The AI in Business Podcast From Multi Agent Systems to Institutional Learning in the Enterprise - with Papi Menon of Outshift by Cisco
13 snips
Mar 19, 2026 Papi Menon, VP of Product Management and CPO at Outshift by Cisco, builds enterprise agentic AI and incubates emerging tech. He discusses why multi-agent projects stall, the gap between mere connectivity and shared cognitive layers, and how to pick low-risk, high-impact experiments. He also covers interoperability, when to build versus buy, and how to preserve optionality while scaling.
AI Snips
Chapters
Transcript
Episode notes
Lack Of A Shared Cognitive Layer Limits Scale
- Enterprises hit friction not from model quality but from lacking a shared cognitive layer that lets agents attach meaning and context across boundaries.
- Papi Menon explains agents can connect syntactically today but cannot collectively learn or refine a shared mission, which blocks multi-agent scale.
Financial Services Network Debugging Use Case
- A financial services design partner built a production network debugging tool that simulates changes on a digital twin to prevent misconfigurations and outages.
- Multiple specialized agents debug different aspects and currently share context one-off, showing the need for a common cognition fabric.
Enterprise Agent Deployments Will Be Heterogeneous
- The agentic enterprise will be heterogeneous with many model sizes, deployment modes, and ownership models working together.
- Papi Menon emphasizes optionality: on-prem, cloud, APIs, native agents and diverse models must interoperate.