LLMs are a revolutionary technology, but to date, little research has examined how they can be utilized by SMEs in a cost-effective way. In a new paper co-authored by our Scientific Director Dr. Joseph Ollier, we examine the application of a LLM-driven conversational recommendation system in the field with real customers.
Results show good performance in the field with real customers (85.5% recommendation accuracy), however, high latency (5.7s) and median costs ($0.04) per interaction challenge the systems viability for lower-profit margin contexts. Nevertheless, we demonstrate the feasibility of the system in a customer-facing role, helping to diffuse the innovative potential of AI down to smaller market players.
Click here to read the pre-print.
