Transcript 0:02 So, welcome everyone to a new episode of the FinOps Weekly Podcast. Today, we have a very interesting interview, um, with Ido Koller, the CPO of Planor. How are you, Ido? I'm good. How are you? 0:20 Doing great. Doing great. 0:21 Uh, today it's going to be super interesting podcast about AI and FinOps for AI, AI for FinOps, and all this stuff that you should, uh, stay tuned to know all the, all the insight that we have prepared for, for you. 0:34 And to- together with me is, uh, Damian Monafo. How are you, Damian? Uh, doing great. It's always, uh, you know, a pleasure to be here and talk about interesting things like AI. Yeah. 0:48 So, you know, you know, it's, it's been, uh, probably the, the word of the year, and it's probably going to be one of the more typical word in, in the events that we have in FinOps, uh, now. So, you know, in... 1:02 from, from someone like, like you that are in the, in the core of the, of the FinOps ecosystem, what do you think is the, the distinction or the difference between FinOps for AI and AI for FinOps? 1:14 And what, which of these two excites you the most? Like, what's your preferred part from, from this? Yeah. Uh, that's a great question. 1:23 And I think AI is probably gonna be, like, the, the word of the decade, not only of the year, like, with the pace that things are going right now. 1:31 Um, and I think both FinOps for AI and AI for FinOps are super interesting topics. So FinOps for AI, I think it's really in the early days of that. 1:41 So that means, like, companies using AI both, for example, when developing and then, like, creating new code using AI, but also using AI services. 1:51 So for example, using OpenAI or Anthropic or, uh, AWS Bedrock and all of those services, and also training their own models. And that's becoming more and more of a big issue for companies and a big spend. 2:02 And if you're looking at, uh, let's say, the forecasts of companies like AWS or, or GCP or Azure, then you definitely see that people are predicted to spend a lot more on that. 2:12 So it, it makes a lot of sense that FinOps will take care of that. And on the other hand, AI for FinOps is, is completely different. 2:19 So it's not about doing FinOps, um, on more things, but it's about completely changing the way we're doing FinOps. 2:25 Because like in the same way AI has been a revolution for, for example, writing code or generally in, in across many of the SaaS platforming industries that we're in, I think in FinOps it's gonna be a, a drastic change as well. 2:38 And we're starting to, to seeing it happening, uh, even today. So that means completely changing and augmenting the workflows that we're currently doing to automate them a lot more using AI. 2:48 But it also means understanding what we can do more, what used to be impossible but now becomes possible through the usage of AI, and how that can enable FinOps team to be even more central within companies and push for much more effective resources. 3:00 Um, so between the two, uh, we tend to think, and me personally, but both Planor at large, that, uh, AI for FinOps is should, like, should be what we're attacking right now, and not FinOps for AI. 3:12 I think mainly because currently, like if we're looking at twenty twenty-five, most companies are still experimenting with AI. 3:19 And I think it means a lot about the place of FinOps within there, both that there isn't necessarily a lot of spend yet, but also that that spend isn't really thought of as something that should be properly managed right now. 3:31 Because even leadership thinks about that as like, let's understand what's working more and working less, and let's find out like, uh, if we can get more market capture through doing those things. 3:41 And that's not, like, a good point in time to try to... like, for the FinOps team to get involved. And we see a lot of people, like, having issues when they're trying to, to push their leadership on that too much. 3:52 But I think even more interestingly, it's because the pricing models are not really set yet. 3:57 So that means that the way people will use AI in two years is probably be- gonna be very different than how they use AI currently. 4:04 So the way they are paying for things, the reservation models, if they're going to train it or use open source models or use SaaS models, and if it's gonna be, like, one company or a bunch of companies. 4:14 And I think that means that trying to, like, create a process out of it too much is gonna be incredibly difficult right now for many FinOps team. 4:22 So at least our, our opinion is that we should wait a bit for things to be a bit more stable, and in the meantime, we can focus on, on changing a lot of other things with AI and making the current work that people do way more effective. 4:37 No, that's, that's probably a good answer. And I think it will probably surprise a lot because, you know, in the, in the feedback that we got, this depends on, on who you ask. 4:46 Uh, people are more worried about, you know, applying the FinOps for AI and, you know, control that. Uh, but that you're oriented on the, on the AI, uh, for FinOps and how, uh, you know, AI can help on the FinOps. 