A shared discipline for control, value, and optimization

AI is moving quickly from experimentation to scaled use across people, teams, and services. As adoption grows, leaders need better answers to practical questions: What are we spending? Which services and scenarios are driving usage? What value are we getting back? And where should we invest next?

FinOps for AI is a shared discipline for managing AI spend and optimizing it for business value. Success requires active participation across the organization, from finance and IT to business leaders, developers, makers, and end users. And as we continue to launch new innovations through usage-based billing, organizations need a consistent way to understand where AI investments are going, connect spending to outcomes, and make informed decisions about where to control, sustain, or expand usage.

The foundation in place today

We are committed to helping organizations stay in control of their AI spend and understand the value they are getting in return. That means giving leaders and admins the tools to manage usage, set clear spending guardrails, track costs, and connect investment to outcomes as AI adoption scales. With the general availability of Copilot Cowork and Work IQ APIs back in June, cost management in Agent 365 has become the foundation for admins to manage usage-based billed services with centralized visibility:

• Usage-based billed services are off by default: Decide who gets access to usage-based workloads by creating scoped spending policies with spending limits.
• Spending controls at tenant, group, and user levels: Create tenant-wide and group-level billing policies, including user-level caps within group policies, and configure alerts when spend reaches specified thresholds.
• Departmental billing support: Manage departmental costs by associating spending policies to the right Azure subscriptions and resource groups, enabling cost accountability via the same Azure billing constructs organizations use today.
• Detailed usage reporting: See usage broken down by user, group, workloads, policies, and payment methods with clear accountability across the organization.

Cost management in Agent 365 is included with your Microsoft cloud subscription.

New capabilities available and coming soon

Today, we are introducing many new FinOps for AI capabilities within Agent 365, Insights, and Microsoft Copilot. Together, these capabilities expand the scope of services, while comprehensively offering new ways to understand and manage usage, connect cost to value, and make informed decisions with confidence.

• Support for the new Copilot (and more): Expanding beyond Cowork and Work IQ APIs, cost management in Agent 365 now supports Code and Copilot Managed Runtime and, beginning in October, will support agents built with Microsoft Copilot Studio. We will continue to add additional AI workloads and make cost management in Agent 365 the central home to manage usage-based AI workloads.

• Manage model choice: Tailor model family access for different groups based on their needs and business objectives with cost management in Agent 365. This gives admins the ability to match the right models to the right user or scenario to maximize value generated. This also impacts which models Auto can select in Cowork. We are rolling out this capability for Cowork with plans to expand to additional workloads soon.

• Measure the value of Cowork: Get a clearer view of how Cowork helps your team execute meaningful work—not just how often they use it with the consumption dashboard in Insights. By connecting credit consumption to value generated, IT and business leaders can identify high-value scenarios, compare results with the resources invested, and make better decisions about where to sustain, expand, or optimize AI usage. This feature is available both in the cost management in Agent 365 and the consumption dashboard in Insights. For more details, check out: Measuring the value of Cowork.

• Cowork end-user spend management: View your credit usage, remaining credits, and usage history directly within Microsoft Copilot. With visibility into their consumption and notifications as they approach limits, end users can better self-manage AI usage, prioritize high-value activities, and stay within organizational spending policies.

• Programmatically manage policies: Use Microsoft Graph API to manage spending policies at scale in your own business intelligence and FinOps tools of choice. This helps organizations automate with the systems and processes they already work in.

• Redirect credit requests to your existing approval flows: Route credit requests from end users to custom URLs. This enables admins to flow credit requests via the existing automation and approval flows that your organization already runs on.

We will continue to invest in FinOps for AI and expand the tools organizations need to stay in control of spend, understand value, and optimize for impact. We look forward to sharing more as we bring new capabilities and experiences to our products.

Join the community!

As you use these controls, share with us what you are seeing and what you would like us to cover next. We will continue to share practical guidance on controls, visibility, value measurement, model choice, and the operating habits that help organizations tie their AI investments to real business outcomes. Join now!