FinOps in Action

FIA - Marit Hughes
===

Marit Hughes: [00:00:00] if you don't know and aren't working together on your business goals, um, your strategic. Initiatives. How do You know what data to even bring to the decision as a FinOps practitioner? Right? for A lot of folks to shift into that cloud, cloud mindset or take risks with their deployments, feels a lot like being dropped with no map into the middle of wilderness for them, and it causes them to make decisions that are not for the betterment of their solution or their monthly bill.

Intro: Welcome to FinOps in Action. I'm your host, Taylor Houck. Each week I'll sit down with FinOps experts to explore the toughest challenges between FinOps and engineering. This show is brought to you by 0.5, empowering teams to optimize cloud costs with deep detection and remediation tools that actually drive action.

Taylor Houck: Hello everyone, and welcome to another episode of FinOps in Action. I'm Taylor Houck, and today I'm super excited to be taking over as host of [00:01:00] the show. And for my very first episode, I couldn't ask for a better guest. She's one of the most experienced thin ops practitioners in the world with over a decade's experience of managing cloud spend.

Today she works at Deloitte helping federal, state, and local governments implement multi-cloud FinOps programs. This is someone that I've been lucky to know since my early days of FinOps when she actually used to come to the in-person meetups I hosted, and I would always make an effort to talk to her because it's hard to walk away from a conversation without having learned something.

Welcome to the show, Marit Hughes.

Marit Hughes: Taylor, thank you so much for having me, and 0.5, thank you so much for having me. I've always enjoyed talking to your entire team. Whenever I get a

chance.

Taylor Houck: Oh, thanks, Marit. I I appreciate that. And, uh, before we get into the meat and potatoes of the show, I wanna ask you something on a, on a personal basis, right, because I know a lot of people in FinOps and you've been doing it for longer than arguably any of them. Can you tell me a bit about what it was like?[00:02:00]

When you first started working in in FinOps, and when did you realize this was something that you were gonna spend a decade working on?

Marit Hughes: Yeah, absolutely. So I processed my first AWS invoice in May of 2012. It was for about $3, give or take. and it was a credit card purchase, uh, for nasa. So I felt very cutting edge and leading, uh, on the cloud in the government in particular. It. Back then, there was no cost and usage report. There was not even a DBR, the detailed billing and resources with, with resources and tags from AWS. There was just a cost allocation report. Um, commitments did not come in all upfront, partial upfront, and no upfront. Our eyes came as light mediums and heavies, and if you ever want someone to try and explain it to you, it should not be me. 'cause I don't know that I ever fully understood them. Um, [00:03:00] and I did a lot of things fully by hand, manually in Excel. Um, the tools didn't exist. Not only did 0.5, which we know came out of self, uh, a year and a half ago. Um, none of the tools that we now think of the old guard even existed yet. Um, so it was a lot of things. Um, in 2016 ish, um, I'd been with the company I was at for about eight years and I was like, oh, what am I doing now? And in the grand scheme of cloud, now what I was doing was not much, but at the time it was huge. I had 13 payer accounts. I had a thousand linked accounts that was a lot 2016, and I told somebody I wanted to help people manage their cloud bills and, but the problem was with nobody wanted to admit that they needed help. Uh, a year or so after that, I ended up at a third party tool. Um, after a brief [00:04:00] stop actually working on contract directly for the government and then ended up a distributor and then ended up where I'm at now, where tools are the lifeblood of everything, there's a third party tool for every niche that you need. Um, and very few people have ever actually looked at the raw curve because it's always coming into a tool. So it's been through this huge journey. There are now entire teams of things that in 2016, people didn't want to admit that they needed help with. So to. from one extreme to the other in less than a decade is quite something to

behold.

Taylor Houck: Yeah, I mean, you really were had that first row seat of seeing the entire industry not only evolve, but even just form in the first place.

And

Marit Hughes: absolutely.

Taylor Houck: now, like, You know, when you. Look back on, on that. Right? And, and through FinOps there's always new people coming into it who are asking questions about, You know, how do we get started?

