Chicago Booth Review Podcast Will Salesforce and Adobe Survive the SaaS Apocalypse?

Marc Knez: I'm looking at my UI that is connecting me to my AI agent, and the AI agent is doing a bunch of things for me, which includes bouncing or connecting and using each of these tools that were apps that I have subscriptions to.

Hal Weitzman: Software as a Service has boomed in recent decades, but now it faces an existential threat from AI. Which companies will survive? Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking insights in a clear and straightforward way. I'm Hal Weitzman. Today, Chicago Booth's Marc Knez gives us his take on the SaaS apocalypse. Are companies really going to stop subscribing to software applications housed in the cloud to run their businesses? And what will they do instead? Marc Knez, welcome to the Chicago Booth Review Podcast.

Marc Knez: Thanks for having me.

Hal Weitzman: We're here to talk about the SaaS apocalypse. There's a SaaS apocalypse coming. Why are you particularly interested in this? You teach tech and strategy here at Chicago Booth. How did you get into thinking about this topic?

Marc Knez: Well, my area of focus and what I teach in the classroom is all about strategy in the context of technological change. And so as a topic, it's a forward-looking topic, which means as new technology emerges as the basis for innovation, and that starts potentially impacting an industry. The question then becomes, "Okay, how things are going to play out in the future?" And if you're an incumbent in an industry, an investor, or potentially a new venture considering entering an industry, you have to have some understanding of and viewpoint about how technology is going to lead to innovation and change and what that then means.

Hal Weitzman: So are you a disruption guy? Are you more interested in disruption?

Marc Knez: No, I'm not a disruption guy. I'm into understanding when disruption might occur and what it looks like. Any time new technology emerges, and then within a specific industry, we start seeing the potential for disruption, that's a sandbox for study. Currently, in the context of the massive potential disruptive change of AI, there's never been a bigger sandbox to study. So that's why it's an exciting time despite my age. Wish it would've happened 20 years ago. Fundamentally, it's about how does one ground one's thinking as they're thinking about what's going to be the impact of the technology. There's always two principles that I start with that I try to train my students with, which is first, when new technology emerges, that's the basis for innovation and products and services. There's going to be some re-optimization process. And that-

Hal Weitzman: Re-optimization meaning?

Marc Knez: Well, you're going to have, and we'll get to SaaS in a moment, but the idea is based on the new possibilities created by the new technology, there's going to be innovations and products and services. That's step one. But then, it's not just, "Okay, the product is better, or the service is better." The re-optimization can carry on from there to you are going to perhaps make changes to the overall business model that create and delivers the product or service, so you have business model changes. And once you start having business model changes, then you have changes to the structure of the industry. So all of this plays out over time, and so the first orientation or principle is, "Okay, we need to understand how this re-optimization process may play out." That's step one or part one.

Part two is economics always wins. It's just a matter of how long it takes, so that one expects that the product or service and business model that creates the most value will win ultimately. So ultimately, it'll be the winner. And the reason that's important is because these things take time and there's all kinds of inertia that slows down the process of us getting to what becomes the new conventional optimal solution. A lot of that inertia is created, first and foremost, by incumbents that aren't super excited about moving to this new world. Now, there are exceptions to that. And so, it's really understanding these two forces and using that as guidelines to think about what might happen next, not next month, not necessarily next year, the next three to five years. And so the so-called SaaS apocalypse is a great sandbox to think about this as it's going on.

And also, there's a lot of, I would call it in part confusion and in part just there's so many different perspectives on this. Having a clear-minded perspective is not easy, but it's what's necessary to really say, "Okay, what should we be doing? Where should we be going next? Which SaaS companies are in trouble, which aren't?" These sorts of things. That's where I come at it from.

Hal Weitzman: Okay, excellent. We'll get into the apocalypse. We'll delay the apocalypse for a few minutes because I want to just remind our listeners about SaaS, software as a service. Just give us a brief potted history. How did it arise? What problem was it solving? And why did it get so big to where we are now?

Marc Knez: Yeah. It's actually a useful starting point because it does get back to this re-optimization and economics always wins. Pre-SaaS, some of us are old enough to remember, back in the day if, let's use CRM software, customer relationship management software as example, because we'll want to talk about Salesforce. Pre-SalesForce.com, the dominant provider of CRM software was Siebel. I remember 25 years ago teaching a case on Siebel software, which was eventually purchased by Oracle. In the world of packaged software, if you purchased Siebel Saw or adopted Siebel CRM software, Siebel would send you packages full of CDs that had the software on it. And then, someone in IT or if it was a big implementation, you would hire someone like a firm like Accenture to come in and implement. You weren't just sticking the CD into the PC because it would have to get configured to the specifications that the particular enterprise wanted. There was this implementation phase and often the implementation was more expensive than the software itself.

