Inside Modern FinOps: Jean Latiere on Agentic AI, Cloud Waste, and Environmental Impact

Chapter 1 — Intro: Meet Jean Latiere

Kevin: Hello and welcome to Unclouded, the Agentic AI FinOps podcast by Cloudgov.ai. Today I am joined by Jean Latiere, cloud financial management specialist at OptimNow, where he helps enterprises cut cloud waste, govern multicloud budgets, and align cost with value across AWS, Azure, and GCP. We’re going to dive into the next evolution of cloud financial management. One where Agentic AI and sustainability go hand in hand, because FinOps is not just a function, it’s a culture. Jean, welcome. Great to have you.

Jean: Thank you. Thanks for the invitation.

Kevin: Absolutely. So, a lot of companies focus on short term savings, one time optimizations, tool rollouts, or cost cutting exercises, but they rarely last. How do you help organizations build a true FinOps culture that sustains itself?

Chapter 2 — Why One Time Optimizations Never Last

Jean: Yeah. So that’s a tricky part. And that’s a tricky part in FinOps: what is the trade off between bringing some one offs and some quick wins, and building that culture, that culture of FinOps in the organizations, that will be lasting and that delivers sustainable value. Because what we see generally when we come in as FinOps experts, generally there has been something like an optimization trap.

Chapter 3 — The “Optimization Trap” and How Enterprises Get Stuck

Jean: The optimization trap is generally they’ve been optimizing with the commitments, with the savings plans and the reservations. And with savings plans and reservations, you can make quick savings, like good discounts against some commitments. And yeah, you can please the management with it. You will have less on demand. You have more discount. You can really show some savings on the compute. But actually what you’ve not done is you forgot to look at all your architecture and all the waste in your architecture.

And yeah, there are surveys that say that there is about 30 to 40% of waste in the cloud today. So resources which are basically unused, and that’s across all the companies in the world and across all cloud providers. So when you start with savings plans and reservations and commitments, you got some discounts. But actually you got some discounts on a good part of resources that you’re not using, and you’ve committed on using those resources. So yeah, that’s a bit of a trap that you fell in, because you need some quick results. Someone came like a manager and said okay, I need to show that we are optimizing, I need to show some results, and they went for it. And yeah, after that, what do you do? Exactly, what do you do?

And that’s actually where you need to start optimizing. You need to bring that culture of FinOps across the organization, have people understand what is cloud financial management, what are the concepts behind it. And that requires a bit of work more than setting up a FinOps theater with some dashboards and some good news. It’s more like a ground work that you need to do.

Chapter 4 — Why FinOps Success Is Cultural, Not Technical

Kevin: In your experience, is it difficult to get engineering teams and finance teams to kind of work together on this problem?

Jean: So yeah, they are not coming from the same world. They are definitely not coming from the same world. They are not talking in the same language, and they don’t understand each other even when they speak. That’s quite funny.

Chapter 5 — Creating Shared Language Between Engineering and Finance

Jean: When you talk with engineers about money, generally there is sometimes a gap. There are things that they won’t get about financial concepts like depreciation, amortizations and this kind of things, or contractual stuff about clauses in your contracts.

And same with finance, they tend to forget that there is an underlying technological infrastructure behind everything that they are caring for. So when you talk to them about Kubernetes, or when you talk to them about the compute VMs or stuff, they start getting a bit stressed and not feeling in control.

So you really need to bridge the gap between those two business disciplines. You need a bit of diplomacy so that they understand each other, the issues and challenges. And you need to explain and you need to educate on each other’s challenges and issues. So you really need to position yourself as the bridge to bridge the gap between those two disciplines. And also procurement, which is a third stakeholder who also has their issues and challenges as well.

Kevin: Is that something that you have to try to do, to bridge that gap and bring all these entities together, so they can all have the same sort of end goal of really figuring out how to optimize and control or reduce the waste?

