Venkat Ramakrishnan is President and Chief Operating Officer of NeuBird AI, creator of the leading Production Operations Agent.
He has decades of experience building and scaling enterprise technology companies from startup to exit.
Before joining NeuBird AI, Venkat served as Vice President and General Manager of the Cloud Native Business Unit at Pure Storage. He was also part of the early team at Portworx, which was acquired by Pure Storage.
Throughout his career, he has also held senior positions at Cisco, IBM, Fusion-io, AMD and Western Digital.
In this conversation we talk with Venkat about how the acceleration of AI is creating a structural operations crisis for businesses – and how to pace ourselves with expectations, management and solutions brought in by AI coding tools.
M.R. Rangaswami: AI coding tools are letting developers ship software faster than ever. You’ve argued that this acceleration is creating a structural operations crisis. What are most enterprises failing to see?
Venkat Ramakrishnan: The math has stopped working. Developers using Claude Code, Cursor and similar tools are shipping at multiples of what they could two years ago. But the teams responsible for keeping it all running haven’t grown. If anything, many have gotten smaller. One side of the house is producing ten times faster while the other is wrangling environments that are exponentially more complex.
AI-generated code accelerates change velocity, and more change means more incidents, dependencies and things that can go wrong. The telemetry coming out of modern infrastructure was already at the edge of what humans could process, and now it’s beyond it.
What makes this a board-level problem is what it costs. NeuBird AI’s 2026 State of Production Reliability and AI Adoption Report found that engineering teams spend roughly 40% of their time on incident management instead of building product. That’s forty cents of every engineering dollar going to firefighting instead of your roadmap, SLA exposure in customer contracts and your best engineers burning out instead of shipping what you promised.
Enterprises have a structural problem they can’t hire their way out of. The run side needs the same AI leverage the build side already has, built for prevention, not just response. Teams have the opportunity to turn reliability into a competitive advantage versus sticking to the status quo and watching build-side acceleration turn into an ever-growing backlog of incidents, longer outages and engineers who never get to do the work they were hired to do.
M.R.: As a fix, enterprises are trying to build their own agentic AI for production ops using internal teams and general-purpose LLMs. Where do you stand on the build versus buy here?
Venkat: I have a lot of respect for teams that want to build internally, but I stand on the buy side. The build instinct comes from a good place. They want control, they want to protect their data and they understand their environment better than anyone. The problem is that building an agent capable of operating on production infrastructure is far more complex than people expect.
These initiatives generally fail at the demo-to-production transition. The agent looks great in demo. Then it gets into a real production environment and starts making confident recommendations that take services down and cost hours of recovery time. The reason is that a general-purpose LLM doesn’t have the operational picture required to make decisions in a live environment. It needs real-time signals, accurate dependency mapping and guardrails calibrated to how that specific enterprise operates.
Building all of that is a 6 to 12 month investment just to reach baseline parity, then it has to be maintained as the environment evolves. We built NeuBird AI to solve the grounding problem. The reason we deliver RCA in under 2 minutes with 95%+ accuracy is because the agent operates on real-time signals across over 50 integrations, not a static snapshot, inside a zero storage, read-only architecture that internal builds spend months trying to get right before they can go anywhere near production.
For a CIO or VP of Engineering, the question should be: “Where is our engineering time creating differentiation?” Almost always the answer is the core product. Buy the platform, put your engineers back on the roadmap.
M.R.: There is a lot of anxiety that AI agents will eliminate SRE and DevOps roles. What’s your take on the topic?
Venkat: We’ve heard this argument before, and history is pretty consistent on how it plays out. Productivity gains create more value, and more value creates more jobs. Email made mail rooms obsolete and the labor economy grew significantly. No one today expects marketers to send campaign emails by hand. The same pattern is playing out with agentic AI in operations.
The more interesting question for a CIO is what happens to the capacity that gets freed up. One SRE with a production-ready AI agent can cover the work that used to require five people. That’s scaling, not headcount reduction. The business grows, the surface area of production infrastructure grows with it, and the team that used to be underwater keeping up with incidents is now doing the reliability engineering and preventive work required to improve the system over time.
We see this directly with NeuBird AI customers. The hours that used to go to reactive firefighting are shifting to work that compounds: hardening systems, improving observability and reducing the conditions that create incidents in the first place. That’s what moves SLA performance and reduces P1 exposure. It also finally evens the scales. Development teams have had AI leverage for two years. Production ops teams are just getting it now, and the ones that embrace it earliest are going to have a meaningful advantage in how much they can take on as the business scales.
M.R. Rangaswami is the Co-Founder of Sandhill.com