Managing AI Value Using FinOps Practice Operations

When organizations adopt and manage Generative and Agentic AI, managing the FinOps practice—described in the FinOps Practice Operations Capability—also becomes more of a challenge. Organizations need to understand the value of AI spending even as the complexity of AI cost and usage data increases and the volume of AI projects grows.

This paper introduces the FinOps Practice Operations capability in the context of AI, beginning with the specific challenges AI brings.

Next we’ll present several Best Practices to managing AI value in the Practice more effectively.

Lastly, we’ll present recommendations for participating in or operating an AI Investment Council to implement these best practices, and help FinOps practices keep the focus on how AI investments are producing value for their organization.

Challenges in Practice Operations

FinOps Practice Operations is defined on its Framework page as driving a culture of accountability by running an effective FinOps team that empowers the FinOps practice through continuous implementation of FinOps strategy and processes.

Major activities revolve around building the FinOps practice (including the extended FinOps champions throughout the org), managing the practice and ensuring the collaboration and decision making that are hallmarks of FinOps are getting done around the organization.

Managing AI value adds significant additional complexity to existing FinOps practices for a few reasons:

1. High volume of AI projects going on all at once
2. Speed with which they are developed, run, and time allowed for decision making
3. Novelty of many AI services, use cases, tools, and the inexperience in accurately managing or forecasting
4. Importance and attention paid to AI by everyone in the organization, up to and including the Executive suite
5. Diverse nature of those using and building with AI, particularly outside of traditional Engineering teams and the ability to quickly reach large scale

Recommended Best Practices

There are several methods that FinOps teams should consider when incorporating AI in order to increase practice effectiveness.

Establish Ownership and Accountability

First, it’s important to decide who is responsible for AI spending and the value it creates. Should it be based on the team, the project, or the department? Owners will need to follow organizational rules and controls to make sure AI is used in a way that’s both ethical and follows company policies (which may also be rapidly evolving). Because AI brings its own risks, having clear visibility and accountability is a must.

Track and Analyze Costs

Understanding cloud costs is already challenging, and AI services further increase complexity. To keep track, it’s important to label or “tag” AI resources in the cloud. Teams also need to decide what’s most important to track:performance, cost, or both. Good data is the foundation for making informed decisions.

Work Together and Communicate

FinOps usually involves people from different parts of the business working together. With AI, new roles like data scientists and machine learning engineers become important in these discussions. Sharing information and making sure everyone understands the costs helps everyone take responsibility, not just the finance team.

Set Plan and Budget Thresholds for AI Projects

Because AI workloads can change so quickly, it’s hard to predict how much they’ll cost. Teams should regularly look at how resources are being used and be ready to adjust their budgets for times when more AI work is happening, like during model training or testing. Working closely with AI teams can help encourage innovation without letting costs get out of control. This can initially be done at a company level to insulate from runaway overall expenses and to allow for experimentation and responsible scaling.

Fund Incrementally

The less predictable or structured the AI project is, the more frequently it should be reviewed. Do not provide months of budget when forecasts are only accurate for the next few weeks. Frequent reviews of AI project spending allows experimentation with a fail-fast mentality, and allows for incremental adjustments as you work through implementation details.

Use the Right Tools

Tools can help track AI spending and spot any unusual patterns, like sudden increases in costs. Dashboards and alerts can give teams up-to-date information and help them take action quickly, instead of waiting until the end of the month to review bills. These high-speed and experimental workloads may also be appropriate for building in hard-spend caps, which should also be clearly visible to users.

Build New Skills

AI is changing fast, so FinOps teams need to keep learning about AI itself, about new tools for automation and cost tracking, and about how to communicate across the business. These new skills will help teams handle future challenges as AI continues to evolve.

Move from Reactive to Proactive Cost Management

In the past, teams might have waited until they got a bill to see how much was spent. With AI, that approach no longer works. FinOps teams now need to be proactive in helping the business innovate while also setting clear limits and making sure spending matches the company’s goals.

Explore New Ways to Measure Costs

Today, teams can monitor costs for AI resources like GPUs and even track costs down to specific models or tasks (like the cost per AI prediction). Unit Economics that measure value to the business are particularly important to compare AI investments against one another.

Target Efficiency in AI Ops and AI Platform Teams

Teams now focus on making sure AI hardware (like GPUs) is used as efficiently as possible. This might mean adjusting the size of clusters or using cheaper resources when possible, which can save a lot of money. FinOps is also being built into the processes that AI teams use, so data scientists can see the cost of their work right away.

Use AI to Improve FinOps

Some organizations are starting to use AI tools to help manage cloud costs, like AI assistants that can forecast spending or spot unusual patterns. Due to a high variance and non-determinism of results and outcomes, even with these tools, people are still needed to make sure everything stays on track.

Measure the Business Value

Measuring and quantifying the business value of AI initiatives has been called out as a major challenge by FinOps teams managing AI. This challenge is typical of a lot of new technologies or innovations. One of the most important things an AI Investment Council might do could be to develop consistent ways to tie cost to value. By defining specific outcomes or KPIs they require AI projects to address, this can drive the Unit Economics discussion higher in the organization, and discussion in this broader forum may help to identify useful business value metrics more generally.

Managing with an AI Investment Council

One of the most effective ways of implementing many of these best practices is to form, or participate in an AI investment council within your organization.

When doing any sort of large scale technology adoption that is not yet fully established with well known architectures, approval processes, KPIs, or outcomes, organizations are best served by bringing experts together from impacted parts of the organization into an ad hoc Tiger Team. This was a common practice years ago when organizations were adopting cloud, or DevOps practices, and given the importance and amount of spending in AI, it is a practice that should be adopted by organizations now. This sort of group may go by a variety of names, but we’ll use AI Investment Council for purposes of discussion here.