4:58 It's, it's very interesting, right. Um, so, uh, Damian, do you have any, any follow-up for, for this? 5:06 Uh, well, it's interesting to see that, uh, today, you know, uh, what I see people using, uh, AI for FinOps is when you ask... 5:18 I, I play a little bit in the, with Amazon Q, and you were asking, you ask like, "Give me the top ten," or, "What is the most, uh, expensive, uh, service?" and stuff like that. 5:29 I see the, the, um, uh, we are in the beginning of, of that, um, and I would like to see again how, how it develop. Um, a- and yes, a lot of, uh, we see also a lot of traction on, uh, FinOps for AI. 5:46 How we're gonna, as FinOps, you know, practitioners, how we're gonna apply that. Um, but it will be interesting to see all the things that, you know, a lot of 5:58 talks about agents are being there, um-I, I, I wonder- Yeah... how that will be applied in, uh, what do you think, how that would be applied in FinOps? Yeah. That, that's a great question. 6:13 And, uh, I think what you're saying also is very correct, that a lot of what we're currently seeing in AI for FinOps is a lot of hype, like the Amazon Q example or the Copilot, where me personally, I, I don't think what they're doing necessarily is super useful because it's super basic. 6:30 So, uh, it, it takes the data, the most basic data that you have today, and it shows you that in a completely new chart that's, like, for example, not interactive and doesn't have to do with your current workflow. 6:41 And I think it's, like, a, a good starting point, but, uh, I think it misses the, the key things that AI can really do for FinOps, which is actually finding from, like, the breadth of data that company have, and contextualizing those things automatically, and creating much more interesting explanations. 6:58 And I think agents are a part of that, so both, for example, using things like MCP servers and creating interactions to go automatically, and using the more advanced AI models. 7:08 I think more broadly, it's, it's about getting, like, the, the right level of explainability such that everyone could access this data. 7:15 And I think that's, for example, let's say, uh, a finance person, uh, wants to use Amazon Q, or a finance person wants to get, uh, like, an answer to something that's going on with FinOps. 7:25 They're probably not gonna know the, the EC2 ID to ask Amazon Q and, and tell them, like, "Oh, look at that resource and tell me what happens, and then look at the CSV of the result of that." 7:36 So the way they could get more value from AI is if they could ask questions in the way they're used to ask questions. 7:42 So let's say talk about margins, talk about the unit economics of their business, talk about the trends that they are seeing and asking for explanations, and then get answers also in terms that they care about. 7:52 So not in terms of this is the EC2 IDs and this is a CSV of its cost day by day, but in terms of who's the owner of that, which project it relate to, which customers it relates to, and that's what makes me so excited about that. 8:05 The ability to democratize, to contextualize, and do that even if those people are not really experts in what they are doing. That's, that's, well, that, that's, that's super interesting. 8:18 And I think that, yeah, we have seen that at the beginning, the Copilots, the, the Qs, they are very helpful, um, at s- at, at some point, but it have a limit. 8:28 And also it's not, like, the way to scale is not that high because you cannot do anything at the scale with a Copilot or with a Q. 8:37 Uh, it's very helpful on, on server side, but the, like, the agentic system is, is way better. 8:42 And talking about the agents and evolving a bit about the, the convolution, like, you know, the agents are super great, but we have seen that they could make mistake. 8:51 Like, they could be great, but they could delete a database or make, like, if they, there's no, like, manual control. 8:57 So where do you think, like, the line of agentic AI is and the improvement that you, do you think, or how people should use it to be better on, on these agentic AI systems? Yeah. 9:09 I think even, like, regardless to, to AI for FinOps, so for FinOps specifically, uh, I think the trends that we're seeing with AI and let's say with the vibe coding, uh, and people generally using, like, very automatically generated code is, is extremely scary from FinOps perspective. 9:25 Like, the ability of someone to just, like, press Tab a bunch of times, and then accidentally having, like, 100 CPUs instead of 10 CPUs because they, they didn't even look at the Helm chart they wrote, uh, is something, uh, incredibly scary. 9:39 And I think it's even becoming more scary when some companies, for example, talk about automatically changing your codes, creating PRs, reviewing them, and p- pushing those things to production. 9:49 And my, my, my ideas on that is that AI should augment what people currently do, but not take the actions itself at the end of the day. 9:59 So that means, for example, because, like, the bottleneck is never, like, someone, let's say, reviewing a PR and pressing Approve. 