Chatbots are easy… My peer Alastair Cooke demoed one running locally on a mobile workstation. These demos are dangerous as they show how “easy” it is to use RAG to extend the usefulness of the 7 billion parameter model. However, it creates tunnel vision for use cases.
How do you use these models to create business value outside the typical chatbot scenario? Where do you start?
It all starts with data. Logic is the difference between GenAI models and search engines. GenAI models take a large data set and add predictive results. So, any data analysis area requiring additional logic is a great start.
You can begin with call center data. What makes one agent’s customer satisfaction score higher than another agent’s or a group of agents' score? It may be too much or too nuanced of a difference. It’s a great problem to try to tackle with a GenAI model.
As infrastructure professionals, think through how you’d make this data available to the model and how you’d protect it.