I, I, I just got tasked with managing my company's cloud spend. What should I [00:05:00] be thinking about based on your experience and, and talking to people at different conferences and getting to know them, what do you think that people get wrong in their FinOps journeys in 2026?

Marit Hughes: I think the main thing they get wrong is thinking that the way that they're doing it in their organization is directly applicable to how other organizations are doing it. Um, I get a lot of requests for benchmarking. Can you benchmark me against this other thing? Well, I can give you their, their data points, but you have very different models, Of what you're trying to do. So Sure, this one has 60% of their resources scheduled, but they process batch jobs like this. That's all they do. So it's a useless metric, right? And as FinOps practitioners start to change, um, organizations and industries. Is where I think they're gonna learn that lesson the most.

Because typically if you stay within your own organization, [00:06:00] even if you change teams, for example, you get like, uh, you go from business unit A to business unit B within the same organization. Those motivations and drivers are, are frequently fairly similar. Um, but when you go from. Entity A to an entirely different one with a different business model. That's where you're gonna realize everything you thought you knew about FinOps is incomplete and in inapplicable to the next thing you're on, what you're talking about is that culture coordination, um, critical thinking, data-driven decisions like those should always be a play. But what data you need, how decisions are made will definitely vary.

Taylor Houck: That is such an interesting observation and one that I can relate to on such a deep level because when I started in FinOps, when I was first hosting those meetups that I referred to earlier. I was working for a software company, right, and what I was doing FinOps on was like it was development workloads, [00:07:00] right?

I then moved on to NBCUniversal. At NBCUniversal. a lot of what I worked on were ERP systems. We're talking about SAP implemented at. Huge enterprise scale. And when you work from, You know, the, the quick dev, Hey, just shut off these idle things. And, and worst case scenario, someone screams to an environment where you have real downstream impact when their performance issues or things go down, it's a whole different ball game.

Where, where have you seen this play out in the, in the real world?

Marit Hughes: Um, well, when I worked at a third party tool, but also when I worked at a distributor and even now is, how should I do X? Well, there are 15 correct ways to do x. I need more data about what it is you're trying to achieve, what problems you're trying to solve. I think the place I see it happen the most is when, Folks come from industry I love you, industry people. I was one of you for a [00:08:00] while. I learn a lot from you. Mostly I'm in awe because you're doing things at a scale that I haven't reached. Uh, the guy from Indeed whose commitments bill is larger than all of my clients Cloud Bill altogether, like bow down and worship you, right?

But, the complexity of, for example, a lot of what I'm doing now is involving government systems. Right. So these, these decisions and these thought processes, you were talking to downstream impacts when your ERP goes down. talk about what happens if the live system for your Medicare eligibility during open enrollment in your state goes down. Um, sure they're both very big systems that impact a lot of people. but it's very different when it's people's healthcare actually at stake versus a can I log in and see that? My, uh, dental plan from my employer that I know what my status of insurance is, right? [00:09:00] Like they're all very different. Um, And so quickly being able to understand that you're talking to people, even at FinOps X, right? You know, we go there, we see a lot of our industry friends there, and at reinvent and at the happy hours, et cetera. We're technically all doing the same job. We, we are not doing the same job. And I think learning from the folks, um, You know, who are doing things at a scale you could never imagine, and honestly, that includes going down, right a lot of people look up to the apples, the indeeds, the EA games, the new banks, right? Sometimes you should be looking down at those smaller companies because they've got of the automation and all of the, the thought behind it, right? And they've had to learn at a granularity that large scale [00:10:00] never has to, large scale is everything but granularity.

Taylor Houck: That's such an interesting observation and, and Marit when you think about, You know, the fact, like you said, you go to FinOps X, you're meeting with all these FinOps people and you feel like you have the same job, but really everyone is. Doing something completely different. But what are the parallels?

Like what are the things that are just true to FinOps that everyone is doing or should be doing?

Marit Hughes: Communication and alignment. I think a lot of communication and alignment is actually education. Um, but if you don't know and aren't working together on your business goals, um, your strategic. Initiatives. How do You know what data to even bring to the decision as a FinOps practitioner? Right? So, um, if you are a FinOps practitioner, and the only thing you're really aware of is your cloud data and what all of your [00:11:00] dashboarding says, but you don't know anything about what are our business goals, what are we trying to build, um, what are the things that are super important?

