Business conversations in the context of user trying to buy something or book an appointment are different from the normal conversations we do in traditional AI chat. And most of the large language models work very well in AI interactions involving coding, software engineering and data analysis. It is surprising that the LLMs are able to achieve this tremendously difficult tasks which many would have thought impossible before 2020.

When it comes to business conversations, we have found that we still need some fine tuning when model is trying to serve local population. After some training we are using good improvements in the quality of responses and success of the business outcomes. Model is able to complete the business users tasks better.

Our business tasks include booking appointments, pricing correctly, finding products and services. We used to have greater errors rates before fine tuning. After looking at the patterns of thousands of interactions in business context, model is performing and getting better.