Chapter 6 — Cost Allocation Without Friction

Jean: So yeah, they need to understand a bit each other’s works. Finance needs to understand a bit the changes and what is a VM. What is an OpEx architecture? What is the storage? What are the challenges of the storage, for example in a regulated environment in healthcare or in finance, what you need to do with your storage in those environments. And same with engineering. You need to bring a bit of financial concepts to them. Like for example, what are amortized costs versus unblended costs, so that you can really measure the efficiency of how you are using your cloud resources.

So you need to define that common vocabulary for all the stakeholders. FinOps stakeholders need to agree on some common KPIs that everybody understands, and how you track them and how you measure your performance against.

Kevin: So obviously we’ve been talking money. But it’s not just the cloud, it isn’t just about dollars, it’s about carbon. How big is the environmental impact of cloud computing and data centers, especially now with Gen AI workloads consuming massive energy?

Chapter 7 — The True Environmental Impact of Cloud and Gen AI

Jean: Yeah. So it was already big and it’s getting bigger. And we know it with the Gen AI and the new wave of all those AI data centers that are being under construction or already delivered. AI data centers need four or five times more energy than traditional cloud data centers as we know them. I don’t exactly have the figure in mind, but I think the IT industry was responsible for about, I would need to check the numbers, but it was about 7 to 8% of the total carbon footprint. Inside that you have the cloud. Well, the cloud won’t make the full of those 7 or 8%.

Chapter 8 — Why AI Workloads Strain Power, Cooling, and Water Resources

Jean: But it is growing and growing and will take about 20% of it. So it’s not so huge, but now it’s really really growing. And it’s not so much the carbon footprint, it’s also water usage, especially in places that are a bit stressed with water. It’s also all the methods, all the resources that you need to build those data centers. So yeah, it’s a very complex supply chain all across the world that consumes a lot of energy and resources. So yeah, it’s significant.

Kevin: Do you think that sustainability should be treated as a FinOps metric like cost or performance?

Chapter 9 — Should Sustainability Metrics Be Part of FinOps?

Jean: Yeah, definitely. You should have it along the FinOps metrics. You should be monitoring the carbon footprint. You should monitor the water utilization of your cloud and your data centers. Actually, the cloud providers like AWS, Azure, GCP provide some dashboards so you can do it. They are starting to have very good methodologies, and you can start trusting the numbers that they share, really. Everything is included, the carbon footprint that they share. So you should have them, and you should monitor the optimization that you achieve against the environmental impact, also the financial impact.

Chapter 10 — Where Cost Savings and Carbon Savings Overlap

Jean: And sometimes very often they are correlated. What you are going to save in money in optimization, it’s going to correlate and you’re going to save in terms of environmental impact. But not always. Sometimes for example, it’s about switching to another region which power is less carbon intense.

Chapter 11 — Region Choice, Carbon Intensity, and Real World Tradeoffs

Jean: For example, France is using a lot of nuclear energy. So the carbon footprint of our power is a lot less carbon intensive for example than Germany, which is using coal during the night and renewable energy during the day. Or Sweden is really good with hydro electric, so very low carbon footprint.

So sometimes it’s also about picking a region. And for example, there are new AI data centers in Sweden that are quite interesting in terms of carbon footprint, also in terms of cost. Because if they need less money to manage the temperature of the data centers, okay, to manage the heat of the data centers to keep it cold, it means that they are spending less. So you’re going to spend less for the same compute. So yeah, generally it’s really correlated. But still you can make some choices and some tradeoffs that will have an impact on the carbon footprint without having an impact on your cost, and the other way.

Kevin: Could Agentic AI eventually optimize for both cost and carbon impact simultaneously?

Chapter 12 — How Agentic AI Optimizes More Than Cost

Jean: Yeah, definitely. I mean, it’s only a matter of defining the goal of the agent. What is the agent supposed to optimize for? Is it supposed to optimize for us? Is it supposed to optimize for carbon footprint or water utilization? Or is it supposed to optimize for the latency and for user experience, for example? And then it doesn’t care about the cost and the carbon footprint. What’s most important is the performance, the latency. So basically the agents will do basically what you tell them to do.