So one of my projects right now is. a migration that had a pilot go live back in the fall. And I have my real go live coming in a couple of weeks every piece of scheduling I was on has now been stopped because I have a business goal of having this government system go live. At the end of January, if I just looked at my data, my cloud bill went up 20% for no reason. was a reason, right? and I think that that's something that regardless, even if you're at a third party vendor and you're talking to your customers, you need to understand what their goals are. Because as that vendor, um, as a consultant, as any of the other kind of ancillary roles around the FinOps team. Everything is driven or should be [00:12:00] driven off of their business goals and what can be achieved.

Taylor Houck: What do you think? Is the best way for people in FinOps to get more tuned into those business goals? I think it's easy for people to get caught up in the data and the optimizations and the forecasting, but how, how would you recommend someone get more tuned into the business goals so they can align what they are doing towards the greater goals of the organization?

Marit Hughes: Yeah, absolutely. So if you are at any sort of product company. Go to your, go to some dev standups, go to the monthly product review sessions, right, where the product team is showing, hey, here's what's coming. some of those roadmap sessions. Um, because that is a tells you a lot about where they're going. If you are at a more standard enterprise where your product teams are more like, how do we enhance this feature of our. [00:13:00] Investment app for our clients, but you're primarily a financial, so think like a bank or an investment company. sure. You know, the roadmap on like what new bond is going to be offered is perhaps not directly relevant, but um, a lot of the information on new analytics that are being put in place, the new features that might come into the user app.

Those things are. Part of those core decisions You know, if there's any of my public sector people here, right? We all know our missions are to serve the taxpayer, but let's, You know, break it down a, a level on that granularity and really think about is our mission to have this system go live, to meet this regulatory requirement, to be in the cloud by a certain date? And what. Is the schedule to get from on-prem to there, right? All of those things. And if you, you talked [00:14:00] about being focused on your forecast, all of those things inform your forecast in a way that somebody might not actively tell you about. Right? I was in a meeting the other day that was like a, a weekly client briefing. attend maybe half of those meetings. Um, and in this one, somebody said the word snowflake. And they were not talking about the weather, and I immediately raised the flag. Okay? Not immediately, as soon as the meeting was over, because I'm trying to interrupt that client facing meeting okay, wait a minute.

If we're planning on Snowflake, there's gonna be cloud costs for that. Oh, it's a SaaS implementation. We're not gonna be hosting it. That data's gotta get there somehow. Right, and no one thought to bring that to the FinOps team because it was a SaaS implementation, right? So in those meetings as much as you can.

You don't have to attend every one, but if you attend some of them and start flagging [00:15:00] things, more people become aware of the things you need to be told about, right? So you create opportunities for that stakeholder en engagement and matrix.

Taylor Houck: Yeah, you're absolutely right. And, and part of the, the interesting thing here is that a lot of times, people outside of FinOps don't recognize the cost impacts of the decisions that they're making as they're making them. And that's why having someone from FinOps in the room to flag that. Or raise to them?

What are the, You know, implications from a cost perspective as to this decision that's made in a, You know, in a product meeting or somewhere else is, You know, it's important to, to be there. I.

Marit Hughes: Yeah. Uh, honestly, what came to mind when you said that was the Hamilton song in the room where it happened? Um, but, um, yeah, the absolutely right. Being, being in there to, [00:16:00] to give that information. One thing I do say a lot to my teams is I don't approve your direction. I tell you the cost of your

decision,

Taylor Houck: That's it.

Marit Hughes: Right? Um, and sometimes you have to take the more expensive option. Because you have a compliance requirement, you have a regulatory requirement, you have a performance requirement says, yep, this one does need five nines multi-region Dr. That is made four maintained four times a day. As long as they're not making every decision that way out of laziness

Taylor Houck: Absolutely. Yep.

Marit Hughes: You know, that's great.