Now, once you purchased the CDs and loaded it up, we're done. You own it, you use it. All the data stays on your servers, all that. If there's an update to the software, we have to go through that whole process again. You have to get the CDs load it up again and that might happen every one to two years. And so, a company like Siebel would have this updating process they would go through and then when they would have a code freeze, okay, no more changes, and they get that onto the CDs, and then who wants to buy the updated version? That's their revenue model, they got to keep it going.

So that's into the early 2000s. Amazon, early 2000, introduces cloud computing to the world. In that step, you can now not only store data, but you can do compute in the cloud in hosting centers. Around that time, you have Marc Benioff, who worked at Oracle have the epiphany, that's the story, that there's no reason to own software. It could be provided in the "cloud". What does that mean? Instead of each user having their own version of, say, Siebel software on their desktop, there's going to be one code base that sits in the cloud. And then, the user essentially links into, through APIs, links into that code base and through metadata, you can customize it for that particular user. All the user's data is held in the cloud. When they're using Salesforce, they're connecting to a hosting center to the cloud and they're working, essentially, the work is being done in the hosting center, not on their PC.

Now, there's variants of this that I won't bother with, but the bottom line is everything is in a hosting center in a cloud and I'm just connecting to it. Well, there's a number of implications of this. First and foremost, adoption costs go from quite high to very low. I can just link in and start using it. Now, there's a bunch of minimal, some steps along the way, got more complicated over time, but at the end of the day, big step from packaged software. Second, and as important as the first, is I can push updates to the software immediately, because I have this single code base, I update it, improve it, and I push it down to my customers. They start using it. There's not this annual or two or three year update cycle.

Hal Weitzman: It's like the transition that Netflix went through where they used to send you CDs in the mail or DVDs in the mail and then they went to streaming.

Marc Knez: That's right. It's a completely different technology paradigm and that's kind of what I'm ... So when there's a new technology paradigm, you expect there to be a new service architecture. Once there's a new service architecture, you get a new business model architecture, because this then, of course, leads to a big change in the business model architecture. Because now, as a software provider, I'm using AWS as a cloud provider. I'm managing the software, I'm updating the software, and then I'm connecting and interacting in real-time with the customer, getting their data. So there's real-time interaction, I have this digital connection with my customer. None of that existed in the pre-SaaS days.

Hal Weitzman: And then, they don't control their own ... Well, they don't host their own data as they used to.

Marc Knez: And that's right. There's variance of that. You can have your own hosting if you want. There's a lot of things that emerged over time. But in general, the idea is all my data is in the cloud. There was a lot of discomfort with that early on because of cybersecurity risk, et cetera. But the economics of it are so powerful that you saw more and more companies moving to the cloud and adopting these types of SaaS applications. There's a lot more to the Salesforce story that I won't get into. It emerged slowly over time, but one of the big steps as well was Salesforce is more than just CRM software or a CRM platform. They're also an app platform.

So third parties, just like on your iPhone, third-party app developers developed customized applications that integrate into their CRM platform. And that then just provided more functionality, more like DocuServe integrates in, lots of things integrated in, making it a better experience for the users. So you get this integration that's happening, but it's all happening in the cloud. None of that was possible on the days of packaged software. There's big scope economies that exist here at the platform level, so it's not surprising. There's network effects, scope economies, et cetera, that drove the consolidation around a company like Salesforce. The last little bit that's important, because that'll be important as we get to SaaS Apocalypse is that we move to a subscription model now, because you're not just paying for the discuss per user. Now, you're paying per user per seat. We go from per package to per seat. And so, it's a monthly subscription model that you're paying, which is extremely lucrative.

What's Salesforce paying for? Salesforce is paying for its hosting costs, paying AWS. It's paying for having people that can update the software, and it's paying for some reps. But at the marginal user case, the cost, marginal cost is zero, but they're pricing that. They're charging, let's say $100 per user or something like that. Varies. You add another user, you get another $100, it's costing you next zero. So it's a very lucrative model at scale. And so part of the excess apocalypse is the risks to that model, which I'll get into. So that's SaaS.

Hal Weitzman: And when did this happen, Marc?

