AI is making software easier and cheaper to build, but why are SaaS costs still rising?
In this episode of The DevOps Sauna, Pinja Kujala and Henri Hämäläinen explore how AI is reshaping SaaS spending, software pricing models, and IT budgets. From build vs. buy decisions to usage-based pricing and growing competition, discover why traditional SaaS spending could soon decline and what businesses should do to prepare.
Speakers
Henri Hämäläinen
Chief Product and Partner Officer
Henri brings extensive C-level leadership to the software business. As the Chief Product Officer at Eficode, he leads global strategy and connects technology with concrete business value. Recognized as an industry thought leader, he created the AI maturity model and the three-portfolio model. Drawing on his background in coaching winning product organizations, Henri implements straightforward processes and structures to help companies build better software.
Pinja Kujala
Team Lead | Advisory + Atlassian
Pinja helps organizations connect strategy, people, and technology to build better software organizations. Driven by curiosity, she brings a broad perspective across the DevOps landscape, helping leaders navigate change and turn complexity into measurable business value.
Transcript
[Henri] (0:03 - 0:09)
In reality, the data is still quite easy to get out from every tool, quite easy.
[Pinja] (0:12 - 1:39)
Welcome to the DevOps Sauna, the podcast where we deep dive into the world of DevOps, platform engineering, security, and more as we explore the future of development. Join us as we dive into the heart of DevOps, one story at a time. Whether you're a seasoned practitioner or only starting your DevOps journey, we're happy to welcome you into the DevOps Sauna.
Hello and welcome back to the DevOps Sauna. Here's one strange thing that is happening in software right now, and as we all know, because of AI, it has never been cheaper or faster to build new software. If we give a developer an AI assistant, an AI tool, they can create faster in a matter of not even days, but we're talking about minutes and hours, what actually used to take one team months previously.
And if we look at the companies and what they're actually buying and paying for software nowadays, the cost is going up. And I think the new spend has entered the SaaS spend category, and that is AI, but the budget is still the same. And there is a big part of that growth in software spending coming this year from vendors, because they are simply charging more for what customers are already having.
And today, this is our topic. So let's talk about the future of SaaS spend, and I'm very pleased to welcome to discuss this topic with me today, our very own Chief Product and Partner Officer Henkka Hämäläinen. Hey Henkka, welcome.
[Henri] (1:40 - 1:43)
Thank you. Super nice to be back here with you again.
[Pinja] (1:43 - 2:04)
It's good to have you back. Previously with you, we have talked about where is the value being created now, and one of your current predictions that I saw you write about on LinkedIn a couple of days ago is that within the coming years, the average amount that companies are spending on SaaS will start to go down. So where is this prediction coming from, and what is behind this thinking?
[Henri] (2:05 - 4:13)
Yeah, thanks. It's almost like the reverse of how it normally goes, because normally budgets shrink, and then you have to really get things done cheaper. But I actually believe that this time there has been this discussion about Jevons paradox that like, hey, the need will always increase when the price of the delivery actually goes down, so that the pricing would be like remaining the same.
But I do believe now seeing so many things happening in the field that like the efficiency gains in the software world have gone up so quickly that it starts to kind of create competition to the world that kind of way. Just like when there's so much competition in some area, the prices normally start to go down, because then the element of competition always becomes one of the things to become the price. So I believe that this is not the normal way of how the prices will go down because the budgets will go down, but actually because the competition in the SaaS world will be so enormous that it has to kind of affect the prices also.
And if you're asking the exact moment when this happened was that like, it was some time ago, like a few months ago, it's really hard to think in the AI world like when stuff happened, because like a few weeks old news are like old news already. But it was the moment when the whole Adobe suite was basically copied in using one of the new, at least at that time, the new models and released as an open source version. So of course, that's not something that we want to happen to any of the companies that the IPR gets copied.
That's not the point. But the point is like how quickly with the very limited resources still stuff like that can be done. Stuff like companies built for tens of years with hundreds of people, and then with one AI, like one person effectively using AI could actually do this thing.
