Kathy Gibson reports – Not only is artificial intelligence (AI) already a widespread phenomenon, it is rapidly maturing into its second age.
The first era put the power of AI at every user’s fingertips; the second era is where agentic AI will be employed to perform tasks, says Mohanad Abuissa, director of Solutions Engineering for Cisco Middle East, Africa, Türkiye, Romania and CIS.
“With agentic AI, agents and applications, robots and humanoids will start proliferating,” he tells delegates to Cisco Connect Southern Africa 2025.
This will have a profound impact on your network, he adds. “AI will make it feel like there are 80-billion people in the world.”
With the massive implications this will have on technology architecture, companies need to ensure their infrastructure is ready to handle the load.
Abuissa explains that there are three stumbling blocks today:
- Infrastructure constraints – this is the power, compute, throughput, and bandwidth needed to cater for AI;
- The trust deficit – “A lot of users, if they don’t trust the model, won’t use it,” Abuissa says. “So you need to provide guardrails.”
- The data gap – most training to date has happened on human-generated data. But there is big potential and untapped opportunity with machine data, and AI could play a big role in unlocking this potential.
“At Cisco, we believe we are addressing these challenges head on,” Abuissa adds. “We believe we are the critical infrastructure for the AI era.”
He explains that Cisco solutions focus on three outcomes: AI-ready infrastructure; future-proof workplaces; and digital resilience.
Where the infrastructure and the workplace meet is where humans and technology intersect, so it needs to be underpinned by secure global connectivity.
AI is driving the largest expansion of data centres in the history of the world. But we need to rethink how we build these data centres so they are able to cater to AI workloads, Abuissa says.
“We need the right compute, storage, and network architectures.”
He adds that these new-look infrastructures have to land in the face of three opposing factors.
The business demand for speed, with more than half of executives keen to adopt AI as quickly as possible – indeed companies believe they need tp adopt AI within the next six months to a year or they may lose their competitive position.
Cost pressures are always present, but magnified by the investment needed to make data centres AI-ready.
Lastly, AI opens up a new threat surface. “It’s a great tool to work with, but it can also increase the threats,” Abuissa says.
Cisco has introduced technologies that address organisations’ infrastructure concerns, he adds.
“We have been working with hyperscalers to provide those technologies and we are bringing that experience and technology to the enterprise,” he says.
“Cisco can address the full stack, bringing together traditional workloads and AI workloads, encompassing platforms, compute and storage, and networking and silicon.”
New innovations include more scalable and flexible GPU servers; Hyperfabric that reimagines the data centre lifecycle; Intersight for unified intelligence across the ecosystem; AI Defense, a common substrate for AI safety and security; and AI PODS, full-stack solutions for AI use cases.
“When it comes to firewalls, we can address the entire infrastructure from branch to campus and from data centre to cloud – because you must have security everywhere.”
Live Protect addresses concerns around patching and zero-day vulnerabilities by giving Cisco networking devices vulnerability protection, while AI Defense lets customers develop, deploy, and run secure AI applications.
As well as these solutions that are available now, Cisco will soon launch additional innovations including the secure AI factory and hybrid mesh firewall, scalable devices ready for AI, and security fused into the network.
For operational simplicity, the company has announced the unification of Catalys and Meraki to provide one set of unified hardware, one set of licensing, and one management console. It will also bring Thousand Eyes into the environment for enhanced multi-layer assurance, AgenticOps to provide a cross-domain, multi-player, and purpose-built model, and AI Canvas for a reimagined user interface.