Enterprise AI is moving from experimentation toward measurable business results. A report produced by MIT Technology Review Insights in partnership with Microsoft shows that technology teams are already using AI agents across practical work in AI, data, and cloud operations, while still drawing clear lines around trust, context, and oversight.
Why tech teams are a natural test bed
The pressure around enterprise AI is changing. Gartner is calling 2026 an “inflection year” for organizations to align AI projects with strategic business objectives, and executives are looking for evidence that AI can deliver financial outcomes.
The technology function is a direct place to look for that evidence. According to McKinsey, IT infrastructure costs are projected to grow two to three times by 2030, while budgets remain unchanged. That creates a difficult operating problem for the engineers, developers, architects, and other practitioners responsible for building, deploying, and improving infrastructure and applications.
Over the last 18 months, those teams have been putting agents to work. The report frames this as more than simple automation. The larger promise is that agents can help manage and coordinate workflows, while humans and agents work together toward business goals.
That distinction matters. Automating an isolated task is different from letting an agent act across a workflow. The more influence an agent has, the more confidence teams need in its ability to operate safely, reliably, and securely.
Where confidence is strongest
The report is based on a survey of 300 global technology experts. It ranks 101 tasks across AI, data, and cloud workflows according to how confident respondents are in agents acting on their behalf.
The strongest confidence appears in tasks where the work is clear, measurable, and easier to evaluate. Technology experts said agents help with everyday work by streamlining processes, improving performance, and reducing repetitive tasks.
Examples called out in the report include generating reports and boilerplate code. These are not trivial tasks, but they are easier to judge than open-ended decisions because the output can be checked against expectations.
The report also points to rising opportunity in multistep workflows and tasks that require advanced reasoning to make decisions. That is where agentic AI begins to move beyond routine support and toward more meaningful participation in technical work.
Data workflows are emerging as the clearest opening
Among the domains covered, data workflows stand out. The report describes data as the breakthrough domain for AI agents because structure can provide a reliable foundation for decisions.
Technology teams showed trust in agents across areas such as:
- data quality monitoring
- visualization anomaly detection
- real-time data stream monitoring
- data profiling
These examples share an important feature: they are grounded in data patterns and operational signals. That does not remove the need for expert judgment, but it gives agents a more stable base for action.
The report also emphasizes the role of domain experts closest to the point of data generation. Those experts can provide context that helps agents act in ways that produce trusted outcomes.
This is a practical point for enterprise AI adoption. Trust is not only about the model or the tool. It also depends on whether the agent has access to the right context at the right point in the workflow.
The main gap is business context
Agent readiness drops when systems lack business context. The report states that more complex tasks require more reasoning capability and a greater need for business context.
That is a central limit for agentic AI in enterprise settings. Technical systems often need to understand why a task matters, what constraints apply, and what outcome the business actually wants. Without that context, an agent may be technically capable but still poorly positioned to act on behalf of a team.
The report describes context-generation capabilities for agents as still early in development. This is especially true when enterprise data is difficult to wrangle and connect into the agent lifecycle at the speed and quality that developers and executives need.
In other words, the barrier is not only whether an agent can reason. It is whether the organization can supply the agent with connected intelligence that reflects its workflows, data, governance, and priorities.
Human oversight remains central
The report is clear that human oversight is a key factor in successful agentic AI deployment. That is especially important because agents may participate in automated decision-making, where mistakes can carry real operational risk.
Jeremy Winter, corporate vice president and chief product officer at Microsoft Azure Platform, connects trust to the way agents are designed inside existing enterprise systems:
“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,” says Jeremy Winter, corporate vice president and chief product officer at Microsoft Azure Platform.
That view suggests a practical direction for adoption. Agents gain credibility when they fit into known operating models rather than sitting outside them. Identity systems, governance models, and operational boundaries are not side issues; they are part of what makes agentic AI usable in serious technical environments.
The report’s broader conclusion is that technology teams are in a pivotal position to lead this shift. The experts interviewed expect agent confidence to accelerate as teams gain more experience and as business environments mature.
For now, the pattern is clear. AI agents are earning confidence first where work is measurable, structured, and close to the systems that technology teams already understand. The next stage depends on whether organizations can connect agents to richer business context while keeping humans meaningfully in the loop.