Marc Knez: Late 2000s is when Salesforce emerged. Intuit, their QuickBooks software transitioned. One of the most notable transitions was Adobe because Adobe had been a package ... because you think about Salesforce, they were a startup, right? They were a disruptive startup. Think about Adobe. Adobe was the dominant provider of creative software and that was packaged software. In around 2011, I could be getting the date wrong, Shantanu Narayen, the CEO, made the decision that we're moving to SaaS. We're going to move to a SaaS model, where he announced to the world we are no longer going to do any development on our packaged software. So for all of our customers out there that own our package software, you're never going to get an update to that where you should anticipate having to move to a SaaS model. There was a lot of angst about this in the market. Matter of fact, the stock price for Adobe went down.

But it was the right decision because he knew something, and this is the foresight. What he knew was the economics of SaaS is going to dominate. Packaged software is going to die. And we are under extreme threat because the ease of adoption of SaaS is going to make it very easy for entrants to come in to our market and start coming after us. And so, there's a lot more details, but bottom line is it's better, because you can refresh it, you can improve it on a daily basis if you want. And second, the barriers to entry into our industry are going to come down. This is going to be a big opportunity for new players to come in with a SaaS model, just the way Salesforce had done to Siebel. So seeing that, he knew, foresight, and he got out in front of it and was wildly successful.

Their earnings went down for a while, stock price goes down, and then there was just a lot of growth, because the economics is very positive, because of the subscription model once it takes hold. And you're continuing to aggregate more capabilities into Adobe, you're charging higher subscriptions, life is great.

Hal Weitzman: Carry the Two is the show that pulls back the curtain to reveal the mathematical and statistical gears that turn the world. Co-hosts, Sadie Witkowski and Ian Martin bring unique perspectives from the fields of mathematics and statistics to convey how mathematical research drives the world around us. With each episode tackling a different topic, subscribe to Carry the Two, part of the award-winning University of Chicago Podcast Network.

Marc, in the first half, we talked about the rise of Software as a Service, SaaS, and all kinds of subscriptions, and we're all paying subscriptions in our personal and professional lives past few years, too many gym memberships that we don't use, et cetera. And there's the equivalent in business. A lot of businesses now looking at their subscriptions saying, "We're really using all these seats." As you say, "It's only $100, but it adds up. We have a lot of subscriptions. Are we really making use of that?" That's one reason that people have been looking at companies like Salesforce and saying, "Are they really sustainable?" There's a backlash against subscriptions. There's also AI. So tell us, how did we get from this situation, where we had this domination of subscriber companies to where we are now?

Marc Knez: Yeah. I'll get to why the subscription model is at risk. To some extent, it won't necessarily go away, but it'll be the amount that they can charge if they can preserve it. We have to start with why AI is so disruptive and the fundamentals, first step is the cost of producing a software application is close to zero. The idea is, the fancy term is democratization of software development, where you and I, as non-software developers with a little bit of training, could vibe code, which basically means we're just, through narration, asking questions, requesting a pack a piece of software, have the following functionalities. Anthropic, for example, who's the leader provider of this kind of capability will create that software for us. There's a bunch of little details in that, but at the end of the day, I don't need to be a software engineer. I don't need to be a very experienced developer. Having development skills helps, but you don't need a lot of it.

So first, Salesforce and any other, Adobe, any other very successful software company is really good at developing software. And so, there's human beings that are specialists in developing certain types of software, certain types of software languages, et cetera. Well, that expertise is trending towards becoming obsolete in lots of domains, because the machine is going to do it. They allow a large language model with the proper tools that's going to do it for you. So that's part one, because there's no doubt that Salesforce, Adobe, these kinds of companies, they're great at software development. They have a lot of smart people that can do software. Microsoft, on the gamut of companies. So you go from that expertise becoming more and more obsolete over time. So that's part one.

Part two is Agentic AI. If you think about what goes on in an enterprise, on a decent-sized enterprise day-to-day, you have a lot of employees in particular roles sitting in front of computers bouncing from one application to another. They might be in their Salesforce application, they might be in their email application, they're doing some word processing. There could be six or seven that they're using over the course of the day and then coordinating with others. In a nutshell, what they're doing day-to-day is typically fairly routine. It doesn't change much. It's just what do they need to get done that day, but the steps are relatively routine and somewhat predictable. Well, what Agentic AI, an AI agent can do is just automate all of that.

