Artificial intelligence (AI) startup Yellow.ai launches the Orchestrator LLM (large language model) which combines the capabilities of 15 AI models including OpenAI’s GPT, Meta’s LLaMA, Anthropic’s Claude for enterprise applications.

Rashid Khan, co-founder and chief products officer at Yellow.ai says, “Our base model is the open-source LLaMA 2B. On top of that, we are offering model-as-a-service on the Orchestrator platform which orchestrates across models for tasks they are best at such as GPT for reasoning and Cohere for ranking and so on.”

The Orchestrator LLM employs precise question recognition, conversation history retention, and cross-channel context switching to tackle the issue of interrupted customer conversations.

Businesses can save operating expenses by sixty percent and increase agent productivity by fifty percent by not having to train the model.

“For instance, in a sales use-case, model integration has reduced the number of interactions going to human agents by 60 percent. Also, for a two-wheeler auto company, bookings made on a went up by 15,000 for a campaign which created a buzz about the vehicle,” Khan says.

Even if YellowG LLM, the company's initial model, will power 60–70 percent of workloads, the Orchestrator LLM will further improve these results. Additionally, it introduced the Komodo-7B, the first model for customer service in more than 11 regional languages in Indonesia.

The San Mateo, California-based customer service automation startup serves over 1100 enterprises, including Sony, Domino's, Hyundai, Volkswagen, ITC Ltd, OYO, Skoda Auto, and Asian Paints.

“In the February-April quarter, the company has doubled its bookings in the US market. We are currently generating $45-50 million annual recurring revenue and inching very close to profitability,” Khan adds.