By Guest: Randy Bias (LinkedIn)
Company: Mirantis
Show Name: To The Point
Topic: AI Infrastructure
The enterprise software landscape is about to undergo another seismic shift. While most organizations are still figuring out how to implement AI-assisted operations, industry veterans like Randy Bias are already looking past that paradigm toward fully autonomous AI infrastructure. His prediction? 2026 is the year the Model Context Protocol (MCP) breaks out as the industry standard, custom agents become obsolete, and enterprises that fail to reimagine their workflows—not just optimize them—will be left behind.
📹 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
From AI Assistance to Full Autonomy
Randy Bias, VP of Strategy & Technology at Mirantis, has spent decades at the forefront of infrastructure evolution, from virtualization to containers to AI infrastructure. His current focus centers on a critical gap in the industry: while AI-assisted coding receives massive attention, AIOps remains comparatively underdeveloped.
“The area I’ve been focused on recently is agents and operations. Some people call it AIOps,” Bias explains. “There’s been the whole MCP thing, but what I find fascinating is that the focus so far has been on AI-assisted coding and AI-assisted operations. Very little work is being done at the level it needs to be.”
This gap becomes particularly apparent in Mirantis’ work on large AI infrastructure deals. The company is grappling with fundamental questions about automating everything top to bottom—networking, storage, bare metal, virtualization, cluster management—to create what Bias calls an “AI factory” that can largely manage itself.
The MCP Protocol Becomes Ubiquitous
One of Bias’ strongest predictions for 2026 is the breakout success of MCP. While last year saw MCP gain momentum and move toward foundational governance, this year will cement it as the dominant protocol for agent communication.
“I’m convinced that this is the year that it becomes a done deal,” Bias states. “I don’t think there will be any other protocols for most agent use cases. People will just keep extending the MCP protocol.”
This standardization matters because it enables a more significant shift: the replacement of custom agents with general purpose agents equipped with domain-specific skills through MCP servers and tools.
General Purpose Agents Replace Custom Solutions
The trend toward custom-built agents for specific use cases is ending, according to Bias. Instead, general purpose agents like Claude Code, Codex, Gemini, and others can be adapted to domain-specific purposes by combining skills and tools through MCP servers.
“It seems like these general-purpose agents can be adapted for domain-specific purposes by combining skills and tools in the form of MCP servers to tackle domain-specific problems,” Bias explains. The advantage? Organizations can ride the innovation wave happening in those general-purpose agents rather than maintaining custom solutions. “I think 80% to 90% of most agentic use cases can probably be solved with those general-purpose agents.”
Recent additions to the MCP protocol specification enable this evolution. The “tasks” feature allows for long-running capabilities on MCP servers that can be checked asynchronously, combining with the ability to run custom code. Bias demonstrates this with “nightcryer,” a proof-of-concept triage agent that automatically deploys when production issues occur—triggered by events, not humans.
“A general-purpose agent can write the code with the rules and actions that create triggers for automated tasks like triage, and it can update that code on a regular basis,” Bias notes. “The MCP protocol and MCP servers can be extended—and they’ll be extended dynamically—by agents making decisions about what they think they need.”
The Enterprise Adoption Challenge
Technology capabilities alone won’t drive this transformation. Bias identifies two major bottlenecks: the standard enterprise adoption curve and a fundamental misunderstanding of how to apply AI agents.
On the first challenge, enterprises must vet and validate new technology while addressing security concerns. Bias recently completed a cybersecurity questionnaire for a major financial services company regarding an MCP server, finding the security team impressively knowledgeable about MCP and AI but understandably cautious about new attack surfaces and regulatory compliance.
The second challenge is more subtle. Early cloud adopters tried forcing inappropriate workloads onto AWS before the industry developed clear patterns for cloud-native architectures. The same thing is happening now with AI agents.
“People are going, ‘An AI agent can just come in and automate everything for me,’ and then, when they get in the trenches and figure out what that actually means, they realize AI agents aren’t the best solution for everything,” Bias observes. “They’re appropriate for what they’re good at—fuzzy logic—and you still need regular, normal software in place.”
Reimagining Workflows, Not Optimizing Them
The path forward requires more than technology adoption—it demands cultural transformation and workflow reimagination. Bias points to a top financial services firm that issued a mandate from the top: every group must work with AI agents.
“People are afraid of AI. They’re worried about it taking their jobs. They’re worried about failures,” Bias acknowledges. “But for your entire organization to level up, everyone has to use the tools—and use them seriously.”
This isn’t about superficial integration, like having ChatGPT write a document. It’s about fundamentally changing workflows. “It’s not about taking your old workflows and just jamming an agent in there. It’s about the new kinds of workflows that become possible when you have AI agents in the process,” Bias emphasizes.
Organizations must approach this through a CEO lens, evaluating specific use cases for business impact rather than pursuing automation for its own sake. The question isn’t whether AI agents can do something, but whether applying them to a specific problem creates measurable business value.
Mirantis’ Dual Focus
For 2026, Mirantis is pursuing a two-pronged strategy: delivering high-quality AI infrastructure, especially for sovereign AI clouds where capacity demand continues exceeding supply, and demonstrating practical applications of agents to real-world business problems.
“It’s about knocking down the low-hanging fruit—the things that get you somewhere—so you can use that as a proof point to build on and develop the DNA, the cultural change inside the business that you need to cross the AI chasm,” Bias explains.
The parallels to cloud adoption are clear. Just as cloud-native thinking eventually replaced lift-and-shift strategies, AI-native workflows will replace the current approach of bolting agents onto existing processes. The organizations that recognize this distinction in 2026 will be the ones that successfully navigate the transformation ahead.