[Pinja] (4:13 - 4:37)
Yeah. And this is something you and I discussed in a previous podcast episode that basically if somebody wanted to set up the software that could support maybe a financial institution, that could be done in a single weekend nowadays. Of course, you cannot run a bank after a single weekend, but the supporting systems can be up.
And as you say, the whole Adobe suite can be copied like that. It is one of the biggest SaaS companies in the world.
[Henri] (4:37 - 5:21)
Exactly. There's one other project I can tell more about later, but like where I've been seeing from the site that one person with a really good AI framework can actually matter of days recreate stuff or create stuff that has taken like 20 people a few years. So of course, this is part of the problem that like this isn't true when there's a lot of legacy, but this is true for the new competitors to the field.
And this is exactly to the point it's like now it's so easy to actually create competition to the existing SaaS in the world that has to affect the prices at some point. That's where my prediction actually comes from.
[Pinja] (5:22 - 5:59)
Yeah. And if we think about the budget that now has to be shared with AI, because that's the spend on AI is going up right now. There's a lot of pressure.
Companies are not only, of course, they want to become more effective. They want to be more efficient in what they're doing. There is the need to match the competition, but there is also the FOMO effect.
In this podcast previously, we've said many times, we've tried to debunk some myths. So we've been talking about DevOps dying? Is DevOps dead?
Is Agile dead? Now, Henkka, are we saying that SaaS is dying as well? So can we make any assumptions?
What does it say that there is still a place for SaaS in the software tooling chain?
[Henri] (6:00 - 7:14)
Yeah, I think we're back to some of the details later. But I would say it's almost the opposite. For sure, there will be a lot of SaaS going forward, and there's even more products in every field.
But the whole point is that now the competition increases so much more. Of course, it depends a lot on what you think about this field. If we think about the core processes, some ERP processes, of course, there's not going to be an easy competition for those.
But every company has tens and tens of SaaS tools for stuff that will have a lot of competition that can be replaced. And it's almost like, maybe this is a bad metaphor, but thinking about changing a car frequently, I think that's the type of world we will get into with some of the tools that you just should do every three years or five years in order because you just get the latest stuff working. So it might be that this is the reality, and it might finally become true what the whole product management people have been saying for 20, 30 years, that you have to actually kill your darling.
So you have to kind of yourself create the next one which kills your current product. And now it's easier than ever.
[Pinja] (7:14 - 7:42)
The product management people will be happy to hear that we finally say that they were right. I think we have said that previously as well. But I guess the one thing we can say that if we think about it from the budget perspective, and as mentioned before, there is a need to now share this budget, the money itself might not be increasing internally in an organization, but where it is being spent, and maybe the direct budgeting to SaaS might be shrinking.
Is that what we might be seeing?
[Henri] (7:43 - 8:33)
Yeah, of course, it depends how you count it. But if we really count the AI spent separately, I'm almost certain that every company will have to save from some of the existing tooling and move that more towards the AI spend. If you kind of want to count that towards the SaaS spend, then it's not going to shrink.
But if we really try to split the pure using some of the models more directly, like with Claude or Gemini or GPT or OpenAI stuff, and if you think that's spent separately, then I'm actually very, very sure that the spend towards the normal tooling, even without AI included, it will keep on going down.
[Pinja] (8:33 - 9:14)
Yeah. And as I say, the SaaS products, especially the legacy ones, they're so heavily integrated in organizations. It's not that we can just go and change our ERP all of a sudden or customer management systems.
But everything else, whatever is loosely coupled in a tool chain, in an architecture, might actually see this kind of change. Maybe not every year. We're not talking about a lease car deal in this context, but maybe getting the newer model and something that works better with the whole rest of the tool chain that the organization has.
But maybe there is also something that might happen with putting pressure on the pricing points as well.
[Henri] (9:14 - 10:17)
Yeah. The whole pricing ideology might be affected with this one. It's almost like the heart, this is debunking the myth again, but at least to my understanding, many of the current tool vendors get a lot of their profit out of the non-usage.