10:06 The bottleneck is often that people, first of all, don't really trust the insight that they see. So they see, for example, that, uh, a database is not utilized enough. 10:16 But they want to check if maybe there is, like, once a week that this database is utilized, and if they're gonna resize it, then it's gonna make an issue. 10:24 Or maybe this is, like, a test project, but in two months this is going to scale up and the usage is gonna scale up, so we should keep the right size right now because otherwise in the future, uh, it's gonna create performance issues. 10:36 So I, I think the key thing is for every insight, and this is like a, a basic example, let's say a right sizing of a database. 10:43 But even for more complex things, you need to know the context of how it interacts with other systems, who's the owner of that, get the right question to those people, get the answers, and then understand if this action can be taken or not. 10:55 And I think that's exactly where we're heading to with AI as well. 10:58 So automatically investigating, integrating with all of those external systems, getting this data, and wherever it, it doesn't have the data, sending a Slack message, requiring a response, opening a Jira tickets, doing all of those extra actions that get the right people involved, and then kind of automating that process. 11:15 And then at the end of the day, it could also allow you to, like, uh, see where is the part of the code that you need to change, et cetera. 11:22 But I think that's, like, the, the easiest part within the process because usually when you have all of the answers, things are quite clear. 11:30 Um, so I, I think that's the way that's gonna evolve as far as, like, remediations and not as far as, like, automatically creating patch and running, like, Terraform modules in your production environments. 11:42 I hope that, uh, it doesn't all, uh, do all of that because even if I do mistakes when I do run, like, uh, Terraform code, I cannot imagine what an agent on, like, auto doing everything, what could, the mess that they could make [chuckles] for, for Terraform.Yeah, for sure. 12:04 Uh, and I think we-- by the way, we, we also see that today, like even in examples that, that we see within customers, that sometimes you, you can see like, uh, currently it's even like very, very basic examples because I don't think it's that integrated within systems. 12:18 But for example, we saw a huge spike in one of our customers for like, uh, a query that ran on some, uh, data lake within their infrastructure that basically costs like, uh, tens of thousands of dollars for one query. 12:31 And when you look at that, it's pretty clear that that query, that SQL query was basically AI generated, and the parameters weren't good enough. It scanned way too many tables, et cetera. 12:41 So I think you, you already kind of see those issues today that obviously also existed with some analysts not being proficient enough, but I think you, you definitely see them more where people are just copying and pasting. 12:51 So that's definitely gonna happen more and more. Damian, do you have any, any stuff, uh, you want to ask me? 13:06 No, I was, I was thinking also that, you know, uh, at this point, I will trust AI also as, uh, you know, something that help you organize and help you nudge the people, uh, which will be like a chatbot, you know, like an agent, uh, that will help you prioritize based on, you know, past experience. 13:23 I totally agree with, uh, what, uh, Ido was, uh, saying. Um, and, um, but, uh, just thinking about something else. Ido, you, you come from cybersecurity background. 13:38 Uh, do you also see the FinOps like, uh, 13:43 like in the first step of se-security, in which, you know, uh, uh, the security guys were like, you know, these people that were bugging you, and today it's more of like, you know, something that is fully integrated into all our system. 13:58 Uh, do you see FinOps the same, and do you think that or when do you think that it's gonna fully integrate in all our flows? Yeah, that, that's, that's super interesting. 14:10 I think there are many parallels between, uh, FinOps and security. 14:13 So, uh, and I think by the way, you see even like with companies starting today within the FinOps space that, uh, many of us founders came from security background. I think it, it, it's-- it makes a lot of sense. 14:24 So, uh, I think very broadly speaking, if you're looking at, at the cloud and like the way people think about the cloud, so probably, you know, DevOps, observability tool are on the forefront of that. 14:34 All of those open source that really invents what we consider cloud today are definitely like, uh, the first line of defense on what's really going on in the cloud. 14:42 I think cybersecurity was like a close second to all of that, mainly because when something new happens, there are risks to it, and people need to take actions on that immediately. 14:51 But I think FinOps, uh, at least up until recently, was kind of left behind. 14:55 So many of the vendors were created at a time that, for example, you, you couldn't even access the, uh, cost explorer in AWS, but you had to download a bunch of like big Excel files. 