We’ll first describe how an AI Investment Council might be set up – the principles, personas involved and FinOps’ role – and then provide some examples of how you might operate a council for your organization.

The Council’s Purpose

• Identify and evaluate opportunities for responsible and high-impact investment in AI
• Advise on portfolio strategy, risk management, and innovation trends in AI and related fields
• Ensure alignment of AI investments with the organization’s long-term mission, ethical standards, and sustainability objectives

Scope and Guiding Principles

The scope of the AI Investment Council is to guide, evaluate, and govern decisions related to investing in artificial intelligence technologies.  While the exact remit depends on the organization strategy, the scope typically spans the following criteria:

1. Strategic – Aligned with business goals, move fast spend intentionally
2. Disciplined – Financially sound, every AI dollar has an owner
3. Responsible – Ethical and compliant, start small, prove value, and then scale
4. Future-ready – Scalable and competitive

These might be seen as the outcomes the organization wants to achieve by having a Council in place, but it should not be relied upon as a guarantee of governance or controls in and of itself. Governance and automated controls will likely increase as AI architectures and experience grows, and as all parts of the organization become comfortable with the technology, but having an AI Investment Council in place while the area is developing provides a regular opportunity to revisit investments to ensure they are consistent with the organization’s wishes.

FinOps Role in the AI Investment Council:

FinOps aids the AI Investment Council by providing financial oversight across Cloud Infrastructure, Model Training and Inference Costs, Third-Party AI Services as well as Experimentation and Prototyping budgets for AI. Other important areas of FInOps support include:

• AI ROI & Value Measurement: Business value metrics, Cost-to-value ratios, Payback periods for AI investments
• Cost Transparency & Chargeback: Clear visibility into which teams, products, or models driving costs, Showback/Chargeback models for AI Usage
• Optimization and Scalability Guidance: Attribution of shared AI platforms and foundation models
• Risk and Compliance Input: Advising on rightsizing compute for training vs. inference, Model selection tradeoffs

In a mature AI Investment Council FinOps’ seat at the table is a strategic partner, involved early (not after bills arrive), focussing on maximizing the value of AI, not minimizing AI ambition.

AI Investment Council Persona Members

When designating/appointing representatives to serve on the council, it is recommended to consider technical and financial persona leads that have seasoned tenure and real experience in architecting and managing spend for artificial intelligence and agentic technologies. The objective is to curate a group of individuals that have the capability and necessary knowledge to execute on business strategy, financial spend goals that are capable of making informed value driven decisions for AI technology. Analysts recommend that the council be chaired or provided direct guidance by a C-level or other senior executive. A leader with the profile of a FinOps Executive Technology Leader is ideal to lead this council. The council might include representatives from:

• Business / Product owners
• AI/Technology Lead
• Enterprise Architecture
• AI or Technology Platform teams
• Infrastructure Leaders where AI runs (Data Centers, Cloud, colo, etc.)
• IT Security / Risk Management
• Finance and IT Finance
• FinOps Leads
• Procurement / Contract owners

Operation of the AI Investment Council

Review is required when:

• A new AI initiative requests incremental funding
• An AI pilot seeks to scale
• AI spend exceeds predefined thresholds
• Variable-cost AI services are introduced

The AI Investment Council criteria is dynamic and should be evergreened to incorporate lessons learned from managing/scaling and approving AI investments.

Low-Cost experimentation below thresholds and aligned with organizational budgeted spend could proceed without review.

Review Cadence and Reporting Frequency of the Council

The Council should meet as needed or deemed necessary depending on the number of AI Investment requests. Many organizations report that they meet monthly. Some report reviewing dozens of AI project funding requests, and meet more frequently for briefer time.

Establish a cadence and meeting structure that works for your organization such that this is a meeting people attend regularly, do not send proxies to, and which does not easily get overrun by other urgent activities. Too frequent and the council members may not be able to dedicate time. Too infrequent and your engineering teams may idle while they wait for approvals.

Primary Goal of Reviews

• Consistent set of information (allow apples to apples comparisons)
• Common Organizational KPIs and Metric targets
• Reporting of actual Spend vs expectations
• Value signals/criteria
• Cost risks and associated anomalies
• Optimization actions/well architected initiatives
• Fund the next steps, not entire projects

Do not spend a lot of time debating detailed cost or architecture decisions in the council; focus on the value, risk and funding decisions.

At the end of each meeting your goal should be to have a short term approved spend list and allow projects to carry forward to their next milestone.

Defining Success: Value Driven AI

The Council works together to determine AI initiatives are successful when they:

• AI investments that show clear value or fast learning – what does good look like?
• Technology investments with accompanying cost visibility and predictable spend patterns
• Scaling decisions which are data driven – what reports and data should be considered

Example AI Investment Council Charter

The following section is an example charter for practitioner use.

Conclusion

AI is changing the way organizations think about cloud spending. To keep up, FinOps teams need to become more real-time, more aware of how AI works, and more automated. By working together, using the right tools, and developing new skills, organizations can make smarter decisions and ensure their AI investments pay off.

Working with AI teams, and using AI for the FinOps practice itself, are enabling FinOps teams to move more quickly away from simpler activities like reporting and toward more business partnering, building value cases, and strategic organizational enablement, which enhances the value of both the FinOps practice and the organization as a whole.

Acknowledgments

We’d like to thank the following people for their contributions to this Paper:

Tamera Armstrong

SHI International Corp.

Tammy Burnitt

UIPath

Becky Canterbury

Shutterstock

Jean Latiere

Sanofi

Aditya Nagalla

Mozn

Alex Landis

Snowflake

Rich Young

Warner Leisure Hotels

Ermanno Attardo

Trilogy