Guest: Randy Bias (LinkedIn)
Company: Mirantis
Show Name: To The Point
Topic: AI Infrastructure

AI agents are generating enormous excitement, but Randy Bias, VP of Strategy & Technology at Mirantis, sees a familiar pattern playing out in how enterprises are actually responding — and it mirrors the early days of cloud.

The enthusiasm is real. So is the hesitation.

When Mirantis recently provided an MCP server to a major financial services company, the security questionnaire that followed was extensive. But Bias found himself impressed rather than frustrated. The cybersecurity team understood MCP. They understood the risks. They asked precise, informed questions about the new attack surface they were inheriting. This wasn’t obstruction — it was exactly the kind of rigorous vetting that responsible enterprise adoption requires.

“The cybersecurity team really understood MCP. They understood the issues, they understood AI — they asked very good questions,” Bias noted. “But you could tell they were going, ‘Okay, this is a whole new attack surface. How do we take care of it? How do we make sure we’re still compliant with all of our regulatory requirements?”

That dynamic — smart people applying hard scrutiny to new technology — is one of two major bottlenecks Bias identifies as slowing AI agent adoption in the enterprise. The other is more subtle.

There is a widespread tendency, he argues, to assume that AI agents can simply walk in and automate everything. It is the same mistake that plagued early cloud adoption, when organizations tried to migrate workloads to AWS that simply did not belong there. The result was the same then as it threatens to be now: friction, disappointment, and retrofitted solutions that create more problems than they solve.

“People are going, ‘an AI agent can just come in and automate everything for me,’” Bias explained. “And then when they start to get in the trenches and figure out what that means, they realize AI agents aren’t the best solution for everything. They’re appropriate for what they’re good at, which is fuzzy logic. And you still have to have regular software in place.”

The fix, in his view, isn’t a technical one — it’s strategic. Organizations need to approach AI agent deployment the way a startup CEO approaches resource allocation: with ruthless clarity about where the investment will actually generate business impact. That means going into the business, auditing the existing portfolio of workflows, and asking a direct question — does applying an AI agent here solve a real problem and create measurable value?

“It’s not automation for automation’s sake,” Bias said. “It’s going in and looking at the business, doing a triage of your own portfolio, and saying, ‘This is a place where we can use agents for direct business impact.’ It’s about really thinking through the business objectives.”

The underlying message is both a caution and a compass. The enterprises moving carefully — validating security posture, stress-testing use cases, demanding regulatory clarity — are not falling behind. They are building the foundation for AI adoption that actually lasts. And the organizations that skip that work in favor of speed may find themselves repeating the painful lessons of early cloud migrations.

For Bias and Mirantis, MCP is already at the center of enterprise AI infrastructure conversations. The technology is real, the demand is real, and the skepticism is productive. The bottleneck, ultimately, is a feature — not a flaw.