Over ten years ago, I had a thought-provoking chat with the operations leader responsible for one of the world's largest Hyperscaler networks. For privacy, let us call him Adam and his company Foo. Adam remarked, “Stefan, my aim is to run this network with just one person, and I’m getting pretty close!”

Foo’s network team had built a robust automation library specifically designed for monitoring and managing their network. This required a major investment, which fits with their company culture and strategy. Plus, Foo’s scale and buying power enabled them to demand that vendors supply the right tools, adhere to standards, open-up APIs, and deliver fast support—if they wanted to keep Foo as a client.

Historical note: Google introduced the concept of Site Reliability Engineering (SRE) around 2003.

Sometime after my talk with Adam, when I was a Software Development Executive at Dell, our internal IT leaders often discussed deploying SRE teams to improve efficiency, reduce costs, and increase reliability in managing Dell Infrastructure. The presentations reminded me at the time of Adam’s comments regarding running Foo’s network with a team of one. Building an SRE team made sense in theory but was a tough nut to crack even for a large company. This called for merging NetOps and DevOps, and for infrastructure to be open enough to support code deployment and feature access. Those presentations and strategies led to no change on the ground.

I am sharing these memories to set the stage for answering the question:

Can you run a network with a team of one?

The SRE team’s mission—then and now—has been to stay lean, serving as both the backbone of infrastructure operations and as developers who can fix bugs or add automation features. But as I mentioned, this is much harder to put into practice, even at big companies. It takes executive support and, more importantly, a cultural shift. It always looks clean on slides.

For more than three decades, I have managed development teams, large and small, with a focus on Orchestration, Telecommunication, and Core Networking. I have been intimately involved in rolling out Open network operating systems like SONiC and the ill-fated OPX (Open-Switch under the Linux Foundation). I have managed networking support and sustaining organizations and served as executive sponsor for various customers. Therefore, I have a unique understanding of the challenges integrating NetOps and DevOps. A simple example is that without a shift in company culture, it is difficult to recruit an impactful development engineering team as part of the Network Operations team. Even if recruited, the Development Operations team will create barriers for the Network Operations team to perform any meaningful development. Therefore, deployment of SRE not only requires the right team, but certainly the right culture top-down.

Need to Look at the Forest and Not the Trees.

Infrastructure orchestration companies have mostly tackled automation for specific, well-bounded use cases—such as monitoring homogeneous data center networks or managing uniform infrastructure throughout its lifecycle. These are the simplest and most common scenarios.

For example, automation for rolling out Enterprise Campus Fabrics is either limited to a vertically integrated solution company (in this case hardware, OS, Orchestration is tied together) or Dorado Cruz. Stacked solutions tend to lock customers into a single vendor, while unstacked ones, though more flexible, are harder to manage unless you use tools like Cruz to create a virtual stack. The teams running these networks are not talking about SRE, CLOS, or Fabric—they are focused on access, aggregation, distribution frames, L2 and L3, and so on. Their main concerns are security and making sure users can connect with the required performance via Wi-Fi or VPN. Additionally, in my experience, cost is a bigger factor in campus deployments, and there is always an ongoing debate over what should stay on-prem versus move to the cloud.

I digress – Can you run your network with a team of one? My detour highlights the significant gap between the complexity of network deployments and the capabilities of enterprise staff, a critical factor to consider when addressing this question, unless you have the right tools, strategies, and innovations to bridge that gap and make the vision of a one-person network team a reality.

Bridging the Gap: Robust Orchestration & Management and Age of AI.

At Dorado, for more than a few decades, we have been a leader in addressing the needs of customers requiring a management platform. Early on, we recognized the headaches of managing mixed-vendor networks. So, we designed an architecture around managed objects, defined their relationships, and built a flexible model that expands as new devices are introduced. We made it scalable by decoupling services so that processes could spread across distributed compute resources, and we built tuning mechanisms for connectivity as deployments grew. We added Fabric Orchestration and monitoring. Why is this important in answering the question? Let us look at the age of Agentic AI and tie it all together.

We are now in the era of AI—specifically, Agentic AI. I have seen plenty of presentations pitching orchestration and management tools that use Natural Language for tasks like, 'Apply VLAN 2 to all my switches.' Interesting, sure—but it misses the bigger picture. It is like focusing on a single tree while ignoring the entire forest. There are no shortages of articles discussing how AI serves as an enabler, not a replacement, in modern orchestration and network management, so I will not dive into the many ways AI fits into the picture by focusing on the different trees in the forest.

Instead, the real question I am addressing is this: Is it truly possible to run a network with a team of one?

That’s the BIG Challenge We Set Out to Tackle.

Once we identified the challenge, we asked what it takes to run a network efficiently. Our conclusion: you need an SRE team. And what does an SRE team do? They use tools to monitor, correlate, and predict, continually develop and integrate, and ultimately optimize network performance. At minimum, the SRE team needs at least one expert who understands operations and oversees the rest; without that leader, you risk blackholing traffic and bringing everything down—I have seen it happen, with millions lost due to downtime. Back to Adam and Foo’s network. Foo had optimized its network architecture for simplicity. Simple to debug, simple to configure and deploy and integrate into a single source of truth platforms. Still, they had to invest heavily in building their libraries while maintaining expertise.

At Dorado we started thinking of the SRE team as trained AI agents. These agents need to be trained not only on the network’s specifics, but also on the orchestration platform’s architecture—so an operator’s query does not accidentally cripple the network. We found Dorado’s Cruz is an optimal platform for integrating Agentic AI. Each agent acts as a member of the SRE team and learns on the job and can be trained on the open architectures of the underlying infrastructure where available, and our Cruz platform.

Maintainability of the environment supporting SRE is another challenge that gets overlooked. I have seen plenty of environments that fell apart once a few key people moved away from the company. What if your SRE team members never move away from the company and continue to adapt, learn, and propose enhancements on your infrastructure?

At Dorado, we believe in thinking BIG—even when you’re scaling down! Our solution, Cruz AI Vision, is the Multimodal Agentic AI Platform designed to revolutionize network operations. With Cruz AI Vision, you won’t need the monumental investments Foo made to establish a one-person NetOps team. It clears the path, making it truly possible to run your network with a team of one. But this is just the scratching the surface— Cruz AI Vision’s full potential deserves a deeper dive, which I’ll share in an upcoming article.

Until then, are you ready to see how Cruz AI Vision can transform your network operations? Learn more here and contact us to explore real-world scenarios and schedule a demo.

Let us show you how we’re making the future of network management a reality today!

10 Key Takeaways

1. The dream of a one-person network operation team is achievable with the right tools and strategy.
2. Deployment of SRE teams requires cultural and executive support and may be hard to maintain.
3. There is a wide spectrum in deploying Automation. For example, it is easier in Data Centers but far more complex in dynamic Enterprise Campus networks.
4. Dorado’s orchestration platform simplifies managing mixed-vendor networks from Campus to Data Center, with scalability and flexibility.
5. AI-powered agents can act as virtual SRE team members, learning and optimizing network operations on the job.
6. An expert is critical to oversee operations and prevent costly errors when deploying AI Agents.
7. Simply relying on Agentic AI could result in outages costing millions of dollars.
8. With the right approach, even small teams can achieve Hyperscaler-level efficiency.
9. The future of Network and Infrastructure Operations is Thinking Big and leveraging AI-driven Agents integrated with leading management and orchestration platforms such as Cruz from Dorado.
10. Dorado’s Cruz AI Vision is paving the way for true one-person network operations.