All of a sudden, all those applications have a new label. They're called tools. So I am no longer going to go into my ... I'm never going to look at this in this future. I'm not looking at the Salesforce screen. I'm looking at my UI that is connecting me to my AI agent and the AI agent is doing a bunch of things for me, which includes bouncing or connecting and using each of these tools that were apps that I have subscriptions to and they are scheduling for me. They're determining what set of sales leads I should go to next. They are crafting the emails. Responses are coming in. The Agentic AI is responding. I'm directing it, it's informing me, but my user interface to the world, to my work is through my AI agent. The AI agent sits between the user and all of these software applications.

In this somewhat extreme world, the software applications, like a Salesforce, using them as an example, they get pushed to the background. What they become essentially is very well-structured databases with rules associated with customer relationship management. It's this system of truth, system of data that the agent taps into. So it's still important, but it's not in front and center.

Hal Weitzman: It's scraping the information.

Marc Knez: It's not just scraping, because there's an API into Salesforce and it's telling Salesforce, "I need you to identify the next 10 leads." And it's doing that, but the machine is doing it. I, as a user-

Hal Weitzman: This is like the AI search versus Google, right?

Marc Knez: Yes.

Hal Weitzman: It's giving you a summary of the things that you need to know.

Marc Knez: Yeah. Instead of me going through and searching and assembling that information, it's just doing it for me, and much, much faster, much faster. As opposed to me as a human going from screen to screen and thinking about what, it's just doing it. And it's doing it in the middle of the night. This is going on continuously. You have two things. The cost of developing software is going to zero, and then you have Agentic AI potentially, the technology's there, Anthropic comes out with new tools regularly. Every time Anthropic comes out with one of these new tools, one of their coworker tools, they're making it easier to build these agents to do this work. Now, companies can use their tools to say, "I want my agent to tap into these databases attached to these applications and have this kind of functionality and that the Anthropic tool will help develop that agent pretty quickly." Now, there's a bunch of bells and whistles around this, but the bottom line is that's where things are going.

If we go back to the subscription model, currently, what protects Salesforce and other software subscription models is Salesforce still has some control over the AI agent connecting to their platform. What that means is they're still going to be able to charge per ... It's almost like every salesperson ... right now, we're charging by salesperson by seat. Now, we're charging by API connections, which are attached to a human being. We can still try to preserve that model, but here's the rub. Let's say I have a hundred salespeople, and so each salespeople is going to have their own AI agent. From a pure technology perspective, I only need one subscription to Salesforce. I don't need a hundred subscriptions.

It's just Salesforce sitting in between what they'll allow to happen as it relates to what's connecting to their platform. This gets back to economics always wins. This is putting a lot of pricing pressure on a Salesforce where, and I'm picking on Salesforce because they're the one that's probably most likely to sustain their position despite this if you're going to pick one, at least high on the list. But the key is, if I'm sitting down with Salesforce and I'm saying, "Listen, my AI agents are doing all these things and you're wanting to increase your subscription price," I'm going to push back on that really hard, because I know I don't like this pricing model anymore, because I really don't need all these 'seats'.

Hal Weitzman: Marc, you've described this as slow-moving, but is an apocalypse of some kind, a big event for SaaS. You've also indicated that not everyone is going to be wiped out. What are the companies that are going to survive? How will they survive?

Marc Knez: Yeah. I've been using Salesforce as an example, but it's somewhat extreme example because there is a large continuum of different types of software companies. What we put most attention on now is if you are a software company, where you have an application, where you're a user interface into some kind of data platform that has some kind of functionality, but that data platform obviously sits in the cloud, that can be replicated very quickly with an LLM. And so, you're at significant risk, because it's too easy to imitate what you're doing and it's not just, there's a term that I like to use, which is called absorption risk, which is it's not just that other players are going to come into the market and compete against you, you're moving to a model where the enterprise itself will simply add that capability to their software stack. If I need this kind of new application, new capability, I, my people will just quickly build that and add it to our capability inside the firm.

And you seeing that happening with CRM software, not large enterprises, but smaller, say a new, if you're a new venture and you're starting to scale, you're not going to go out and pay for salesforce.com right now. You're just not going to do it. If you have any tech capabilities and you're paying attention, you're going to quickly build your own CRM application. Now, if you're a large enterprise, and this takes me to who's not necessarily in the near term going to be in trouble, if I'm using CRM as a large enterprise, a lot of very core processes inside my business run on Salesforce. I can't just flip a switch and stop using Salesforce. I can start using Agentic AI to become more efficient and effective. I'm still having to use Salesforce though.