So, um, You know. Making sure that ops is seen is not a, um, a yes no, but as a, how can we help get there? And then if you said something was gonna cost X, how can we do our best to help you keep it at the price of X versus, You know, ever increasing cost [00:17:00] because forgot a lifecycle policy.

Taylor Houck: Yeah. And what, what came to mind as you were describing that is just another example from when I was with NBC, where there were teams that were making decisions on essentially if they were gonna spin up a new staging system to test some new features that they were working on. And essentially it would've required copying a database that was hosting over six terabytes of, of in memory data.

Right. Which is not an expensive machine and. They were pretty much the decision that they were facing was, are we gonna do it or not? And what they were not considering is the ability to, that, like, essentially with the usage-based pricing, you could spin this thing up when you needed it and turn it off when you didn't.

Again, these are people that were, You know, from the on-prem world. So they were, even if they were considering costs. And did the math out the break even point or the, the, the decision. Would've been based on a 24 7, 365 model versus, uh, [00:18:00] You know, a scheduled model.

Marit Hughes: if we need it for two weeks, 24 by seven and then after that we can, we can go

down.

Taylor Houck: Exactly.

Marit Hughes: Um, and, and those come in all sorts of ways. Um, recently I've actually been doing of, um, people put into their emails one thing and they're using their terminology and you have the FinOps practitioner has a panic attack. Uh, not describing my week at all, uh, because I got an email that said we needed to double the size of prod, and I was like, Ooh, that's a lot of money. That is a lot of money. Uh, just kind of sent a panicky email like, Hey, sure we need to do this? That's gonna like do a 30% increase in our total cost line. Um, and it came back that, uh, the [00:19:00] person who wrote the email in their head was only thinking about three servers that needed to be doubled, but it was only in their head. But the email. Just said prod, right? sometimes what FinOps is flagging isn't the cost, it's the communication fail, right? Um, they were communicating just not as cleanly and concisely as we need, and sometimes numbers us people to have to be very precise with their language.

Taylor Houck: Absolutely Marit. Um, I wanna shift gears just a little bit and talk about a concept that you brought up to me at reinvent this year when you and I were chatting and that concept was how fear plays a role in FinOps practices.

Marit Hughes: Yes.

Taylor Houck: your thoughts as it relates to this.

Marit Hughes: Yeah, absolutely. So I've been obsessed with this topic for two to three years now. Um, my degree of [00:20:00] obsession varies, uh, but you've caught me right on my kick again, it really comes down to how many decisions are being made because people are afraid. Of, taking any sort of risk. So they're afraid that CPU over 60% is gonna lag their machine, so they double the size without ever having confirmed that perhaps. They don't get lagging until 82% they could just stay right where they're at. Um, some of the scheduling stuff, the amount of teams I encounter who don't wanna schedule things to turn off nights and weekends because somebody might need to work at 11:00 PM or on a Sunday afternoon. I'm like, there are startup scripts for that. could we? Shut it all down. And then if somebody needs it, here's a script in a repository that is called the Startup Script, and they just put in their instance name. Like, could we solve that? Um, a lot, You know, particularly on-prem folks coming into the [00:21:00] cloud, you were talking about your SAP folks who only think of big databases that please, dear God, don't turn them off.

They take two hours to come back online. Um, and they only think in a prod mindset and they're afraid. And I don't think it's like a conscious fear that they're shaking in their boots, keeping them up at night. Right. But they're not familiar with it. And they'd rather do the thing they're familiar with.

'cause let's be honest, most of us would. Right. We'd rather stay in our neighborhoods, ensure we wanna go to a town that we haven't been to, but we want a map and we want our cell phones and all of that stuff. Right? And for a lot of folks to shift into that cloud, cloud mindset or take risks with their deployments, feels a lot like being dropped with no map into the middle of wilderness for them, and it causes them to make decisions. That are not for the betterment of their solution or their monthly bill.

Taylor Houck: [00:22:00] How do you help people adjust and get out of that mindset? Maybe if they don't even recognize that they're in it.

