Gartner has declared 2026 an "inflection year" for enterprise AI, as organizations race to align agent deployments with measurable business outcomes. The pressure is intensifying: McKinsey projects IT infrastructure costs will grow two to three times by 2030, even as budgets stay flat — making the tech function a prime target for agentic automation.

The State of Agent Confidence

A new report from MIT Technology Review Insights, sponsored by Microsoft, surveyed 300 global technology experts and ranked 101 tasks across AI, data, and cloud workflows by respondent confidence in AI agents acting autonomously.

Key findings include:

  • Confidence is highest for measurable, repeatable tasks — generating reports, writing boilerplate code, and streamlining defined processes
  • Data workflows are the breakthrough domain, with strong trust in agents for data quality monitoring, visualization anomaly detection, real-time stream monitoring, and data profiling
  • Complex, judgment-heavy tasks show lower readiness, primarily due to insufficient business context being supplied to agentic systems
  • Tech experts believe agents meaningfully help with everyday work — reducing repetitive tasks and improving performance

Where Agents Fall Short

The core bottleneck isn't capability — it's context. The more complex the task, the more an agent depends on deep business knowledge that is still difficult to wire into agent lifecycles at the speed and quality enterprises require.

Enterprise data remains hard to wrangle, and context-generation capabilities for agents are still maturing. Human oversight continues to be a critical factor in successful deployments, particularly for high-stakes or multistep workflows.

Governance as a Trust Accelerator

"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, CVP and Chief Product Officer, Microsoft Azure Platform

The experts interviewed for the report expect confidence to accelerate as hands-on experience deepens and enterprise AI environments mature. Embedding agents within existing governance frameworks — rather than treating them as separate systems — appears to be a key unlock.

What's Next

Microsoft is pushing further into this space through Microsoft Fabric for data workflows and Microsoft 365 tooling, with executives emphasizing systems thinking and keeping humans in the loop as foundational principles for responsible agent deployment.

For technology leaders evaluating where to deploy agents first, the data is clear: start where structure is strongest, build context pipelines early, and treat governance as a feature, not a constraint.