The Internet of Agents and What It Means for Enterprise Leaders - with Vijoy Pandey of Outshift by Cisco
Summary
Vijoy Pandey, Special Vice President and General Manager for Outshift by Cisco, discusses the enterprise shift from deterministic to probabilistic, agent-driven architectures. This paradigm shift introduces challenges in interoperability, governance, and safe AI deployment at scale, particularly in regulated and mission-critical environments. Outshift, Cisco's internal incubator, focuses on reducing technology and market risk for emerging tech like agentic AI. Pandey highlights the need for new access control mechanisms, such as task- and transaction-based access, to manage autonomous agents operating at machine speed. He also emphasizes the importance of open, interoperable platforms for agent discovery, identity, collaboration, and evaluation, citing a multi-agent healthcare contact center workflow as a practical example of agents working in teams to achieve business outcomes.
Key takeaway
For AI Architects and Directors of AI/ML guiding enterprise AI strategy, recognize that scaling agentic AI demands a fundamental shift from static role-based access to dynamic, task- and transaction-level controls. Prioritize building interoperable platforms that support agent discovery, identity, and multimodal communication to manage autonomous systems effectively and reduce operational risk in critical environments.
Key insights
Enterprises are transitioning to probabilistic, agent-driven AI, necessitating new approaches to interoperability, governance, and access control.
Principles
- Agentic AI operates probabilistically at machine speed and scale.
- Zero trust principles must extend to task and transaction levels for agents.
- Multi-agent systems require open, interoperable collaboration platforms.
Method
Implement task- and transaction-based access control for agents, moving beyond traditional role-based methods. Develop platforms for agent discovery, identity, communication, and autonomous action to facilitate multi-agent workflows.
In practice
- Use multi-agent systems to reduce configuration errors in telecom.
- Streamline patient routing in healthcare contact centers.
- Employ simulation layers to validate operational changes.
Topics
- Agentic AI
- Probabilistic Computing
- Multi-Agent Systems
- AI Governance
- Enterprise AI Adoption
Best for: VP of Engineering/Data, Director of AI/ML, AI Architect, CTO, Executive, AI Product Manager
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by The AI in Business Podcast.