Open Data

Orienting AI agents: How Phasis builds models for enterprise - SiliconANGLE

Orienting AI agents: How Phasis builds models for enterprise - SiliconANGLE

Phasis Inc. addresses a critical gap in AI development by creating a "World API" that serves as an orientation framework for autonomous agents. Recognizing that reasoning capabilities are advancing faster than domain expertise, the company utilizes graph-based ontologies to map complex realities, such as culture or enterprise structures. This approach ensures agents do not operate in isolation but are anchored in verified, up-to-date information before executing tasks. The core innovation lies in the distinction between reasoning and orientation. While models get better at logic, they lack an innate, durable understanding of the world. Phasis provides this missing layer by delivering relevant context, evidence, and uncertainties to agents at the start of any workflow. This pre-task orientation prevents agents from piecing together fragmented knowledge repeatedly, allowing them to act with precision and reliability based on a shared, maintained state of reality. This strategy is highly relevant to open data because it demonstrates how structured, verified knowledge graphs can enhance AI utility beyond raw processing power. By treating real-world domains as maintainable, evidence-backed states, Phasis highlights the necessity of high-quality, accessible data infrastructure. It suggests that the future of effective AI depends not just on larger models, but on robust, open frameworks that allow agents to continuously orient themselves against accurate, shared knowledge bases.

Source: siliconangle.com
Published on 2026-10-13

Related news