Gartner is calling 2026 an “inflection year” for enterprise AI agents, and McKinsey has identified the structural cost pressure that may force the shift. MIT Technology Review published a synthesis of enterprise adoption signals on Sunday that frames agentic AI as operationally urgent rather than speculative.
The Cost Forcing Function
McKinsey projects IT infrastructure costs will grow two to three times by 2030 while budgets remain flat, according to the MIT Technology Review report. That gap creates an environment where AI agents in technology operations become a cost-reduction necessity. The math is simple: if infrastructure spending triples and headcount budgets stay constant, something has to absorb the difference.
Gartner’s framing reinforces the timeline. The firm forecasts global AI spending will grow 47% in 2026 and positions this year as the moment organizations must connect their AI projects to strategic business objectives or risk falling behind.
Where Confidence Is Highest
The MIT Technology Review report is based on a survey of 300 global technology experts who ranked 101 agent task categories by confidence in agents acting on their behalf. The findings show two clear patterns.
First, confidence is highest for structured, measurable tasks: generating reports, writing boilerplate code, data quality monitoring, visualization anomaly detection, real-time data stream monitoring, and data profiling. These are areas where output quality is easy to verify and failure modes are well understood.
Second, confidence drops where agents need business context. Complex tasks requiring multi-step reasoning and decision-making scored lower, not because the models lack capability, but because enterprise data remains difficult to connect into agent workflows at the speed and quality developers need.
“As we design agents to operate within the same operational boundaries, identity systems, and governance models that teams already use, they start to behave more like the systems organizations already trust,” Jeremy Winter, corporate vice president and chief product officer at Microsoft Azure Platform, told MIT Technology Review.
Data as the Bottleneck
The report identifies data workflows as the breakthrough domain for agent adoption. Technology teams trust agents most where data structure provides a reliable foundation for decisions. This tracks with recent enterprise deployments: New Relic’s Autopilot handles incident triage and root cause analysis, Servicely’s multi-agent orchestration manages service automation, and HP’s OpenAI Frontier integration scaled after one engineer processed 122 pull requests across 43 projects.
The common thread across these deployments is that agents succeed where the data pipeline is clean and the task boundaries are defined. The gap between current agent capabilities and the remaining 2-3x cost expansion McKinsey projects is an integration problem: getting enterprise data wrangled, connected, and available to agent systems at production quality. Model capability is not the bottleneck.