Optimizing E-commerce Conversion with a Proactive AI Sales Agent

The client is a startup making the product for mid-sized e-commerce retailers with a broad customer base across multiple countries.

The client would like to develop a universal e-commerce solution adaptable to be implemented in online shops of a wide array of industries, from real estate to art and fashion. The sales logic is different for each domain, so the solution should be adaptable for different good catalogue structures, goods features, and sales pipelines.

A solution must function as a proactive sales assistant, not only answering questions but guiding users through the sales funnel, anticipating their needs, and driving them toward a confident purchase decision.

Quantum developed a custom multilingual AI Sales Agent based on the AgentFlow platform, specifically for the client’s e-commerce environment. The Agent aims to serve not just as a support tool but also as a proactive sales driver.

The Agent is powered by a two-level approach for user intent analysis:

1. Intent Classification: At its core, the Agent deeply understands where each conversation is within the sales funnel, evaluates whether all conditions for purchase have been met (questions about shipping, price, or features), and handles objections accordingly to the user’s sales strategies.
2. Intent Handling: Based on the classified intent, a specialized component is tuned to handle the request – deploying a product proposal component for browsing, using a knowledge base to resolve objections, proactively pushing the conversation forward, asking strategic questions to guide the user toward a decision and close the sale, or initiating a seamless handover to a human operator for complex issues.

These components can be adapted for each user based on it’s unique needs.

To ensure the Agent could perform with high accuracy and a natural ToV-powered language, we tuned and configured it using a vast amount of the client’s business data:

• The Product Catalog: client’s product database, including SKUs, specifications, pricing, images, and real-time inventory levels.
• Corporate Knowledge Base: a knowledge base containing detailed information on company policies, shipping and return procedures, warranty details, and Frequently Asked Questions (FAQs).
• Historical Customer Dialogues: anonymized chat and email logs from the client’s CRM were used to tune the Agent on the nuances of customer inquiries, common objections, and practical, human-like response strategies.

Fully integrated via API with the client’s online shop and CRM, the Agent was deployed across the client’s most critical channels – website, WhatsApp, Instagram, and Telegram for a consistent customer experience.

Quantum’s AI Sales Agent is powered by a variety of LLMs like Mistral 7x8b, GPT-4o, Claude Sonnet within the Langchain framework, optimizing query responses in customer service. This combination, featuring PostgreSQL for precise document retrieval, was chosen for its accuracy, effectiveness in complex query handling and flexible deployment options: cloud, on-premise, hybrid.

Front-end integration options ensure a smooth customer experience, with choices such as web chat widgets, mobile SDKs, and messaging platform integrations. On the back end, the AI Sales Agent provides seamless integration with existing corporate solutions via OpenAPI.

• 70% Automation of Routine Queries: The Agent successfully handled repetitive customer questions, allowing the human support team to focus on complex, high-value interactions.
• 24/7 Sales and Support: The client’s store became an always-on sales channel, capturing revenue around the clock and improving customer satisfaction.
• Expanded Market Reach with seamless multilingual support.
• Streamlined Operations through API integration with the back office.
• Increased Engagement: With multichannel communication capabilities, the AI Sales Agent engaged customers on their preferred platforms, creating a seamless and consistent customer journey.