Marit Hughes: Yeah, I think that's a lot of what the FinOps team's job is, whether that FinOps team is one person, that it's 20% of their job, or whether the FinOps team is a team of 15. Right is that education, um, is offering them the thing you learned while you were talking to somebody in the hallway at FinOps X, or maybe it was a session at, um, one of the FinOps community days or one of the CSP conferences. Right. Hey, I heard, You know, Marit and Taylor talking on the Point five podcast about having launch scripts. So if somebody needs to work on a weekend, could we just create a launch script? Now we can turn things off because they're not thinking of that. And I think that's our job a lot as FinOps practitioners, is to take all of the lessons either we've learned ourselves or that we see our colleagues learning and sharing in the slack [00:23:00] in the 0.5 community.

Um, The cloud efficiency hub, right? There are so many lessons learned there that as FinOps practitioners we know about, we can access and we need to be helping those teams, um, think through those options because they don't know what those options are, right? Um, especially with the CPU utilization thing, everybody knows that they should do performance testing and know when they're gonna start seeing performance lag, but they're trying to move at the name of speed. Right. So if you can say, Hey, here's a script that somebody shared for testing this, or, Hey, I know that we already have an internal script for this.

Can we do some metrics before we upsize? Or alternatively, if we need to move fast 'cause go live is coming. Can we all commit to watching that performance data for the next 45 days and the week before the [00:24:00] second patch window? 'cause generally after go live, there's like shake out, nobody touch anything.

It's not a bug, let's just keep it running. Um, that we will revisit this and then we track and help them revisit that. Um, because a lot of it is they're just trying to move fast and. Get to the next thing, how can we help them realize that there are multiple roads to

Rome.

Taylor Houck: Yeah, it's, it's almost like when there's a lack of data telling you the right answer, people are gonna fall back to their gut feeling and it's emotion. Right.

Marit Hughes: Yep.

Taylor Houck: when a decision is made based on emotion, it may not be the optimal one that you would make if you had all the data.

Marit Hughes: Yeah, for sure. And I think sometimes it, they, it doesn't feel emotional at the time. It feels very obvious. We have a time constraint. I know how it works this way. I don't have to go research anything. I don't have to validate the script you found on somebody's [00:25:00] GitHub, You know, any of that. Um, and they're just, they're trying to move forward.

But, You know, it's funny because I see sometimes conversations about commitments. How do I, I don't wanna overbuy my, my compute savings plans, so they don't buy any at all.

Taylor Houck: Are you really seeing people that aren't buying any at all?

Marit Hughes: Yes.

Yes. Right.

Taylor Houck: They're longstanding entities that aren't going away.

Marit Hughes: correct, um, some of them, now to be clear, in government we have like a lot of procurement rules and laws about timing. Of like when you make those purchases. I'm not talking about that. talking about like regular companies, regular industry, where they hit decision paralysis is one of them. Like how much, um, are they, they too many options. Um, but a lot of them, they, the. Even though compute savings plans exist and now DB SPS [00:26:00] exist, they still mentally pro process them as if they are reserved instances that are restricted to instance type region size os though you can spend all day telling them compute savings plans don't work that way. Right. Um, it, it's so ingrained in them to not that decision. More and more are buying some. But funnily enough, those are almost always the people who are in fact, still buying our eyes and haven't switched to savings

plans.

Taylor Houck: do you think? That those that are not purchasing commitments, that feel confident that they will remain in the cloud, do you think it's because they know that their usage is inefficient and they're not ready to commit to it?

Marit Hughes: Yep. Or we've got

plans.

Taylor Houck: Yeah.

Marit Hughes: have a, I have a team that I've, I've been working with on and off. They kind of periodically bat phone, Hey, can [00:27:00] you come take a look at things? And they were gonna change, um, aloe, configure, they were changing something. Some configuration be done in the next three months. We'll be done in the next three months. I've been hearing that for 16 months now. Hasn't changed. Um, right. So they didn't wanna make the commitment. I was like, but if you buy compute savings plans, and they're like, well, we're gonna go to serverless. Great. It will, it will cover that. Maybe don't buy the full amount, but buy just that dollar an hour.

