How does FinOps enable AI, and what challenges do organizations face in optimizing cloud spending?
In this interview, Anne Johnston and Jerzy Grzywinski from Capital One share their insights on their journey! They cover key challenges and the future of FinOps in an AI-driven world.
As AI adoption accelerates, FinOps plays a crucial role in ensuring efficiency, cost transparency, and value-driven investments. At the upcoming Data Innovation Summit 2025, Anne and Jerzy will take the stage to explore how organizations can maximize their cloud investments while driving innovation. Read more to discover their key strategies and insights.
Hyperight: Can you tell us more about yourselves and your organization? What are your professional backgrounds and current working focus?
Anne Johnston: Currently, I lead the Cloud Evolution organization accountable for increasing the effectiveness of Capital One operations in the cloud; with FinOps as one of the pillars. I’ve been at Capital One for almost 10 years and have worked across our technology organization in several roles enabling customer experiences and securing infrastructure. Before Capital One, I worked in consulting with Accenture. I started in the technology industry during the dot com era with AOL.
Jerzy Grzywinski: I head up the FinOps pillar at Capital One. In my role, I lead a team responsible for architecture, standards, and tooling to maximize the efficiency of operating in the cloud. Before this role, I had the pleasure to work in different engineering roles from full-stack development to DevOps, to Architecture. I’ve been with Capital One for 16 years where I have had the opportunity to tackle different engineering challenges and different Capital One products.
Hyperight: At the upcoming Data Innovation Summit 2025, you will present on FinOps for enabling AI. What can the delegates expect from your presentation?
Anne and Jerzy: AI is the epicenter of innovation that is sweeping the industry. We are seeing FinOps and AI intersect across three themes:
- Success in FinOps is driving savings that can be reinvested in high value initiatives, such as AI
- FinOps is optimizing efficiency for how AI solutions are architected, implemented, and maintained
- Leveraging AI in our FinOps tooling is delivering higher impact for the organization
Our presentation will focus on the first two themes. We will also share how Capital One maximizes value for its AI investments.
Hyperight: Can you tell us about Capital One’s journey with FinOps and how it’s evolved over time?
Anne and Jerzy: We started our journey to the cloud in 2016, with our last data center closing in 2020. When it comes to FinOps, we took a “Crawl, Walk, Run” approach as we scaled our cloud adoption:
- Crawl: For several years, our focus was on migrating to the cloud. Many teams were new to using the cloud and the emphasis was on accelerating adoption. In this phase, FinOps was limited to central finance managing our cloud bill and invoicing.
- Walk: As our cloud footprint grew, it was important to create greater transparency. Transparency of where costs were coming from, to make the most of our cloud investments. This is when we started expanding our pre-purchasing strategy and outlining optimization opportunities for the company. We also began allocating our tech investment to implementing a solution that identifies spending and chargeback transparency.
- Run: This phase uncovered how critical it was to generate greater awareness and efficiency in the cloud. Especially as our cloud spend continued to grow. To do this, we increased investment into tooling and optimization initiatives across the organization. With increased scale, we invested in automation to replace manual touchpoints. We also expanded our focus moving up the stack to seek efficiencies at the software layer and across SaaS providers.
Now, we continue to invest in FinOps, while broadening our scope to SaaS and GreenOps as well.
Hyperight: What are some key challenges organizations face when managing cloud data costs? How does Capital One approach these challenges with its FinOps strategy?
Anne and Jerzy: One challenge organizations often face is scale. At Capital One, we have 14,000+ tech associates working on over 2,000 applications. Driving best practices at this scale requires us to be creative in how we execute change.
The goal is to fully automate each best practice, ensuring it’s clear and intuitive for developers without needing additional training or influence from local teams. Unfortunately, not every best practice can be fully automated. To address this, we invest in features and culture initiatives. These initiatives will help us ensure reach across all application teams. This minimizes developer effort and increases adoption of good engineering practices that in turn, drive high efficiency.
Hyperight: Anne, your team leads the charge in promoting a culture of cost awareness. How do you foster this mindset across the company, and what role does leadership play in driving it?
Anne Johnston: End-to-end software application ownership is so much more than developing the business logic that creates a customer experience. Something like cost awareness can be seen as just another task, another thing to do. But, it doesn’t have to be just another ‘add on’ that an engineer needs to remember to do.
At Capital One, we embed cost awareness into each part of the lifecycle so it’s in line with the lifecycle. Whether it’s during design, when creating an initial cost estimate, or during development when teams receive recommendations, various milestones trigger our engagement model in a helpful, non-binding way.