15:05 So obviously having like a basic BI tool was enough, and those people were like BI people, but they weren't really cloud people. 15:12 And you see that with the fact that they were very late to, for example, understand Kubernetes properly or understand data lakes properly or understand how, how do things interact with one another. 15:22 And I think one of the key things that we bring from, from cybersecurity is that understanding that the cloud is much more like a graph or a map of resources and not really like a table for finance, and that's it, and that things really have a lot of interactions within them. 15:36 And that it's kind of alive a lot more than people think with how things are constantly being deployed and evolved, and you need to take that like shift left approach and correlate that with what engineers are doing. 15:47 I think that really is why also we're seeing a lot of the trends you're currently mentioning of like pushing things more and more to the engineers, getting them involved, and getting them to, to take actions. 15:57 I think that's also when you see a lot of the problems that used to happen in cybersecurity also becoming a real problem within FinOps. 16:04 So I think like the most common problem peop-people talk about in cybersecurity is like, uh, the noise of alerts and the fact that there are many things that are irrelevant. 16:12 All of the tools throw you like a million ticket, and you don't really know what to do. 16:16 And I think recently you see those same trends, especially when engineers are the one facing those tickets i-in FinOps as well, where people are saying, "Uh, there are too many anomalies, there are too many recommendation opportunities. 16:27 What can I do? How can I prioritize? Does it make sense? Is it a false positive," et cetera. 16:31 I think a lot of the cybersecurity mindset is also about like, what are the good tricks to look at that and find kind of the, the needle in the haystack or to find the interesting thing within like the noise of these alerts. 16:43 And that's also like a lot of our focus as well using AI actually, because AI is incredibly good at like smelling what's not good enough. Like let's say you, you had an anomaly. 16:54 It's really hard if you're just using like a, a forecast model to, to look at, let's say, uh, a cost that happened, let's say a marketplace purchase. Uh, so you bought through a marketplace some tool. 17:05 It obviously looks like an alert. Any tool is gonna say that's an alert. 17:08 But using AI, if you bring the context of that to an AI layer, it easily can tell you that that's not something that's really problematic, but probably like something that the business is doing and is not like something that you should open an, an urgent issue on. 17:22 Although it's like a, a hundred k of you spend on a day because it relates to a marketplace, it relates to a vendor. 17:27 So contextualizing those things is actually like another extra layer of both, uh, reducing noise, both compressing things. 17:34 So looking at many like unrelated alerts and combining them to one like big incident that have all of that context together. 17:41 But also to make those, uh, alerts or incidents or recommendations much more actionable by the people getting them and routing them to the right person. 17:50 And, uh, I, I think that's like definitely a, a lesson learned from, from security, and you can see that in, in many security tools as well today.Yeah. Yeah, indeed. 18:00 And also like the actionability and the remediation and all this, you know, interaction with additional components, it's, it's a different... and also the, the feature that, that is more common. 18:10 But, and I wanted to, to, uh- And, and one, one... I want to add one thing. Sorry, Victor. It's, um, uh, that there is a good thing that you brought up, you know, it's, you know, the false positive. 18:19 Especially when, you know, we all the time as practitioners say, "Let's put not only anomalies, let's put budget alerts," right? But when you put budget alerts, where do you put it? Like, 10%, 20%, 50%. 18:30 I make sure you're getting all these, [laughs] all these nets, you know, that, that doesn't give you anything. You... First, if you don't correlate it with time, which could be one way to avoid it. 18:40 A-and secondly, when you have, you know, uh, now with this direction of multi-account strategy, you have so many accounts, so many, uh... I'm not sure you'll get overloaded, but, uh, this, this is just very much- Yeah. 18:52 I think the main pro-... I wanted to add. Yeah. Like, I, I think the main problem with alerts is, uh, like, uh, at what level of obstruction you're looking at. 19:01 Because let's say you're alerting on, on every single resource. You're doing, like, for every resource ADC, it changes. 19:07 Obviously, you're gonna get both a lot of false positives, because let's say, uh, there is an auto-scaling group that turns off and on new machines all of the time. 19:16 You're gonna be alerted on any new machine if it's, like, big enough. 19:19 But also it means you're lacking the context, because usually this machine went up, but also this bucket went up, and also this networking cost went up, and they are related to one another. 