Observe.AI, the AI start-up founded by Swapnil Jain, Akash Singh and Sharath Keshava Narayana in 2017, has launched its contact center LLM (large language model) based on 30 billion parameters. The company maintains that the first-of-its-kind LLM is found to be 35% more accurate than GPT3.5 in automatically summarizing conversations and 33% more accurate in identifying customer or agent sentiment during calls.
Observe.AI’s LLM is built specifically for contact center use cases on various parameters (7B, 13B, and 30B parameters) so that it performs well for use cases such as summarization, generating coaching tips or responses to customer questions.
The model is said to effectively address some fundamental issues associated with out-of-the-box language models like GPT. “The downside of out-of-box LMs like GPT, is that they are trained on what is called “clean text”, e.g. language that is simplified and distilled to root worlds that are easy for a machine to understand. This clean data looks very different from the natural conversation between agents and customers, where the disfluencies (stuttering, drawing out sounds, or repetitions) and non-grammatical utterances are common,” wrote Satarupa Chatterjee, Senior Director, Product Marketing for Observe.AI, in a blogpost.
Observe.AI has introduced a number of Generative AI offerings based on its proprietary contact center LLM, namely Knowledge AI, Auto Summary and Auto Coaching.
The San Francisco-based startup says that its new generative AI offerings mark a new era in agency efficiency and performance. The company has a workforce of over 250 employees and serves over 200 businesses globally. It has raised $213 million to date, with its round of Series C funding in 2022.
