The race to build loyalty with humans and agents is accelerating. The percentage of retailers that have deployed agentic systems remains in the single digits. However, retail leaders are well aware that the companies that control the agentic buying journey will be the true winners of modern commerce. McKinsey predicts that agents will influence $3-5 trillion in global spending by 2030, including $1 trillion in the US B2C retail market.
A leading US wellness company is helping to lead the charge. With a strong focus on providing personalized, scientific wellness solutions, the company’s leaders saw an opportunity to implement an agentic AI system to support its direct-to-consumer (D2C) sales model by delivering personalized product recommendations on its website. The company’s products, such as nutritional supplements and diagnostic tests, target a diverse array of buyers, including individuals, healthcare practitioners, athletic trainers, and professional athletes. The wellness company’s goal was to increase conversion rates, improve customer engagement, and drive repeat purchases.
Tredence worked with the firm to deploy a customer-facing GPT solution that offers personalized health coaching, product assistance, and recommendations.
Designing a Scalable Chatbot for High Performance in a Regulated Industry
The wellness company operates a multi-cloud IT landscape across AWS and Azure, with the majority of its data and analytics workloads running on Azure Databricks as the enterprise data and AI platform. However, it had customer data siloed across multiple operational systems, limiting its ability to build a unified customer 360 (C360) view.
The firm needed an AI-ready centralized data foundation to support advanced analytics and AI use cases. The customized GPT-based assistant would then be trained on the company’s product catalog, customer purchase history, and health-related interactions. And governance and guardrails would enable the company to maintain compliance, safety, and trust, as it offered targeted health and wellness recommendations.
Tredence designed a phased, agentic solution roadmap built natively on Azure Databricks to integrate data, intelligence, and activation seamlessly. This would enable the organization to evolve from a basic conversational assistant to a fully personalized, autonomous AI agent.
Developing an AI-Ready Data Foundation on Azure Databricks
Tredence established a robust data foundation on Azure Databricks, serving as the backbone for customer GPT, analytics, activation, and insights use cases.
The team next built a centralized customer 360 data foundation on Databricks to ingest, unify, and govern customer data from multiple sources, including:
- AWS-hosted data: Transaction, product, customer, and subscription data
- ERP systems: Loyalty programs, returns, and order data
- Health data: Health profiles, quizzes, and test data hosted on AWS
- Salesforce Service Cloud: Contact center interactions and support data
Using Delta Lake and Databricks workflows, Tredence implemented data harmonization pipelines that embedded data quality checks, customer identity resolution, and segmentation and feature engineering. We further ensured AI/ML model readiness and integrated MarTech and downstream activations. With this work, we established Databricks as the single source of truth for customer intelligence.
Deploying an AI Product Assistant
Next, Tredence delivered an AI-powered product assistant within a controlled, compliant scope, training it on factual product knowledge to deliver immediate business value while minimizing risk.
The customer-facing chatbot enables shoppers to discover and search for products and use natural language to gain detailed product information and explanations of how features align with the Dietary Supplemental Health and Education Act of 1994 (DSHEA)-aligned compliance controls.
The Databricks platform provides:
- Multi-modal RAG pipelines using Databricks Vector Search
- Dynamic knowledge base updates from curated product data
- Scalable and secure model deployment using Databricks Model Serving
Integrating a Lifestyle and Health Knowledge Agent
Next, Tredence developed a GenAI-powered lifestyle and health conversational agent that delivers prescriptive responses based on user intent and context. The agent:
- Enables rich, multi-theme conversations across lifestyle and health domains
- Adheres to enterprise-grade guardrails and regulatory sanitization
- Personalizes interactions using long-term conversational memory
The agent is backed by Databricks capabilities, including:
- An agentic architecture with thought, planning, and execution layers
- End-to-end LLMOps and a responsible AI framework on Databricks
- Secure storage and retrieval of past interactions using Delta tables
Providing a Personalized Wellness Tech Assistant
In the final phase, Tredence delivered a fully personalized, autonomous wellness assistant driven by customer purchase behavior, health context, and historical conversations. The assistant provides:
- A personalized AI health coach that has deeper and broader conversational abilities
- Rich text and image-based interactions
- End-to-end customer workflows—from discovery and recommendation to purchase
It is enabled by advanced platform capabilities on Databricks, which enable the wellness company’s teams to:
- Build, deploy, and serve custom ML and GenAI models
- Integrate with multiple downstream systems for fulfillment and activation
- Benefit from a fully autonomous, self-learning agentic system with dynamic model and knowledge updates
Solution Architecture at a Glance
Tredence delivered a Gen AI-ready data platform architecture.
The solution architecture comprises two core layers: a data foundation and a GenAI application layer, both built natively on Databricks.
Tredence implemented a data foundation using Databricks Delta Lake, leveraging a medallion architecture (bronze, silver, gold), to ensure scalable ingestion, data harmonization, and analytics-ready data. End-to-end data governance, security, and access control were enforced using Databricks Unity Catalog, enabling centralized metadata management, fine-grained permissions, and lineage tracking.
Curated, business-ready datasets from the gold layer were exposed in a consumable semantic format, serving as the primary knowledge source for the customer GPT and agentic workflows.
The entire platform was orchestrated using Databricks Workflows, ensuring reliable scheduling, dependency management, and end-to-end pipeline automation. Machine learning and GenAI models were developed directly on Databricks and tracked, versioned, and published using MLflow, enabling seamless model deployment, monitoring, and lifecycle management.
Agentic Architecture on Databricks
Tredence built a full-stack, agentic AI application natively on the Databricks platform, leveraging Databricks’ unified capabilities for data, AI, and governance.
The solution used the LangGraph and LangChain frameworks, combined with Databricks Foundation Models, to design and orchestrate a multi-agent system. These agentic workflows were deployed as Databricks Model Serving endpoints, which were securely exposed to the front-end application for real-time inference and interaction.
To enable an end-to-end, customized Customer GPT experience, the solution incorporated multiple specialized agents, including:
- Orchestration agent: Coordinates agent selection, execution flow, and context passing
- Text-to-SQL agent: Translates natural language queries into optimized SQL against curated Delta tables
- Data processing agent: Performs real-time data enrichment and transformations
- Pricing agent: Provides pricing intelligence and promotional insights
- Inventory agent: Delivers inventory availability, constraints, and fulfillment context
- Recommendation agent: Generates personalized product recommendations based on customer behavior and context
Together, these agents operate as a cohesive, intelligent system, enabling context-aware reasoning, personalization, and scalable GenAI-driven interactions—all governed and deployed through the Databricks Lakehouse and AI platform.
Getting Ready to Win with Agentic Commerce
By leveraging Azure Databricks as the unified data and AI platform, the wellness company was able to:
- Establish a trusted customer 360 foundation
- Accelerate personalized recommendations and digital engagement
- Enable compliant, scalable GenAI adoption
- Lay the groundwork for a future-ready, agentic AI ecosystem
The company’s consultative sales model will help it connect with discerning consumers, build trust, and sell product bundles for higher ROI.
AUTHOR - FOLLOW
Maulik Divakar Dixit
Senior Director, Data Engineering, <br>Databricks Champion<br>Databricks MVP
Next Topic
Beyond Excel: Building a GenAI-Powered Spend Intelligence Engine for Strategic Procurement
Next Topic
