Earlier this year, Constellation Research analyst Michael Ni described agentic AI as "merely a feature." A feature. The way spell-check is a feature. You stop noticing it's there.
My first reaction was that this was premature. Then I looked at the MCP timeline. Open-sourced in November 2024. Under Linux Foundation governance by December 2025, with Amazon, Google, Microsoft, and OpenAI as platinum members. Thirteen months from release to institutional stewardship. And I started thinking Ni was diagnosing a phase transition.
Every consequential technology has a boring phase. Boring meaning: the technology stops being the subject of strategy meetings and starts being the substrate on which strategy meetings happen. Cloud computing had one. AWS launched its core services in 2006. Somewhere around 2017 or 2018, "cloud strategy" stopped being a C-suite agenda item at most large companies and became a line item. That's roughly twelve years. And those twelve years were where the actual transformation happened. The long unglamorous middle where organizations figured out what the technology actually changed about their work.
Agents appear to be entering that phase at a pace that has no real precedent.
Gartner projects 40% of enterprise applications will have built-in agents by the end of this year, up from under 5% last year. IBM's Chris Hay put it plainly in January: "We've moved past the era of single-purpose agents." Protocol convergence. Governance formation. Feature absorption. The boring phase arriving in roughly two years instead of twelve.
Cloud computing took roughly twelve years to cross from strategic initiative to assumed infrastructure. Agents are compressing that transition into two.
I keep sitting with what that compression means.
The boring phase of cloud computing was absorption time. Deloitte's David Linthicum has noted that much of early cloud adoption happened when employees used their own credit cards to sign up for SaaS systems, without IT's knowledge or approval. That's what absorption looks like in practice: people stumbling into the technology, getting it wrong in small ways, building intuition about what it changed before anyone wrote a policy document. Those twelve years were the period during which organizations learned, often by accident, what mattered.
So what happens when you compress that window from twelve years to two?
Maybe organizations have genuinely gotten better at integrating new infrastructure. They've been through cloud, through mobile, through SaaS proliferation. The muscle memory exists. But then you read that EY's Raj Sharma told Fortune most enterprises can tell you how many human users have access to their financial systems, and few can tell you how many AI agents do. Only 21% of organizations report having mature governance for AI agents, even as three-quarters plan agentic deployments. The technology is moving into production faster than the organizational thinking that makes production meaningful.
When boring arrives on schedule, it usually means the hard questions have been mostly answered and the remaining work is execution. When boring arrives ahead of schedule, the reading gets harder. Organizations might be absorbing the form of infrastructure, the protocols and the integrations, without doing the slower work of understanding what actually changes about how they operate.
The interesting thing about this particular boring is that nobody seems sure whether it was earned.
-
The web's second audience: Chrome 146's early preview of WebMCP quietly introduced a structured interface that lets AI agents interact with websites through tool calls rather than screenshots, which may be the clearest sign yet that "boring" means the web itself is reorganizing for machines.
-
Governance as rate limiter: ModelOp's 2026 AI Governance Benchmark Report found that commercial governance platform adoption surged from 14% to nearly 50% in one year, yet enterprise AI value delivery still lags behind ambition.
-
The 70/20/10 investment split: BCG research suggests winning organizations allocate 70% of AI investment to people and processes, not algorithms or technology, and those who invert that ratio by spending heavily on tools see flat adoption.
-
Workforce reorganization, already underway: An Al Jazeera report found that AI agents are not just completing tasks but assigning them to human workers and managing each other, with some engineers now spending 20–30 hours a week interacting with agents.