19:28 And if you're getting three different alerts, you're missing the full story. 19:31 But on the other hand, if you're looking at things like, did my cost of this account went up, then usually it's really hard to know what happened. 19:39 Because even if you see there is a change in the account, now we need to go in and understand what service it relates to, what's the team, which, uh, resource, et cetera. 19:48 So I, I think, like, the, the way we're looking at that is that doing alerts on the higher level things makes more sense. If you can, once you get an alert, explain it really, really well and find the root cause. 20:00 So we put a lot on effort that whenever we have an alert, for example, in Panor, we automatically analyzed all of those things and find all of those like, uh, mini alerts within it that really trigger the cost change, and then tie it all to that one big incident and one big story of like, that what caused the change within your testing account. 20:18 Um, and then it makes more... those things also much more like, uh, less noisy and more actionable. And, and learning also- It's, it's really difficult... 20:28 things that you're doing, because if you're doing a report, eventually it's look, uh, very interesting. Nice. Thank you. Yeah, for sure. And, and yeah, as mentioned, it's really difficult because for someone that... 20:40 for me, myself, that I come from an infrastructure team that is very, you know, you get a lot of alerts from Kubernetes, a lot of stuff, and being able to differentiate yourse- like, differentiate what's a good alert from a noisy, like, you know, a pod there and this stuff, but it's, uh, not meaningful. 20:58 So the way that you approach, uh, uh, making an interesting change or what has changed or what is an actual problem, it's, it's a great approach, and thanks, thanks for sharing that, and I'm sure the, the people can do like a good job. 21:10 But a- and, and diving more into, you know, the, the long term, uh, or what we think... 21:17 could think that the long-term benefits would be for, for AI, like in FinOps, what do you think it's going to be the long-term benefit from, from AI, or what do you think it's really going to be changing apart? 21:31 Like, what's going to be empowered? Uh, is it going to be like a FinOps, a augmented professionals or less, let's say, less manual work and more AI automated, more oriented on the agents? 21:44 Do you think we are going to have like an agent team of, of FinOps and like only one human or something controlling it? What, what do you think is your... the approach that we are going to have with, with AI? 21:56 So, uh, I think there are going to be two things happening simultaneously, which are, uh, incredibly interesting. 22:02 So first of all, there are so many, like, uh, menial tasks many FinOps teams do today that are gonna be completely automated, and that's gonna be great. 22:12 Um, so instead of like, for example, there are many people I, I work with that spend probably half of their week chasing teams about, like, cost changes and anomalies and recommendation opportunities, and verifying what really happened, does it make sense, can you make this change, et cetera. 22:28 If all of that, like, investigation time, finding the right owner, et cetera, is going to be automated, it's gonna save people a lot of time. 22:36 And also when regarding, like, the more, let's say, reporting aspects of FinOps, if it's gonna be very easy, basically with a click of a button, to describe what report you want to get, get that report, and also get the five most interesting trends that happened last month compared to the previous month or last Q compared to the previous Q, and then present that to leadership without taking two weeks every quarter of creating that presentation, that's also gonna save plenty of time to, to what many of the teams are, are doing today and what's expected of them. 23:06 And that's not even that far away. Like, uh, we, we see many of those insights even being generated today with current FinOps tools or with, like, AI augmented FinOps in some ways. 23:16 But I think, like, if you'd only look at that, you might think, "Oh, we're gonna need less FinOps people." But I think that's not gonna be the, the, like, the gist of it at all. Actually, quite the opposite. 23:25 Because the thing is that today, uh, hiring a FinOps person is, is often like their impact is limited because they are drawn to all of those things that not necessarily have, like, incredible value for the organization, but take a lot of the time of that person. 23:41 That probably limits how much you can even, like, pay on FinOps, not necessarily on a specific person. 23:46 But, like, if I bring one more person to that FinOps team, it's not gonna necessarily bring a lot of values because they're gonna be blocked on getting all of those answers, and they aren't gonna bring necessarily that much value to your organization. 