The recommendation makes financial sense based on the information available to the agent, but executing it without the right context could introduce a much larger operational risk.

As AI agents become capable of making and executing cloud cost decisions, situations like this raise an important question for technology and finance leaders: who determines what an agent is allowed to do?

Forrester sees a shift in how FinOps operates

At FinOps X 2026, Forrester observed cloud providers and FinOps vendors moving toward greater use of autonomous AI agents.

In its June analysis of the conference, Forrester described a progression from reactive cost management toward systems that can investigate spending, recommend corrective measures, and increasingly execute optimization decisions.

AWS, Microsoft, Google Cloud, IBM Cloudability, and Flexera were among the vendors Forrester identified as advancing AI-driven financial intelligence and automation.

The direction is also reflected in Forrester’s Q2 2026 IT Financial Management research, which highlights growing emphasis on decision support, contextual recommendations, and AI-assisted financial analysis.

For enterprise FinOps teams, this could change how routine investigations and cost management activities are handled. It also places greater importance on the financial and operational context available to AI agents.

A cost-saving recommendation needs business context

Cloud consumption data can tell an agent which resources are running, how much they cost, and whether they appear underutilized. However, the information required to make a sound financial decision often extends beyond the cloud account.

An environment may support a critical application, carry contractual commitments, or serve a business function whose value is difficult to infer from utilization alone.

Even a straightforward cost adjustment can have consequences for budgets, forecasts, cost allocation, and service delivery.

This is where the connection between FinOps, IT financial management, and business ownership becomes important.

An effective agent-assisted investigation should be able to examine spending alongside the financial model, relevant contracts, application dependencies, and the teams responsible for those costs.

Without that context, an agent may correctly identify an expense while recommending an action that is unsuitable for the business.

Who should approve an AI agent’s financial decisions?

The answer depends on the consequences of the action.

An agent identifying unusual spending or preparing an analysis presents relatively limited risk. Changing infrastructure capacity, modifying allocation rules, or committing expenditure requires a different level of oversight.

Before allowing agents to execute financial or operational changes, organizations need to establish which actions are permitted, what evidence is required, and when approval must come from an accountable person.

For CIOs and FinOps leaders, the key considerations include:

These controls help organizations introduce automation without compromising financial accountability or service reliability.

How Yarken approaches agentic technology finance

As agentic FinOps develops, the quality of financial decisions will increasingly depend on how well automation understands the business it operates within.

For organizations evaluating these capabilities, the immediate priority is to establish where agents can make reliable recommendations, which decisions require additional context, and where human approval remains necessary.