There's a lot of seats that are not actually used and customers keep on paying on those. And then if we go more towards this paper usage type of thing, then that will already affect the whole pricing and cost you too many of the things. But I think the obvious thing in any case is that companies will need to find somewhere the money to be spent on AI.
And I think the one question that we didn't really cover that much is about the whole, should we then try to do stuff directly with AI? Or you did ask about the dying of SaaS, but I think we still diverted a bit from that one.
[Pinja] (10:18 - 11:00)
We diverted a little bit. Let's go to that now, because this is the age old question that has been very relevant to many organizations already before AI and AI tooling, which is the build versus buy. And now with AI, it is an even bigger question, especially if we don't use our legacy tooling, we don't touch our ERPs.
But before we go to that, let's touch upon the overall cost that SaaS is bringing. Because if we were to build this ourselves, if we gave a couple of developers a couple of weeks with our product management people, with architects and everybody else, what are we not getting with the build? So where's the SaaS side coming in?
[Henri] (11:00 - 12:56)
And I think that's in the very heart of the question when you analyze the whole build and buy discussion is that what you get from someone else actually running the applications is the maintenance, the security and compliance and the SLAs and overall the trust to all of the latest security updates and everything else have been put into a place. And I think that's in the core of the build versus buy question that, as an example, that if you need to have a dashboard built for your management team or something like that, no one really cares whether there is a 24/7 maintenance for that dashboard, right? So I think stuff like this, of course, you should build with AI, you can actually do a super good dashboard to match your needs at a very rapid pace.
And I think there's many other things that don't have the same security and compliance and maintenance requirements. But then when you're talking about anything that actually then requires our customer facing or like really in the core of your processes, then if you're going to build it, now we go back to the whole DevOps thing, you have to have like a super good DevOps practices inside the house in order to be able to maintain the whole stuff. And I think you've had it on the podcast already, but like how quickly the vulnerabilities are exploited nowadays, it's matter of like minutes or 10s of minutes rather than 10s of days, what it used to be.
So it's something that if someone specializes in, it's much easier than doing it yourself.
[Pinja] (12:56 - 13:16)
It is. And it might be in the core, it might be some of the very compliance or regulated heavy things that an organization needs to run, for example, an HR platform. It is usually that support that is also what we pay for and that expertise that comes from the SaaS companies that are providing those.
[Henri] (13:17 - 14:10)
And maybe that's a good thing, this goes back to like the, at least to my ideology, is why you should also like to think about changing the system once in a while is that normally the systems are like representations of the thinking of that time they were built. If you get what I mean is that like, if you take the HR platform from the 80s, it has the HR ideology of the 80s. If you take it from the 2020s, it has the ideology of the 2020s.
And the same goes with everything, even though they try to catch up, but like if the base has been made, like there ain't that many examples that something was built like 30 years ago and it was able to implement the latest way of working to the system, it's easier to change the system sometimes.
[Pinja] (14:10 - 14:49)
It is easier to scrap completely and get a new version. So many of us who work with legacy software, I'm not saying we are at Eficode, but many organizations that we see, it is very hard to change. I think one example we have used previously has been Atlassian's JSM.
So basically service management tools versus some of the competitors. So if it's integrated, changing from one SaaS to another is not just a snap of our fingers. So it's been there and implementing the ways of working around an old tool might be extremely hard because of how the tool has been built back in the day.
[Henri] (14:49 - 15:06)
Exactly. And it's also, we are diverting a bit, but still want to say this, but like the worst thing you can ever do is when you take a new tool into use is just to say, let's just build the same stuff to the new tool that we used to have. Now that's always a possibility to actually improve.
[Pinja] (15:06 - 15:41)
It is an option. It is a choice, but I wouldn't recommend it definitely to go in that direction. But hey, let's go back to our pricing and the cost discussion here.
So if we think of what will likely happen to the price, this is all of course predictions. So now we're looking at, let's say pricing per seat. There is an option to go more towards usage-based pricing and then outcome-based pricing.
What are the differences then and what are the risks or maybe the aspects to take into consideration as an organization if you have a SaaS product more into usage-based pricing model?
