In order to create an Einstein Service Agent customized for an enterprise, its developers need to access the Einstein 1 platform, start creating a new copilot by giving its specific name and purpose, and then expose it to service topics, Nichols explained, adding that once the service topics are mapped and a LLM is selected, developers can assign knowledge-based actions to the agent.
These knowledge-based actions, according to the company, make the agent autonomous or allow the agent or chatbot to take actions without having a human in the loop, differentiating it from regular programmed chatbots or service agents.
One of the main differences between an older predictive chatbot and the Einstein Service Agent is its ability to understand intent because of its LLM base, said Keith Kirkpatrick, research director at The Futurum Group.