23:57 But unlocking that possibilities for those people to ask much more higher level questions, so to reason about what sort of ar-architectural changes might be needed, or how does the unit economics of this feature really works, and does it make sense, and where should we put more efforts, or be involved much more earlier on within the process of developing a new feature. 24:17 So when you both unlock the time for people to get involved in that, but also give them the capability to be way more powerful with the time to be able to do that properly, then it unlocks a lot more new opportunities for what FinOps means for those organization and make those... 24:33 these teams way more central. 24:35 So I think that's also probably a trend we're gonna see that these, like, these people are gonna have way more power to influence what's going on within the organization because these issues are, are critical. 24:45 Like, at the end of the day, people are spending millions, hundreds of millions of dollars on cloud, and getting even, like, 5% more efficient, 20% more efficient is incredibly impactful for the company. 24:56 So it makes sense to, to put resources on that if it makes an impact, and the ability to make an impact is gonna be much larger when you have more tools to do that properly.Yeah. Very good. I, I, I... 25:10 Yesterday we were in a- we had a Spanish, uh, uh, summit, and we were seeing a very nice, uh, you know, demo about, you know, where, where, you know, the AI is, is here, and we, we just discussed about it, that the AI at this point is not to replace us, not, not even close, but it's there to make us better, to help us be more efficient also. 25:35 Uh, and that's, that's a, a very good point. 25:37 I always say that unless you have, like, a, a robot that is, [laughs] you know, mingling with the people, uh, until that happens, I don't see how AI and something else, uh, uh, uh, replace us. 25:50 I think that we are still valuable, still we have a long, [laughs] a long to go, a long way to go. I'm sure, I'm sure we'll get robots soon as well. Uh. [laughs] Yeah, but think about it. Yeah, it's gonna be fun. 26:02 Robot... And by the way, I was thinking about it while I was, you know, thinking about what, [laughs] what to say. But even robots, eventually think about it. When a robot is gonna come, e-eventually, I think it's... 26:13 We are a little bit shifted, but eventually we'll come, uh, um, also like people, we're gonna, you know... It's gonna be like you think, "Oh, that, that is a robot." I don't know. 26:22 It's gonna be some kind of like a, you know, distrust, you know. It's never gonna be trustable. And still, I think we have many years [laughs] there to go until, you know, AI and robots replace us. But, uh... 26:33 And that's good, you know. Yeah. We, we have some time. Uh. [laughs] Yeah. Yeah, for sure. And, and I wanted to, to, you know, dive into an interesting topic because I... 26:46 You know, when, when we talked in, in FinOps in, in Barcelona, then we, we, uh, get to know each other, and you, you tell me that you were champion in, in debating and, you know, you are very experienced, uh, in communication, and you can tell more now to, to the audience because I think it's a very interesting story. 27:04 But related to that, like, uh, communication, you know, uh, discussions in FinOps is a very well-known topic, and especially one of the main things is the getting executive buy-in. 27:15 So from someone that has a lot of experience and a lot of skill in communicating and, you know, at convincing people, um, how do you think these skills, uh, can be of help to FinOps? 27:27 And how has it been helpful for you, like, in your career? And you can explain a bit about your, your story because I think it's, it's very interesting. Yeah. 27:36 Um, so, uh, a-as you've said, I do have some experience in, in debating. 27:41 So, uh, I've done, uh, like, uh, university debating and debating at large for, for years in my past, both, like, from Tel Aviv University where, where I studied, uh, a-and generally more globally. 27:54 And in there, uh, I basically, I, I won the, the Israeli championship, the European championship, and, and many other tournaments, and then also helped organizing some of those tournaments with- for, for, for other people. 28:05 And I think people usually when they think about debating, they, they think about, like, people who like to argue. They think actually it's, it's quite the opposite. 28:13 It's more about people who like to listen and, and find out, like, what the other person really means. Because the way debating is, is being done and being adjudicated is mainly about engagement. 28:23 That's like the word we mainly use. 28:25 So finding out the main, like, points of contention, and then talking specifically about them, and finding out what did exactly the other person say about that and what would change their mind about that topic specifically. 28:36 And I think that kind of mindset is often needed, uh, both in FinOps but at large, because usually people come with their own perception about what's going on in reality, and then they end up talking about two completely different things and not really listen to each other, and then, like, they can't really get to any improvements. 