A New Class of Multimodal AI Agents Can Plan, Reason And Explain

Openstream.ai, a provider of multimodal neuro-symbolic conversational artificial intelligence solutions, has received a U.S. patent for its system and method for cooperative plan-based utterance-guided multimodal dialogue.

AI virtual assistants, AI avatars, and AI voice agents created with Openstream's Eva (Enterprise Virtual Assistant) platform can understand end-user intentions by leveraging multimodal inputs and context to infer user's plans to dynamically generate human-like dialogue in real-time across any channel without a script or hallucinations.

"When people engage a brand's agent, they expect an empathetic and knowledgeable agent that understands them regardless of modality, channel, or language," said David Stark, Openstream.ai's chief marketing officer, in a statement. "AI agents must generate and understand context-specific dialogue using all end-user inputs in real time to seamlessly engage them in unscripted conversations that feel natural and help them to achieve their goals. Without this, brand perception and long-term relationships are at risk."

AI agents created with Eva can engage in complex, multimodal, multiturn, multi-obstacle conversations continually informed by context across any channel or language 24/7/365.

"This patented technology shows how advanced our approach for dialogue management is and how it can handle conversations that are hard for many other contemporary solutions," said Magnus Revang, Openstream.ai's chief product officer, in a statement. "Our system goes far beyond intent-slot, and frame-based systems to reason and plan using very complex constraints and domain knowledge."