Cisco, through its Outshift division, is championing a shift from 'vertical scaling' (building larger monolithic AI models) to 'horizontal scaling' (creating an ecosystem of interconnected, specialized AI agents).
The core vision is the 'Internet of Cognition,' a new infrastructure layer with protocols (proposed as Layers 8 & 9 of the OSI model) to enable agents from different vendors to discover, communicate, and collaborate securely.
Cisco is actively building this future through the open-source Agency.org project, which provides foundational plumbing for multi-agent systems, and has proven the concept's value internally with its CAPE system, which automated 40% of SRE tasks.
For enterprise adoption, the focus is on critical needs like robust guardrails, multi-vendor management, agent observability (extending OpenTelemetry), and a new security model called 'tool, task, and transaction-based access control' (T-back).
10 quotes
Concerns Raised
The current AI paradigm of vertical scaling leads to dangerous concentrations of power and capability.
A lack of interoperability standards is the primary bottleneck preventing agents from collaborating to solve complex problems.
Enterprises will not adopt agentic AI at scale without effective guardrails, security, and observability.
Opportunities Identified
Establishing an 'Internet of Cognition' to unlock a new wave of AI-driven productivity and complex problem-solving.
Creating open standards (via Agency.org) to foster a decentralized, competitive, and innovative multi-agent ecosystem.
Solving complex enterprise challenges by automating workflows that span multiple vendors, departments, and specialized agents.