And so, Salesforce has quite a bit of time before they start having their medium-to-large enterprises dropping those subscriptions. There's going to be pricing pressure on those subscriptions, et cetera. That's going to occur. What is Salesforce doing? Salesforce is adding Agentic AI into their platform. So now, as a user of Salesforce, this Agentic AI capabilities I'm talking about, I can tap into those, those kind of capabilities, within the Salesforce application, and that giving me those benefits. And if I'm a large enterprise and I'm somewhat risk-averse because this whole Agentic AI, this is risky stuff, but I can easily just start being AI-enabled by leveraging the AI capabilities that my existing trusted software providers are giving me, I'm going to go there. I'm going there because this idea of reoptimizing my whole business around Agentic AI, that's too much.

Hal Weitzman: The companies that survive, the SaaS companies or former SaaS companies that survive will be those that can use AI and change their models and make it easy for the customer.

Marc Knez: There's one caveat to it though. I would say yes, near to medium term. If you ask me where is Salesforce going to be in 10 years and what it's going to look like, that's a different answer. That's why, when I say ... people say Apocalypse, they're talking 2027, that's crazy for a lot of the large enterprise software, Workday, et cetera, SAP, even, of course, Microsoft. However, Adobe is another story. If you think about Salesforce is tapping into all kinds of different processes inside the enterprise. Adobe, there's a specialist, a creative, that's using that software as a part of some development step. They're using that software, and we're paying an expensive subscription to Adobe for that. Adobe's great, but at the end of the day, which might be the end of the year, and this is happening quickly, all of that functionality can be provided by an LLM.

An LLM application may be developed on Anthropic, et cetera. I think I forget which LLM tends to be better at this kind of stuff, but bottom line is I can go to my creatives and say, "Okay, I want you to stop using ... I'm going to go from 20 Adobe subscriptions down to three or four because there are certain things I still need Adobe for because the LLMs haven't quite gotten there yet, but the rest of you are going to start using AI for this. And oh, by the way, I don't need as many of you." But the point being, I can just tell them, because they have this discreet job they're doing, that's very different than ripping out Salesforce, which is this enterprise software system that's tied to a broader enterprise software system. It's those more bespoke applications, as that can be replicated, those capabilities can be replicated with AI, even the larger companies like Adobe are a threat.

If you look at how Adobe's performed, it's been rough because it's just too easy. It's become too easy to replicate. And then, you've got other players coming in like Canva and et cetera. It's a matter of how easy is it to adopt the new AI capabilities and new AI solutions.

Hal Weitzman: But do you think that those, just quickly, those that survive, like the dinosaur apocalypse, those tiny mammals survive, will they be smaller and more nimble?

Marc Knez: Well, I think 10 years from now, the way we think about software and the software industry is going to be completely different. I think the idea of someone in their current form with their current business model "survive" and 10 years from now they're still there doing what they're doing today, that's not going to happen. That's just simply not going to happen. Some are going to do it. If Salesforce has incredible functionality within its CRM, that functionality that Agentic AI is still dependent upon is going to persist.

Now, more and more of it will get automated, but Salesforce will use AI to automate more and more of that within their platform. The problem they face is their revenue model, how they get paid and I think they'll survive. I think it's going to be much more difficult for an Adobe to survive because of what I described earlier. Think about the change of, there's not a app-specific user interface going from a world where that goes away and I'm bouncing from app to app and the user interface is now an agent, that's going to have a significant impact on what plays out in the future.

And there's one last little wrinkle to this. The problem that Salesforce has, as big as it is that at the end of the day, it's called "surface area" as it relates to the data that the agent can leverage as it's doing its Agentic AI is limited to what's inside of Salesforce, and that's going to be a problem for Salesforce. As a net large enterprise, I want one big massive properly designed data set that my Agentic AI runs on. I don't want to have the stovepiped data sets. If you ask me where we're going to be in 10 years, in 10 years, large enterprises are not going to have stovepipe datasets. So what is that going to then mean for a Salesforce? I think that's going to be challenging. They're going to have to figure that out.

Hal Weitzman: Okay. As you say, Marc, this is moving fast. We'll have to be back maybe at the end of the year to see how the apocalypse is playing out. For the moment, thank you very much for coming on the Chicago Booth Review Podcast.

Marc Knez: Yes. Good. Thank you. Enjoyed it.

Hal Weitzman: That's it for this episode of the Chicago Booth Review Podcast, part of the University of Chicago Podcast Network. For more research, analysis and insights, visit our website, chicagobooth.edu/review. When you're there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research. This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe and please do leave us a five-star review. Until next time, I'm Hal Weitzman. Thanks for listening.