[Henri] (15:42 - 17:03)
Yeah, I think almost like we could cover these all separately, but the seat-based pricing, I think there the big risk is really around just our customers willing to accept the price, but it's easy for everybody to understand still and easy to compare and easy to compare between tools and vendors. So I think from that end, I think that will always be there, but there is the heaviest competition going to be. That's also easiest to do the comparison and say that we are 30% cheaper with the seat-based price type of an approach.
So that's in some sense the clearest, but then the usage-based pricing, I think that's like for the CFOs, that's a horrific idea without like, hey, we have a tool where we will get a bill based on how much we've used. As a CFO, you never want to have it. You want to kind of as much as possible to have the knowledge of what the bill will be.
And that's like if we mention some of our partners, for example, GitLab went to this flex pricing where you still kind of know what it is, but it's flexible throughout the period. And I think that same is something that others will also try to implement and have already implemented.
[Pinja] (17:04 - 17:32)
And that is something where the risk is shifting more towards the buyer, as you say, when we're using the usage-based pricing. And there's a study that showed that 60% of IT executives consider the variable. This was now on variable AI costs, but variable AI costs were hard to forecast because of the usage-based pricing.
And it's not just AI tools, but SaaS overall, as you say, the CFO's horror story when they don't have the prediction of what the spend is going to be like. Yeah.
[Henri] (17:32 - 19:09)
And AI is almost like a different beast because it's not only about usage, but it's about usage of the same tools in different ways. For example, we have this tool, the AI Insights tool that we've actually created at Eficode, which analyzes the LLM usage across the whole software development lifecycle and can kind of showcase how much spent there was and from there onwards. But what brought it up is that we did an analysis of our R&D leader, Sanjin, who actually did an analysis of 500 of his own commits and 140 of those, if I remember by heart, he thinks he actually used a too good model.
So there is the other aspect of like, hey, you need to know already, like, hey, how you shouldn't always be using the latest model because you don't know, like need it for everything. So it's like, there's a different aspect of the whole usage-based thing. And we've seen this in many other things.
Sorry for us to tell another example of creating with AI, these offers are based on templates. We used to do it fully by AI when we realized that it's actually better to create the template by AI and then use Python script to actually fill in the stuff and the costs went heavily down. So there's a lot of room to wiggle on that one.
[Pinja] (19:09 - 19:38)
And this is, as you say, AI costs are a different beast. If I need to use our AI tools to spar me in thinking about what can I, what obscure question can I ask my team in the next team meeting as an icebreaker, I'm not going to use the fastest model. I can wait.
Basically, this is a very, very simple example, but this is something we've been talking to our customers about the model routing as well, that then it would be more automatic and the cost would be more in control because there is the usage-based pricing model in use.
[Henri] (19:39 - 20:30)
Exactly. Yeah, exactly. But the usage-based model for service there to stay, and it also kind of like, if you go back to the SaaS world now for a second, I think that still at least then creates this difficulty to compare the prices.
You can't really, like, if you had a SaaS product A where your usage was something, then it's really hard to compare that with something else because you don't actually have the same usage in another tool that you could directly compare. So the price competition will be harder with the usage-based pricing. And then the whole outcome-based pricing, I think that's going to be, it always feels like a good option, but the difficulty is in defining the outcomes well enough.
[Pinja] (20:30 - 20:39)
It is the, when we're talking about something more obscure, like outcome, output is easy or easier, usually. I'm not saying it's always easy, but it is easier.
[Henri] (20:40 - 20:53)
Yes. But it's not still like you order a hamburger in Burger King so that you can kind of measure when it comes from there. It's not the same with the SaaS tool at least, but like, yeah.
[Pinja] (20:53 - 21:11)
No, it is not. But if we think about the SaaS providers and their pricing models and how they're offering that to their customers, who do you think and what kind of pricing model will help on the SaaS company side? Who are the most likely to be the players to win in this pricing game?
[Henri] (21:12 - 23:32)
I think that's a super good question. Like I discussed, the usage-based pricing will help to obfuscate the costs a bit, but then of course you can still at the end, just like see the total costs in a year and divide by users. So you get some kind of an ideology, but I still feel that maybe this comes more to my ideology than to any facts, but that the winners are the ones who are not trying to get every penny out of the customer, but almost the opposite, try to make every penny of the customer count.
