By Enterprises are under pressure to embrace agentic AI, but most are moving without a map. Costs are spiraling, standards are shifting, and security frameworks are incomplete. Yet the business stakes are too high to ignore. Companies that figure out how to deploy AI agents securely—especially close to sensitive data like healthcare records and financial systems—could gain a decisive advantage. Those that stumble risk becoming part of the 40% of projects Gartner predicts will fail by 2027.
This is where Mirantis is positioning itself. Randy Bias, VP of Strategy & Technology at Mirantis, argues that the industry has seen this movie before. The early days of cloud native were marked by similar confusion: excitement over potential, a rush to experiment, and widespread failure until the right frameworks emerged. “It feels a bit like 2010 to 2012 for cloud native,” he told me. “People are experimenting, but they haven’t figured out how to run agents in production, especially against sensitive data.”
📹 Going on record for 2026? We're recording the TFiR Prediction Series through mid-February. If you have a bold take on where AI Infrastructure, Cloud Native, or Enterprise IT is heading—we want to hear it. [Reserve your slot
The glue holding much of this experimentation together is the Model Context Protocol (MCP), introduced by Anthropic as a way for agents to communicate with tools and data sources. But MCP launched without security or authentication, a flaw that underscores how immature the ecosystem still is. Enterprises need more than experimentation—they need a way to operationalize AI without violating compliance or locking themselves into brittle technologies.
That challenge is what Mirantis is targeting with MCP AdaptiveOps, a services-led toolkit designed to provide both a secure foundation and a path forward as standards evolve.
From Cloud Native to AI-Native
Bias described AdaptiveOps as “a blueprint for deploying a secure MCP control plane, with zero-trust networking, proxies that enforce guardrails, and a framework that evolves as the ecosystem changes.” By starting with open source components and layering services on top, Mirantis is offering enterprises a way to build confidence that today’s deployments won’t become tomorrow’s dead ends.
The comparison to cloud native isn’t just rhetorical. In those early days, companies like Netflix proved that re-architecting for AWS could unlock massive advantages. Bias sees a similar inflection point: “If you can figure out AI agents and you can do it before your competitors, then you get a step ahead. But you have to move from what runs on a laptop to what you can run in production, and that leap is not trivial.”
The pace of change in the AI landscape makes that leap even harder. Bias admitted he hasn’t seen a technology evolve this quickly. “Every week there’s a new coding agent at the top of the heap,” he said. Betting on a specific technology is dangerous; betting on adaptability is safer. AdaptiveOps is built on that philosophy: components can be swapped out as new standards like the MCP Registry emerge, without forcing enterprises to start over.
Failure, Bias warned, often comes from trying to do too much too soon. He recalled how many organizations in the early cloud era attempted to lift and shift legacy applications to AWS without re-architecting, only to see projects collapse. With agentic AI, the risks are different but just as real: choosing the wrong workloads, underestimating compliance requirements, or scaling before teams have internalized what agents are actually for. “You’re adding new DNA into the organization,” Bias said. “It’s not just technology, it’s cultural.”
Building with Flexibility, Not Lock-In
AdaptiveOps engagements reflect that reality. They begin with strategy workshops to prioritize use cases, followed by a fast-track deployment of a secure MCP control plane. In parallel, Mirantis offers training and educational services to help teams adapt to agentic workflows. It’s a structured approach designed to reduce risk and accelerate learning curves, while keeping compliance at the center.
For Mirantis, the bet is that their track record in helping enterprises cross the chasm in past technology waves—OpenStack, Kubernetes, cloud native—will resonate again. “AI-native looks like an extension of cloud-native,” Bias said. Many of the same principles apply: design for scale-out, lifecycle management, hybrid architectures. What’s new is the mix of deterministic software, nondeterministic agents, and the glue in between. Enterprises need a partner that can help navigate that complexity while staying flexible enough to evolve with the standards.
Bias acknowledged that Mirantis itself is still in a learning phase, just like the rest of the ecosystem. But that is precisely why AdaptiveOps is services-led rather than product-first. The goal is not to lock customers in, but to give them confidence that whatever path the MCP ecosystem takes, they won’t be stranded. “Everybody wants maximum flexibility right now,” he said. “We’re committed to making sure our customers can do that.”
Guardrails, Culture, and Compliance
Security is not just a checkbox exercise. Bias pointed out that agentic AI introduces new categories of risk, like data leakage through unauthorized model calls or cultural misalignment in global deployments. “If you’re in Thailand, it’s against the law to say anything against the King. You need guardrails that reflect not just regulatory requirements, but cultural and corporate policy too,” he said. That means compliance frameworks must be customizable, not just borrowed wholesale from external providers.
The lesson from cloud native is that early adopters who paired experimentation with discipline were the ones who ultimately won. Today, the same pattern is emerging in agentic AI. Mirantis is betting that enterprises will need both a secure foundation and the freedom to adapt. For organizations staring at the uncertainty of MCP adoption, AdaptiveOps offers a way to hedge bets without stalling progress.
Bias admits the MCP ecosystem is still volatile, but Mirantis’ strategy is clear: stay flexible, stay open, and prioritize customer adaptability. Just as the company helped enterprises navigate the cloud-native transition, it now aims to help them cross the AI chasm. “Everybody wants maximum flexibility right now,” he said. “We’re committed to making sure our customers can do that.”
For enterprises weighing agentic AI adoption, the message is simple: secure your deployments today, but design them to evolve tomorrow. AdaptiveOps is Mirantis’ bet that the two can—and must—go hand in hand.