Get yourself. Something. but it is a lot of like, we're going to, we're much like before we started recording, uh, Taylor and I were talking about the mess that is my office, right? Like, I'm gonna get to it, I'm gonna get to it, I'm gonna get to it. And I think a lot of times, if it's not decision paralysis or that fear, similar to like our eyes, they don't look at like, Hey, my breakeven point is 7.2 months. Do I really, truly [00:28:00] believe I'm gonna get this commitment done or this ch architectural change done in the next seven months? Like, let's be honest with ourselves. Um, and much like have I been honest with myself about cleaning out that corner of my office? No, I have not been. Right? And I think a lot of people aren't when it comes to those things. The flip side are people who are so aggressive 'cause they wanna save every penny. And then they end up in a situation where now they're over committed because they did all of these things and then they're mad that Amazon won't refund it. Well, okay, you probably still broke even on it. Just let it, you're the math.

Math, right? So, um, You know how, how to balance. That's always a challenge.

Taylor Houck: Marit. Where do you think fear, or how do you think fear plays into optimization? Strategies, usage, optimization.

Marit Hughes: many times have you gone to tell people that they have old [00:29:00] snapshots and they don't wanna delete them in case somebody needs them, orphaned ones. Particularly orphaned one sitting in prod from six years ago that no one is going to return to Um, and it's not like a data retention requirement, um, snapshot.

It's, it is literally like the golden image that they use to build prod with. They don't want to, because what if. What about all of the times that, uh, You know, you're like, Hey, you have 42 stopped EC2 instances with your EBS volume, still incurring costs, your IP address is still incurring costs. Can we rid of them?

And they're like, no, somebody might need them. They literally haven't been turned on for 18 months. Can we snap and delete them? Oh, I don't know. Um, my favorite are the VMs with like 2% CPU, that nobody knows who owns them and what's on them, [00:30:00] particularly in government because there are sometimes systems that they're very low usage, but they have to stay there. And it's been forgotten about in kind of the weeds of time, and it's kind of a legitimate fear. If we stop this thing and this job doesn't run, what are the downstream ramifications? Scream tests are fun until it's a production outage on a mission critical system. Um, so I think in all of those cases, like somebody's afraid that they, they're, they're gonna need it. They're gonna need it. And if they delete it, then what?

Taylor Houck: It begs the question then Marit. How do you, how do you ensure then that they don't need it? How do you ensure them that taking this, this server and decommissioning it won't cause that outage? What, what do you do to, to calm that fear?

Marit Hughes: Uh, well, so as a consultant, I can officially never insure anything. Um, [00:31:00] we always have to caveat we are required to. Um, but generally speaking, I offer them ways around it. Like, can we snapshot it? And hold that snapshot for 45 days. That way if somebody does come screaming, the restoration process is pretty quick. If the instance has been running, can we just agree to stop it? We don't even have to snap it. Let's just stop it for 45, 60 days and see if anybody comes screaming. If they come screaming. It's a really quick restore. Right When it's old orphan snapshots, sometimes it's a math equation. Literally like, Hey, is holding on to all of these snapshots worth the risk that you might need something that is untagged that you'll never be able to locate again, versus, and this is not an exaggeration, a 20% reduction in my monthly bill because we had so many snapshots worth it.

Like what would be your process to get the right snapshot back? [00:32:00] And generally speaking, people slowly start to come out of that just in case and start thinking it through.

Taylor Houck: there's also other. Alternative solutions as well, right? Like if you, if you think you might need the data, well, you don't need to keep it as a snapshot. You could actually take that data and put it into S3, you could dump it into a glacier tier, and then you're paying practically nothing compared to holding it as a snapshot in EBS itself.

Marit Hughes: hot snapshot, you do have to be careful of. 'cause when you move them, then incrementals can become fulls and there's all sorts of like weird trade-offs there. But, um, the thing that just came to mind while I was talking about that is we discovered at one point, um, uh, DB was being backed up three separate ways. Um. Re it was a DB hosted on vm, so regular snapshot R man export to S3 and EFS. And I've never felt more empathy than being on a group call with like six people being like, we gotta, we [00:33:00] gotta, we do not need three backups of this command line delete of all of the EFS attached to production databases.

