The ‘Internet of Agents’: How organizations will manage thousands of new digital employees

The rapid development of AI agent technology is prompting companies to integrate thousands of specialized AI assistants across their operations. Cisco’s Outshift division is leading efforts to create an Internet of Agents (IoA) that would enable these AI systems to communicate and collaborate effectively at scale.

The scale challenge: Organizations are expected to deploy upwards of 20,000 AI agents each, with large enterprises potentially managing hundreds of thousands of agents across their operations.

• Each AI agent functions as a subject matter expert within its specific domain and context
• These agents need to communicate across vendor boundaries, security domains, and different operational profiles
• The complexity increases as agents exchange multimodal data including video, images, and text while managing workflows

Technical requirements: The Internet of Agents requires sophisticated infrastructure beyond traditional internet capabilities to enable effective AI collaboration.

• AI agents must share complex states and reasoning processes while making coordinated decisions in real-time
• The system needs to maintain security and trust while handling probabilistic communication between agents
• Agents require access to various tools, training data, knowledge bases, and dynamic database systems

Code development example: The software development process illustrates the need for universal AI agent infrastructure.

• Multiple specialized agents handle different aspects like syntax checking, security, scaling, and compliance
• These agents often come from different providers and must work with developer-specific code bases and APIs
• The entire system needs to operate in a trustworthy, scalable, and explainable manner

Open ecosystem imperative: An open, interoperable framework is essential for the Internet of Agents to succeed.

• Closed systems and walled gardens limit value creation for developers, operators, and consumers
• Historical precedents like Linux, Apache, and Kubernetes demonstrate how open ecosystems drive technological progress
• Standardization and open specifications are crucial for agent discoverability, reputation management, and communication

Industry collaboration: Building the Internet of Agents requires unprecedented cooperation across the technology sector.

• Organizations must work together to develop specifications and standards in open source
• Cisco is actively bringing together stakeholders who share this vision
• Software builders, consumers, and operators are being encouraged to participate in developing the framework

Future implications: The success of the Internet of Agents hinges on the technology industry’s ability to overcome significant technical and collaborative challenges while maintaining security and interoperability standards. The initiative’s outcome could determine how effectively organizations can leverage AI agents to transform their operations and create value across different domains.