28:53 So it might be the case that, like, the FinOps person tells, like, an engineer, "Uh, you could have saved so much money. I, I had... Like, why didn't you do it? It wouldn't have taken you so much time." 29:04 And the engineers tell them something like, "But I have my tickets. I listen to my product manager. I can't do those things." 29:10 And both sides are going to be, like, especially if it, if things get a bit emotional, then it can be, uh, a bit challenging. 29:18 And, and I think that the key thing in many of those things is find out the incentive structures and find out what really drives people to do things, then find out who can make those decisions. 29:26 And I think that's also, like, for example, why a, a lot of, of my focus when helping organization integrate FinOps, uh, like, to their processes, it's about creating accountability across the organization. 29:38 So I don't think it's necessarily about just, like, chasing people on, on Slack, but it's more about making sure the teams feel accountable and creating things that people can agree on and say, "I really drive this cost. 29:50 I'm really the owner of this cost, and therefore, I should be the one that looks daily or weekly on the trends, on the opportunities, and finding ways to reduce it." 29:59 And once you have that, then all of the communication becomes a lot easier. 30:02 But to get there, usually there are a lot of processes around, like, agreeing on who really is in charge of what, how you should allocate things like shared costs, et cetera. 30:12 And for that, I think that definitely people have to come prepared. 30:16 And I think maybe if I have, like, one tip, and I don't think it's, like, from, from debating, but I think it's more like, uh, about how business is really done, is that whenever you're coming to, to a meeting where a decision like that needs to be made, let's say meet with a lot of teams and, like, uh, kind of split ownership on some resource or, or create a budget and come to an understanding about budgets, always talk to the important stakeholders before that. 30:42 So make sure you get to a one-on-one agreement, because getting to one-on-one agreement is, like, a lot easier usually. 30:48 And once that happens, then the big meeting and the, the important meetings become kind of like, uh, very quick and very easy. 30:57 And then, like, you, you don't really get to all of those disagreements that create, like, an explosion, and then nothing really happens, and thing takes month because you, you can't really reach a consensus about those things. 31:08 So, uh, I think that's definitely the thing to do both in FinOps but, uh, even more largely.May- maybe you can share with us some, uh, links to improving debating. Uh, but I totally agree. 31:23 You know, there are some times that people ask me, "What are the skills that I need?" You know, "Do I need to know how to program? Do I need to know..." 31:29 I say, "First of all, you need to be a, as a FinOps, you need to be a people, a people, a people guy." 31:34 You know, you need to get along because eventually, you are, FinOps is in a point that you need to communicate with so many different teams and so many different type of persons that it's your... 31:45 The first skills you need, less if you know how to, you know, create a Lambda or, or, or, or any, you know, automation. More important is people, people skill. So, uh, definitely, uh- Yeah, for sure... 31:57 maybe we'll talk more about it, about debating. [laughs] Yeah, and I, I think it's very important to, to kind of remember that most people want to do good, and most people want to save money. 32:09 Like, uh, engineers are not, like, the enemy, and definitely finance is not the enemy. Not everyone's trying to harm you or harm the business or, like, s- waste your time or their time or close your project or whatever. 32:23 But it's about what they're being measured on and what can they also use for their own goals. 32:30 So talking to them, understanding them, and helping them ch- create a shared story about how you can both help them and help your goals is, is probably the thing to do whenever you're trying to push a project within an organization where the ownership structure is not really clear, like FinOps is usually is in most organizations. 32:51 Indeed, like, uh, uh, the, the human factor in FinOps is one of a crucial thing and, you know, being able to communicate itself and being able to put your arguments in, in place and also especially relating to people and being able to empathize and to put, like, what's the benefit from them to do this or what's the benefit from the company and how they are going to impact. 33:12 I think it's, let's say, one of the sector that is easier because, you know, it's linked to business and it's linked to, uh, margins, and it's easier to highlight the numbers. 33:21 But also, like, how these people is- should be interested in that and not in availability or, or whatever. Um, so thinking about this and thinking, you know, what we've been discussing, 33:34 uh, you know, we, we like to do some predictions or some, uh, comments or some, you know, thoughts for, from, from the interviewees about, you know, what's going to happen in five to 10 years in FinOps? 