Intelligent AI Sales Concierge

The Challenge

Most website chatbots feel robotic because they lack a "memory." When a customer returns, they often have to repeat their preferences or needs, leading to a frustrating user experience. E-Dawat needed a solution that could act as a friendly salesperson—one that could answer complex service questions while simultaneously building a "profile" of user interests to make future interactions feel personal and seamless.

The Solution

I engineered a sophisticated AI agent using n8n and LangChain. The system was designed with a dual-layer memory architecture:

• Instant Recall: Integrated PostgreSQL to manage active chat histories, ensuring the AI maintains context throughout a single session.
• Long-Term Memory: Built a custom "Save Memory" tool linked to Supabase. This allows the AI to identify important facts—like a user's specific requirements—and store them for future visits without the user needing to repeat themselves.
• Grounded Knowledge: Connected the agent to a digital documentation library (Vector Store) so it provides 100% accurate, company-verified information while politely declining unrelated topics.

Key Results

• 24/7 Expert Engagement: Replaced static FAQ pages with an interactive assistant that provides instant, expert-level responses at any hour.
• Enhanced Personalization: Returning users are "recognized" by the system, significantly improving customer satisfaction and trust.
• Data-Driven Sales: Automatically logs noteworthy customer insights into a database, providing the sales team with pre-qualified lead data.
• Brand Integrity: By restricting the AI to specific documentation, the system eliminates "hallucinations" and ensures every response is on-brand.