Like we were all just holding our breath. 'cause we knew it was triple redundant or double redundant. And yet there was still that moment of panic in all of us about what's gonna happen when we delete it. Right? Um, so it's not that as FinOps practitioners, we don't also feel the fear. Right. But the data had shown us that we didn't need it. Um, so, You know, there are options as you mentioned of how you handle that and how you educate people through that and sometimes walking in, I didn't know that that's what our problem was. I was really focused on our EC2 snapshot problem. I'm assuming that if the engineers have an EFS on there, they need one. It was clearly being used. Right. I just had no idea it was part of the three. And that's why [00:34:00] you have to work with your teams to, to figure it all out.

Taylor Houck: Such an interesting and, and, and it happens all the time. And what's scary, you're talking about fear, is how many of these potential inefficiencies could be out there that you're not aware of, especially as you get into. More scale, larger system, hundreds if not thousands of accounts with millions of resources across many services.

You get into, you mentioned other providers, snowflake coming on, Databricks is coming on. People are experimenting with these new AI tools, consuming, You know, tokens and, and streaming. You know, just adding more usage-based pricing models into the cost structure of their engineering organizations. How do you solve that?

Marit Hughes: solve. That's a, that's a bold ask. I don't know that there is a solve, 'cause that sounds like a permanent fix. know, way back at the beginning we talked about kind of what was it [00:35:00] like back in the day? Um. And it, the one thing that is consistent from, what did I say, 2012, so almost 14 years and now, is that it's changing faster than you can keep up with.

Right. And that's probably, I think the FinOps practitioner's biggest fear is what's the next thing I don't know about? Right. And how, how am I gonna figure that thing out? so I. When the scope is getting larger or more things are coming in and you're being asked to forecast things you don't know about, right?

I don't know how much data we're gonna be sending out of our system and into a SaaS version of Snowflake. don't know how many AI tokens we're gonna use on the next thing. do you forecast for that? It's gonna be a little bit of best guess and like for the best and not letting fear cause me to triple my estimate. Right. Like put some reasonable thought into it and some [00:36:00] best guesses, document the crap out of them and move on. Right? Don't let analysis paralysis become the FinOps practitioner's primary source of fear.

Taylor Houck: so Marit. One of the things that you just touched on is, is how You know what FinOps is and what it looks like is changing rapidly and you've seen it, You know, from when you first started in the industry. Um, to now, and it feels like the change is only accelerating.

If you look forward, say three to five years, what do you think is gonna be different about FinOps?

Marit Hughes: I think the last couple of years has been a heavy focus on data. Getting the data, understanding the data. I think the next few years is really gonna be, You know, we talk a lot about shift left, meaning shift left into the engineering layer. a lot of it is gonna be shift left into the planning. Um, and a lot of really understanding implications of the data versus, Hey, I've got a pretty dashboard with the data, [00:37:00] right? because more and more automation's gonna take care of the data itself. Right. Um, AI is gonna take care of a lot of that, uh, low hanging fruit, or it should, um, if your teams haven't become so fearful that they've put barricades around the low hanging fruit, fruit, deer, fencing, nobody can get in here. and I think that's where having that FinOps practitioner really starting to focus on those communication planning areas and starting that engagement. Way, way to the left. You know, not to be the person who quotes herself, right. If security is job zero, FinOps should be job 0.5. That's really what I'm talking about. Right. How do you get in at those planning stages to make sure. Everything from that initial thing is setting you up for future success, your tagging standards are, what your priorities are, what your [00:38:00] scheduling's gonna be. If you can start thinking about that earlier in the planning process. And I do mean planning. Once you're building, you're building and you can't ever get anybody to go back. Right? Or it's a big fight. As, as I know you've experienced Taylor. Um, think a lot of ops time is gonna go on that. And not on, um, the data itself because so much has come so far on the data

side.

Taylor Houck: I see FinOps getting so intertwined with cloud architecture decisions, especially

with new builds, where you're gonna see the personas kind of come together in, in a lot of ways. And I mean, as you touched on with the, the data, there's. So much data that we have access to and AI is only going to improve the ability for an individual to make sense of it and draw the insights that they're looking for.

