Frontier AI Needs a Sheriff: A Conversation with Broadcom Field CTO Gordon Barr
Frontier AI Needs a Sheriff.
That might be my favorite way to describe exactly where we are with enterprise AI right now.
While at VMware Explore 2026 in Las Vegas, I sat down with Broadcom Field CTO Gordon Barr for a new episode of The Architect’s Edge. We moved past the AI hype and dug into the operational realities: what actually happens when AI moves beyond experimentation and becomes part of the infrastructure that runs the business?
As Gordon put it, Frontier AI today can feel a lot like the Wild West. We are rapidly moving toward environments where organizations will deploy massive numbers of autonomous agents acting on behalf of people, applications, and business processes.
Once an agent has the ability to act, the technical requirements completely change. It is no longer simply about whether a model can answer a question. It becomes:
- What is this agent actually allowed to do?
- How do we control its identity and access?
- What happens when it makes a mistake?
- Who is accountable?
Gordon makes a point that I think every architect needs to hear: if security is not baked in up front, organizations are unlikely to come back later and completely re-engineer the system to add it. That is critical when AI systems are no longer just generating text, but are accessing your infrastructure and private data.
We also challenged a massive assumption I hear constantly: "We need GPUs before we can really start doing AI."
Not necessarily. There are plenty of workloads where organizations can validate ideas and run production workloads without immediately building an enormous GPU estate. Gordon shared a great healthcare example where smaller, specialized models (SLMs) operate close to private data on CPUs—delivering exactly what the workload requires without throwing the entire compute farm at the problem. The goal shouldn’t be to deploy the largest model possible; the goal is to deploy the right model for the use case.
This wasn’t just a conversation about how exciting AI is—it was a masterclass in how we actually architect it to be secure, private, and scalable.
🎥 Watch the full conversation here:
This is the first of several conversations I recorded live from VMware Explore 2026, and I’ll be releasing more through The Architect’s Edge.
I’d love to hear from the architects, engineers, and technologists here: As agentic AI becomes more autonomous, where do you think the biggest architectural challenge will be—security, identity, governance, data, cost, or something else entirely?
#TheArchitectsEdge #VMwareExplore #AgenticAI #PrivateAI #FrontierAI
This article was originally published on LinkedIn. Read the original post and join the discussion here.