Kevin: Yeah. What do you like about Cloudgov.ai’s Agentic AI vision and platform to help solve this problem?

Chapter 13 — AI Agents as Enforcers of Tagging, Cleanup, and Governance

Jean: Yeah. So I saw them already starting to put some agents, for example, regarding the start and stop of the instances, and to be able to autonomously, with some controls and some guardrails, to start and stop some instances, especially in dev and in quality. Probably you don’t want to let an agent do it in prod, but you could do it, but with strong guardrails. So I see that they are already integrating it, and building the Agentic platform that allows you to put that in motion.

Kevin: You’ve been exploring this intersection of FinOps and AI for years. How is AI changing the way companies approach cost optimization, and what do you think Agentic AI means in that context?

Chapter 14 — Balancing Automation with Human Oversight

Jean: So yeah, I believe the impact of agents is going to change quite a lot the way that FinOps is achieved. So it’s going to be a lot less about dashboards and reports, recommendations and then remediations all powered by human and with steps of validations, but much more about setting up some agents to themselves screen and check the environment and what is going on in the cloud platform themselves. Scan the resources, how they are used, are they used enough or not enough for example, what is it, is it right sized or not?

Chapter 15 — Forecasting, Planning, and AI “Clones” of Cloud Environments

Jean: Is it the right kind of disks which is supposed to be running on the VM? Let’s say for example, there’s a classic one on AWS: you should move your GP2 disk to GP3 because it has better performance for better cost. There is no business case to keep a GP2. Sometimes they keep it, but because they forget. So the agent should be able to scan all this and to take some actions that we define in guardrails, what the actions it can autonomously do.

So first, things like having some agents do some work of autonomously optimizing. But also it’s going to help us make some estimations, some prototyping, just like having a virtual clone of the environment on which you can make an estimation. Or “I’m going to build that feature, tell me what I’m going to need in terms of architecture, how much it’s going to cost, and what is going to be the return on investment of this feature according to some business metrics I’m going to give it, like the number of streams or number of clients.” And that’s going to help me guide my decisions in terms of investments.

Chapter 16 — What Agentic AI Looks Like at Enterprise Scale

Jean: So also for the executive or for the business really, business guys, it’s going to change a lot the way that they are going to estimate the cost, that they are going to plan, what’s next for the platform.

Kevin: Do you see AI as more of an advisor for FinOps, or are we moving toward full AI execution?

Jean: Definitely an advisor for everything which is about building your road map, about evaluating the cost of new features, about the value actually of the cloud. We really have the FinOps assistant, the FinOps assistant helping me make analyses that take a long time to do, and take a few consultants. With good agents, you can do it in a few hours, or in an afternoon you can do a very good study.

Chapter 17 — The Next 2 to 3 Years of FinOps Evolution

Jean: And for optimization, yeah, we are going to have FinOps being more like orchestrators of agents and defining the guardrails and the policies of those agents, and what they are supposed to do and what they are not supposed to do, more than themselves acting and doing some stuff. It was already there with everything which is infrastructure as code. It was already coding infrastructure to provision compute resources. So yeah, we’re going to automate a lot more actions.

Kevin: If you fast forward five years, what does your ideal world of FinOps look like once Agentic AI and sustainability are fully integrated?

Jean: No, I can’t move forward five years.

Kevin: Probably at best three.

Jean: Yeah, let’s say three because…

Kevin: Yeah, I suppose five was ambitious.

Jean: Five is like, I got no idea.

Kevin: How about two?

Chapter 18 — Kubernetes Waste, Broken Tagging, and Cultural Blockers

Jean: Yeah, two. No, three. Okay, let’s say because ChatGPT, you know, is probably it’s not, it’s three years old now, like the first GPT that was released publicly. And when we were first playing with it, we were making fun of it because we would ask it “what’s 2 plus 2” and it would say five or something like that. And now it can do so much. So that’s why it’s hard to say in five years. But three years, yeah definitely.