Leadership plays a strong role in not just awareness but also accountability. Our leadership team is aligned on prioritizing effective spending, which paves the way for us to remediate wasteful spending.
Hyperight: Jerzy, you emphasize the importance of “Good engineering is efficient engineering.” How does this philosophy contribute to Capital One’s cloud cost management and operational efficiency?
Jerzy Grzywinski: Capital allocation is finite. In order to deliver the highest amount of engineering value, using the capital effectively needs to be a key pillar. For startups, finite capital is felt by every employee, which translates to a culture of frugality and creativity. As the size of the company grows, the distance in relationship between capital allocation and engineers can grow apart. Our FinOps goal at Capital One is to bring it closer and make it more inline with the startup mentality.
When efficient engineering is part of the engineering culture, every day decisions always end up considering cost. Decisions and solutions can be tweaked in real time with little effort, often eliminating the need for correction at a later time which comes at a higher cost. That is why we believe that building a culture of efficiency is a key aspect to cloud cost management.
Hyperight: Anne, as part of your team’s work to enable genAI and software partners, how do you ensure these emerging technologies are supported with cost-efficient cloud resources?
Anne Johnston: When exploring genAI or other emerging technologies, we collaborate with finance and business teams to identify strategic areas for technology investment and pinpoint opportunities to reduce costs. While not always, those reductions are oftentimes what is funding much of the emerging technology investment. There’s usually an ebb and flow to where the technology budget is focused. Once we understand the investment areas, it’s coming to a joint understanding about what good and wasteful spending looks like. Based on business priorities, growth projections, or other factors, good and wasteful spending can look different across projects. We can then put cost guardrails in place so teams are unable to exceed thresholds by accident – these can include reactive alerting, such as detecting anomalies, and preventative measures like restricting GPU provisioning.
Hyperight: Jerzy, Capital One serves thousands of tech teams. How does the FinOps team ensure consistent cost management practices are followed across such a large and diverse group?
Jerzy Grzywinski: We want to make sure local teams have the flexibility to drive the action with the highest ROI. To ensure consistency, we measure efficiency across our largest resources through a metric we call ‘Spend Efficiency (SE).’ The SE score starts at a resource level that is then rolled up to an application, a division, and ultimately the enterprise. Creating a single metric that can be used with any developer or executive helps clarify the outcome we are working towards, while giving local teams flexibility in how to achieve it.
To support developers with consistency, we have consolidated our tooling and education into a single system. This is the place where anybody at Capital One can go and look at relevant information and recommendations for their applications. Similar to the SE, the tool provides an executive view and then drills down to a division, an application, and a resource. At each altitude, the tool highlights key metrics, highROI campaigns, and supporting data/insights to power the change.
Hyperight: What opportunities have you identified to reduce cloud costs without compromising on the value of data? How do these opportunities influence funding for emerging technologies?
Anne and Jerzy: We have invested in improving our data management efficiency by pursuing two actions: decrease the unit cost and the volume of data.
There are many services available for storing data, ranging from in memory caches, all the way to cold storage. To drive down the unit cost, we wanted to get smarter at using automation to leverage the lowest cost storage tier that meets the need of the application. Using tools such as AWS S3 Intelligent Tiering, automation and cloud native solutions allowed us to leverage access patterns to determine the lowest cost tier for storing data.
To manage volume, we focused on reducing duplication. Finding data across different resource types was our largest challenge, and once identified, it was trivial to delete.
Between these two efforts, we saw significant savings. This allowed us to fuel other investments, without compromising on what data we store and for how long we keep it.
Hyperight: How do you see FinOps evolving over the next few years? Especially as AI and other data-driven technologies continue to grow.
Anne and Jerzy: As the cloud and AI continue to scale, FinOps too will continue to grow in importance. Federating access to resources is increasing in both the cloud and SaaS offerings as there is a strong craving for all technology costs to be well understood and efficient. FinOps will be the vehicle that delivers the observability and, most importantly, the solutions that help maximize value through efficiency.
We are also seeing GreenOps being merged into the FinOps practice. With the exponential need for power to fuel new technologies, understanding power usage and efficiency will fuel a greater focus on FinOps.
Don’t miss Anne and Jerzy’s presentation at the Data Innovation Summit 2025! They’ll explore how FinOps is driving AI adoption and cloud cost optimization at Capital One. Learn key strategies to balance efficiency with innovation and discover how FinOps helps reinvest savings into high-value AI initiatives. If you want to understand how FinOps can transform AI strategies and streamline cloud spending, this session is a must-attend!
Add comment