value for more open-ended queries.
The Problem
The Opportunity
With advancements in LLM (Large Language Model) technology, my team saw a chance to:
Understand natural language descriptions more effectively
Generate informative and persuasive product summaries
Improve overall containment by guiding more calls to completion without agent transfer
My Role & Design Approach
I owned the product design and prompt architecture for the new generative AI experience, focusing on two primary LLM-powered enhancements:
Backend LLM Decision Logic:
Determines if a customer provided enough information in their description
If not, the IVR reprompts the caller for more detail
If yes, initiates a search and sends context to the next LLM layer
Product Summary Generation:
Custom responses based on customer input
If the input is an item number: concise summary with name, price, and availability
If the input is a description: persuasive and contextual summary highlighting use cases and benefits
I crafted detailed flow designs that included:
Success paths
Failure scenarios (API or LLM failure)
Max attempt fallbacks
Failover to existing product availability IVR experience to maintain continuity
Prompt Tuning & Testing
I led the design of prompts for both LLM calls:
The backend input analysis (signal quality, intent clarity)
The customer-facing product summary (tone, structure, information hierarchy)
We piloted the experience in one store, then expanded to 10 stores based on initial performance. I coordinated bi-weekly reviews with business partners to iterate on prompt designs and ensure alignment with business rules.
Iterative Improvements
Production conversations revealed the need to further differentiate between item number and description flows. I redesigned the prompt structure accordingly:
Item number: factual, transactional response
Description: persuasive, conversational response
Results
Containment for product-related calls, our highest volume category, improved 5% overall after the enhancement, and improved 14% for callers who specifically heard a GenAI response.
An increase in additional monthly revenue attributed to enhanced containment.
Nationwide rollout to all Lowe's stores after successful pilot.
What’s Next
With the transition to a new tech platform, I am redesigning the entire product inquiry experience. Future plans include:
Deeper generative AI integration for vague product requests
Smart routing based on context
Seamless transitions into other IVR features like lead forms, scheduling, or enhanced back-and-forth voice experiences
Impact
This project exemplifies how thoughtful LLM integration, strategic design, and prompt engineering can transform a legacy IVR experience into a smarter, revenue-driving system that aligns both user needs and business outcomes.