33:46 What do you see FinOps in the next year? 33:48 So I want to know, like, what do you think is the biggest myth in, in the next, uh, five to 10 years in FinOps, whether or not what's going to happen in FinOps, and something that you don't think people is talking about and maybe how, you know, Pelano is, uh, approaching or, uh, highlighting that. 34:06 Do you, do you have any, like, crazy or prediction that you think nobody's, is talking about in FinOps? So I, I think basically that probably the, the unitless dollar is gonna die. 34:18 So like, like what you've said, that people, it's easy to talk about it because it's numbers. 34:23 I think the numbers we talk about today could be considered meaningless or almost meaningless and will be considered almost meaningless. 34:30 Like saying, "We've spent," I don't know, uh, uh, "two million dollars on AWS last month," uh, isn't gonna be, like, the conversation, but it's more gonna be about the cost per value, and that's the way people are gonna change the way they're talking. 34:43 So the cost per API call, per customer, per interaction. 34:47 I think we're already heading there in some sense, and more and more platforms are, like, changing even their pricing model, their basic pr- pricing model to be that, m- moving even from, like, users in SaaS or, like, a license to something that's much more based on a specific interaction, and that's also gonna drive the business to talk about things in that way. 35:07 Because at the end of the day, if your business is growing and if you're getting more value, then it doesn't matter that you're spending more money. 35:13 I think, by the way, especially with AI, it's gonna happen more and more, that you're gonna spend a lot more but get more money because people pay you for that interaction. 35:20 So the thing that should make sense a lot more is to talk about how that changes over time and how that trends with new feature releases or with new model releases if you're using a new model or with the patterns that people are using your tool, and that conversation only makes sense when it relates to units. 35:36 So I think people and FinOps and, and leadership is less going to talk about how much we spent, but more about are we doing that efficiently? How much value do we get per dollar we spent? 35:48 And that's gonna be a, a massive shift that, that hadn't happened yet in most places. That's a good one. 35:56 Uh, haven't thought about that, but yeah, if you relate to what AI is doing and, you know, how the credit system is working for, for APIs and all this, uh, like what Cloud or what OpenAI does, uh, I think it makes sense. 36:11 And I, and I hope that especially once the, the provider, you know, after all the providers do relate based on service consumption in some places. 36:20 Um, so I think the only need, the, the only step away is linking it to business value easier and being able to provide the unit economics in an easier way. 36:30 Um, one, I hope AI can help on that, um, and automation can help on that. But yeah, I th- I also agree that, you know, we, FinOps is always... 36:39 I, I always understood since the, since the beginning that it's linked to business value, and it's not about the cost itself because, you know, cost can go to zero if you don't have anything in the cloud. 36:48 Uh- Yeah, exactly. But, uh, if you go, [laughs] if you go that way, um, you need a, we need to be better at automating unit economics and to, you know, being able to link it. 36:59 And I would say that probably the, the, the cloud or the, the FinOps ecosystem should evolve to that. Like, okay, this is related to this, uh, business. 37:08 Okay, let's, let's make that API, uh, related cost, or let's make it more relative and not, like, the absolute huge mega bill from, from AWS or from any of the providers, right? Yeah, exactly. 37:21 And the fact that it's not simple, the fact that it's not easily taggable doesn't make it, uh, less valuable. I think that's where a lot of our focus is in Pelano, making that process more autonomous and more easy. 37:35 But, uh, regardless, uh, I believe it's something companies should definitely consider doing internally because it brings a lot of value when you do it properly. Right. 37:45 I'm not sure we are going, we need to go into a world that is, does it make sense? Why? Does it make sense from business perspective, from usability perspective? And yeah, it's good that we are going in that direction. 37:59 Indeed. Um, so yeah. Uh, I, I would like to, to first of all thank you for, for having the interview. I think it's been a, a pleasure, uh, to have you today. 38:10 And I don't know if you have any quote or any, like, golden minute to, to share with, with the audience, and if you have any, any message to share before we, we wrap up, and, and, you know, pleasure to, to have you. 38:25 Yeah. It's, it's great being here. Uh, and thank you for, for inviting me. Uh... 38:33 Um, pleasure to, to have you, Ido, and thanks for sharing your, your insights on the, on the AI, and thanks, Aman, for, for coming back today. And hope you, you guys enjoyed the, the episode, and see you in the next one. 38:47 Bye bye. Bye bye. Bye. Uh, stop recording.