So those who kind of help customers to say that like, hey, did you notice you have 50 licenses you don't actually use? So we could kind of ensure that if it's some other, then you could spend that to the AI features of us or something like that. So at least those are the ones who will gain trust in the market and they might not be that eager to be chased.
But overall, of course, the key thing is how deeply you are integrated into the daily operations and I guess this boils down a bit on the data. I don't want to go too deep on this because it's not always as important as it has been said. When the AI movement started, everyone was like that it will be all about the data.
But in reality, the data is still quite easy to get out from every tool, quite easy. But then on the other hand, it's like how much something is integrated into daily life. So if you have like 5,000 users of a software daily, then the data isn't the one that actually ties it to the organization, but it's the 5,000 users which you would need to retrain and everything.
So I think I do believe that there will be pricing pressure, but the most important thing, maybe that's dull to say, but it's still about how much value you bring to the daily operations of the organization.
[Pinja] (23:32 - 23:59)
I think that's a fair prediction, I would say. But if we now go to, let's say, a makeup scenario that we've now said what the pricing might be looking like and how the budget could be, would need to be split between the SaaS and the AI and our own build software. So when it comes time to actually make those tooling decisions, what needs to be taken into consideration in this day and age?
Where to start the evaluation and the thinking process?
[Henri] (24:00 - 26:33)
Yeah, I think the core thing to think about is where we already discussed how much there are these 24-7 security needs. I think that's one of the directions you need to look at, how much needs there are on this kind of maintenance front. But overall, I would almost be very brave now.
There's a lot of options, there's a lot of options coming every day. And as an example from our own world, let's not name anyone, but we used to have an HR system which was literally from the 80s or 90s, or at least it looked like it was. And then we changed it to a new one a while ago.
And it's amazing how much at the same time the price went down and everything was so much easier and better. And yes, it was a chase project. But in any case, like, hey, that's still paid back quite quickly.
And the same is with so many others. So I would kind of start looking from first, getting some good experiences on that, maybe giving another example. For example, on the contract management system side, we changed from one to another, which kind of with almost 2000 customers and a lot of contracts, it kind of sounded like a huge job.
But in the end, it wasn't that huge a job. And it's not that hard to learn to use a contract management system. So there is stuff like this that can easily like bring you more value and even more cheap.
And maybe I just want to say like, why we are talking about it, I think that's the core thing is like, we're not talking about it just that like, hey, we think this is a phenomenon, which is also interesting. But the core thing is that because of what's happening in the whole software business, organizations need to change their cost structures, like heavily changing the cost structures, like to be much more agile on where the costs can be put. And like, also overall be more open to change on all of the tooling.
And I believe that that's the reason it's important to keep options open. On top of that, there will also be cost savings on this front.
[Pinja] (26:34 - 26:39)
And there will be, as it wass discussed before, this budget is now being split also to AI.
[Henri] (26:39 - 26:40)
Exactly.
[Pinja] (26:40 - 27:02)
Which was not there a few years ago. So there is, again, we’ve be talking about business agility in other contexts as well. But business agility also comes into play here to divert those, make new decisions and try to see from the data what is the best for us.
Also in this, in this front. And I think on that note, we're going to end this discussion today. Henkka, thank you so much for joining me here today.
[Henri] (27:02 - 27:03)
Thank you. It was a pleasure.
[Pinja] (27:04 - 27:13)
And thank you everybody for tuning in. We'll see you in the sauna next time. We'll now tell you a little bit about who we are.
[Henri] (27:14 - 27:18)
Hey, I'm Henri Hämäläinen, Chief Product and Partner Officer at Eficode.
[Pinja] (27:19 - 27:34)
I'm Pinja Kujala. I specialize in agile and portfolio management topics at Eficode. Thanks for tuning in.
We'll catch you next time. And remember, if you like what you hear, please like, rate and subscribe on your favorite podcast platform. It means the world to us.
- AI
- Software Organization
Related podcasts
