- Cisco is working to advance AI agents beyond simple communication
- Its so-called Internet of Cognition proposes a new semantic protocol architecture
- Telcos need to prepare for semantic interoperability layers in future networks
Cisco is hard at work building a new protocol architecture — a so-called "Internet of Cognition" — designed to let AI agents not only communicate, but also reason, collaborate and innovate in the next era of AI intelligence. Telcos need to prepare for what’s coming.
The vendor has hitherto been working to enable what it calls the "Internet of Agents" through the development of protocols like AGNTCY, which enables agent discovery, identity and access management, and communication. But the company is now laying the groundwork for an Internet of Cognition that will move the needle beyond simple information sharing to enable agents to share intent and context.
In a nutshell, Cisco wants to move from syntactic communications to semantics.
“These agents are right now just dealing with syntactic communication through MCP, A2A and all of the stuff that AGNTCY provides…But there is no understanding of the semantic connotations behind the payload of that communication,” Vijoy Pandey, GM and SVP of Outshift by Cisco, told Fierce.
He continued: “So, we are going a layer up from syntactic and we are looking at the semantic layer, and we are saying can we do those three things: Can we share intent? Can we share memory and knowledge in context? And can we enable collective innovation across agentic workflows?”
The Internet of Cognition comprises three pillars: cognition state protocols, which address shared intent and coordination; a cognition fabric, which offers the ability to create institution-wide working memories; and cognition engines, which can either be used to accelerate innovation or enforce guardrails for compliance.
What are the Internet of Cognition hurdles?
AvidThink Founder Roy Chua told Fierce that Cisco has undertaken a rather tricky task.
“The challenge of ensuring agents have genuinely aligned semantic understanding — not just compatible message formats — requires solving deep problems in representation learning and ontology alignment,” he said. “There's plenty of work to go from ‘shared message schema’ to ‘shared meaning.’”
Getting everyone to adopt Cisco’s cognition state protocols could be complicated by the fundamental differences that exist today between model architectures, as well as security and intellectual property concerns, Chua said. Additionally, he pointed out that while the cognition fabric may solve multi-agent context issues, the industry will still have to sort out multi-agent coordination problems.
What does it means for telcos?
So, what does all of this mean for telcos? According to Chua, Cisco’s concepts “align well” well ongoing telco investments in intent-based networking and multi-vendor orchestration. He added that given that the 6G standardization process is already underway, some of these ideas could very well end up gaining traction there as well.
Telcos should pay attention to progress on protocols like AGNTCY, model context protocol (MCP) and Agent-to-Agent (A2A) and “prepare internal architectures for eventual semantic interoperability layers,” Chua said.
Indeed, Pandey said AI is bringing two new layers into the traditional seven-layer communications stack: syntactic protocols at Layer 8 and semantic cognition protocols at Layer 9.
“That implies that every networking stack moving forward will have to deal with not just Layer 4, not just Layer 7, but also the probabilistic communication at Layer 9,” Pandey said. “Telcos need to prep for what’s coming at Layer 9.”
For now, Chua warned telcos against committing significant capital towards new architectures until there’s more clarity on specifications and industry adoption trends.
“We'll likely know in 1-2 years whether this becomes a foundational standard or an effort like intent-based networking, which until recently, couldn't meet its own stated goals,” Chua concluded.