We can generate dashboards, we don’t need dashboards because we can generate them like this. Everything which is cost allocation and virtual tagging, it’s a lot easier now to manage and to remediate when you are missing some tags, for example, which is a big part of the work for FinOps.

Yeah, I see FinOps being a lot more about business and less about really technical FinOps and the optimization itself. I see things being a lot more automated. For example, everything which is Kubernetes optimization, I see it being automatically automated, will do a better job than us.

Kevin: All right. I would love to get an answer in French from you. I’d love to know, we’ve already kind of talked about it, but to get a French version for any French listeners we have, how do you see Agentic AI transforming FinOps, and along with FinOps to transform cloud management?

Chapter 19 — Agentic AI Explained for a French Audience

[Jean answers in French, discussing how Agentic AI transforms FinOps through policies and metrics, echoing the themes covered earlier in the interview.]

Kevin: I couldn’t agree more. I could see that. I could feel that. I feel that. I assure you, with me.

All right, let’s move towards our rapid fire round. I’m going to give you some sentences and I want you to finish them for me in a sentence or two in French or English.

Jean: We’ll do this one in English so I know what you’re saying. Okay.

Chapter 20 — Rapid Fire Round: Sustainability, Waste, ESG Value

Kevin: All right. So finish this line. FinOps becomes truly sustainable when…

Jean: When you’re able to compute the value of your cloud. Okay. Because sustainability is making sure that it’s valuable, that there is some value to what you are delivering. It can be dollar value, but it also can be dollar for the society. So it can be the value that you deliver in terms of less environmental impact, or social value as well. So yeah, for me sustainability, it’s not only about the green things and the flowers and so on, but it’s about making sure that you are delivering some value.

Kevin: One FinOps habit that every company should adopt is…

Jean: Calculate their waste. How much waste they have. Because saying that 30 to 40% of cloud resources are not used. So yeah, they should be tracking the waste. And the other one, they should be tracking what is called the ESR, the effective saving rate, which is how much they are efficient when purchasing savings plans, reservations and also with their ADP. So yeah, couple of ratios: how much waste they have, and how efficient they are, committing to their spend with savings plans and reservations.

Kevin: All right. Do you have a favorite AI tool or experiment right now?

Chapter 21 — Jean’s Favorite AI Tools and Experiments

Jean: Yeah. So I love Cloudgov.ai, of course, for FinOps because I saw the new features. The AI tools I like, so ChatGPT. I love Claude as well. And I started to play a bit more with MCP, which is this protocol that allows to expose some data and services to LLMs. And that’s pretty awesome what you can do with MCP, not only to get some data, but also to take some actions on the cloud itself. So yeah, been playing with this lately.

Kevin: All right, Jean Latiere. Thank you so much for being here. The company is OptimNow. If anyone wants to learn more about you or your company, what’s the best way to get in touch?

Jean: Yeah, can go on the website. I can share also, we have a GitHub where we have been sharing a few things on how to set up MCPs for example on Claude, and you can do some pretty fun demos and doing some stuff like FinOps with Claude. So GitHub of OptimNow, the website, yeah.

Kevin: Awesome. All right. LinkedIn of course.

Jean: Of course, yeah. LinkedIn.

Kevin: All right. Well, thanks so much for being here, Jean. Really appreciate your time and look forward to chatting with you again.

Jean: Thank you. Thanks. See you.

Kevin: And thank you all for joining us today on Unclouded. If today’s discussion made you rethink how cost, culture, and sustainability all connect, check out Cloudgov.ai, the world’s first Agentic AI FinOps platform. We help organizations move from reaction to automation, managing cost, performance, and environmental impact together through intelligent autonomous FinOps agents. Until next time, stay unclouded and stay curious.