So I see less time being focused on massaging and playing around and making the data into a place where you can make sense of it and [00:39:00] more of the actual making sense and drawing those conclusions.

Marit Hughes: And I, I think that's gonna really, um, heavily influence where forecasting ends up going as well. Um, because how can you start taking the things from meetings and integrating that into your forecast, and then how will AI start to, to play with your forecasting projections, because now it could have meet data from your meetings.

And mean, thinking kind. out in space on that. Like, oh, well we know Bob always misses his deadlines by three months. So like your forecast is automatically being like Bob said June, guess that's September.

Taylor Houck: Oh my gosh.

Marit Hughes: and I know that's a long way in the future. Right. But that's a lot of what as practitioners we do now when we're forecasting is being like. Bob aint never made a deadline in his life. Uh, let's shift that. Right? [00:40:00] but you never know, right? Uh, the world's changing so fast, we could, none of us know what we're doing. I can tell you in 2012, even when I did my first AWS invoice that I was not expecting that this has all become. had become within 10 years.

Right? And now we're pushing the 15 year mark. So where are we gonna be five, 10 years from now? Anybody's guess.

Taylor Houck: It's gonna be awesome to see it all play out Marit, and I know that, uh, you're gonna be there with us watching it and making it happen. Um, thank you so much for coming on the show. This has been amazing. Um, before we wrap Marit, You know, we talked all about ops in the professional world, but You know, that's, believe it or not, that's not the only thing that, that we do.

Um, I know you've got a couple, a couple dogs. Are they in there? They're with you when we're recording this.

Marit Hughes: Yes they are. I have, um, the nickname, her nickname is the Void. As you can tell, you can't actually see [00:41:00] her, uh, a black standard poodle. Uh, and then over here, and if I say his name, he will probably get up. Um, my white standard poodle Eddie that I rescued, uh, about 18

months ago.

Taylor Houck: This is how I know that you have well-behaved dogs because you're recording a podcast and trusting them to be there with you. Most of the time, my dog, Myla is here with me in in my office, but when I come on to record, she's gotta be away from the situation to avoid the pickup.

Marit Hughes: that also knowing that when we did our pre-call that my black female jumped straight into my lap, like she was not 55 pounds just placed herself directly in front of the camera. So, uh, thankfully it's just afternoon nap time. In another hour, she'll start trying to make it very clear that dinner should be an hour early.

Taylor Houck: That's amazing. And you got horses too, do you? You like horses.

Marit Hughes: I don't have a horse currently. I lost my horse back in the, uh, spring of 25. [00:42:00]

Taylor Houck: Sorry to hear that.

Marit Hughes: a friend's horse, uh, as she's recovering from some medical issues. So I'm at least getting some time in the saddle. and like a tro an ops practitioner on somebody else's dime,

Taylor Houck: Where? Where do you ride? I didn't.

Marit Hughes: Yeah, there. Uh, so for listeners, I live in the suburbs of DC um, and I go about another half an hour, 40 minutes out and ride at a

farm There.

Taylor Houck: That's amazing. Um, I always love hearing what people are into, um, outside of, outside of work, and I've got family that likes horses and I, I have some horse experience. I'd say the horse and I, we didn't quite get a lotg. I don't know if the horse liked my energy, although I will say my horse's name was Rebel, And so they, they put me on a horse named Rebel and, and.

I, I didn't have all the control in the world. I've got more control of my, my cloud spend than I had of Rebel. I'll tell you that much.

Marit Hughes: Um, well, it is a 1200 pound flight animal. Um, so honestly, it's like your engineering manager when he is heard, prod, has gone down. That [00:43:00] is basically most horses. Uh, fear is what drives them, and you just have to be there with the cookies and, uh, the soothing words to, to bring them back to

Taylor Houck: Amazing. Well, it's been another amazing episode of FinOps in Action Once again, thank you so much Mary Hughes for uh, joining me on the show. And until next time.

Outro: That wraps up another episode of Fit Ops in Action. Thank you for joining. For show notes and more, please visit fit ops in action.com. This show is brought to you by 0.5, empowering teams to optimize cloud costs with deep detection remediation tools that actually drive action.