Tradition and Technology
Aroma di Montagna is renowned for its “Farm-to-Cup” excellence. But how do you bring that deep expertise from the roast master to thousands of simultaneous digital customers? At Thiago Dias, we took on that exact challenge.
The Challenge: The Omnipresent Expert
Specialty coffee consumers are demanding. They want to know everything: from the altitude of the farm and the sensory notes of the roast to the ideal brewing method for each specific bean. Keeping human experts available 24/7 to answer these technical nuances is a logistical and financial impossibility at scale.
The Solution: Layered AI Architecture
We didn’t just want a generic chatbot; we wanted a digital extension of the brand’s wisdom.
1. RAG (Retrieval-Augmented Generation)
We linked Aroma di Montagna’s internal knowledge base—which includes technical specs for every harvest and detailed preparation guides—to a vector-based search engine. This ensures the AI never “hallucinates” or invents information, always citing precise data about the products.
2. Finetuning Open-Weights Models
To ensure brand voice and data privacy, we opted for Open Weights models (like Llama 3 and Mistral). We conducted a thorough finetuning process so the AI could understand the specific terminology of the specialty coffee market and adopt Aroma di Montagna’s signature tone: welcoming, yet technically rigorous.
3. Advanced Prompt Engineering
Behavioral guidelines were refined so the AI could act as a true “Coffee Sommelier,” capable of suggesting pairings and extraction techniques based on a user’s unique flavor profile and past preferences.
Impactful Results
Today, Aroma di Montagna’s AI agent doesn’t just answer questions; it educates the consumer. The result was a significant increase in average order value (AOV), as customers feel more confident exploring limited-run, high-end beans after receiving a personalized, expert-level digital consultation.
Ready to transform your company’s knowledge into an autonomous competitive advantage?