The SaaSpocalypse Is Coming — But Don’t Count Out the Incumbents

Ben Lorica

Ben Lorica and Evangelos Simoudis on Robotaxis, the SaaSpocalypse, and the AI Stakeholder Squeeze.

Subscribe: Apple • Spotify • Overcast • Pocket Casts • YouTube •  AntennaPod • Podcast Addict • Amazon •  RSS.

Ben Lorica is joined by Evangelos Simoudis for a wide-ranging check-in on how AI is reshaping mobility, enterprise software, and corporate strategy. They compare Waymo and Uber through the lens of customer experience, examine whether the “SaaSpocalypse” is a real threat to incumbents or an investor overreaction, and unpack the emerging “AI stakeholder squeeze” facing CEOs, employees, consultants, and outsourcing firms.

Subscribe to the Gradient Flow Newsletter

Interview highlights – key sections from the video version:

Jump to transcript

Related content:

Support our work by subscribing to our newsletter📩

Transcript

Below is a polished and edited transcript.

Ben Lorica. Alright, so we’re back with my regular check-in with Evangelos Simoudis. His blog is at corporateinnovation.co. We are recording this on June 11, 2026, and we’re going to cover three topics, roughly 10 minutes each. Topic number one: AI and mobility. Evangelos, go.

Evangelos Simoudis. Good morning. Good to see you again — good to work with you. I was recently on the road quite a bit, and I had the opportunity to use both Waymo and Uber. I write about the fact that when we talk about robotaxis, it’s about customer experience — it’s not about the autonomous vehicle per se. The vehicle is important, but it’s about the customer. Because of these travels, I had the opportunity to compare the two, and the first point I’ll make is that when we compare the two, we really need to compare Uber Comfort with Waymo. It is unfair to compare UberX with Waymo — they’re two different products. Now, when we talk about customer experience, as some people know, ride-hailing is a low-margin business. So when you want to offer a superior customer experience, you have to wonder where AI makes a difference in that experience, and for what part of the experience the company can charge a premium. When I think about what the traveler cares about, the first thing is ETA — how quickly can a vehicle come to me? And here’s what I’ll say: Waymo still falls short, and this is where you see the power of the Uber network. It doesn’t matter how good your algorithm is if you don’t have enough vehicles. I was looking for a Waymo and it gave me a 10-minute ETA. Uber was there in two minutes — a very big difference. The second thing is the personalization of the cabin, and here Waymo is much better. There is consistency. With Uber, I don’t know if I’ll get into a Lucid or a Toyota Pathfinder or whatever. And—

Ben Lorica. And the driving style is consistent, and the UX on the screen is consistent, right?

Evangelos Simoudis. Exactly, all of that. And by the way, the only thing I can control with Uber is to say I don’t want the driver to speak to me. And many times the driver doesn’t even speak English — if I wanted to talk to them, I couldn’t — and that’s a problem, because it relates to the next point, which is safety. Safety has two parts. One is how the driver is actually driving, and we’ve seen a great variety of driving skills among Uber drivers in New York, Boston, and Los Angeles — again, in Uber Comfort, not UberX. With Waymo, once I believe the autonomous vehicle is safe enough for me to get in — which, for the geofences where I use it, I do — the experience tends to be much more consistent. And in terms of ride quality, there’s also cleanliness. I’ve gotten Uber vehicles where the driver is a smoker, and I’m not a smoker.

Ben Lorica. You mean Uber?

Evangelos Simoudis. Yes, I’m sorry — Uber. Whereas with Waymo that’s not an issue. This is why I keep saying that for people who have already embraced the autonomous vehicle as a safe way to travel within a geofence — San Francisco, Austin, wherever — it’s no longer about whether the vehicle will take me from point A to point B safely. It’s more about what happens during that process. Another thing I’m seeing, and I think it’s an opportunity for these companies, is using AI to offer much more end-to-end mobility. So if my destination is the Transamerica Tower in San Francisco but Waymo can’t drop me off in front of it — either because of congestion or traffic regulations — the ability to direct me after I get out of the vehicle is an important role for AI. So that’s what the consumer sees. Now, flipping it—

Ben Lorica. One feature I’d give Uber a shout-out for is scheduling. You can schedule a pickup, and more importantly, you can schedule a drop-off, which is huge. If you have an appointment, you can be dropped off five minutes before it.

Evangelos Simoudis. Very good point. Now, when I look at the flip side — what’s happening within the enterprise — when I look at what a TNC, a transportation network company, has to offer, let me elaborate. They need to worry about customer acquisition, customer attrition, maintenance — everything that relates to the customer. Then they have to deal with ride management, vehicle operations, vehicle technology, and vehicle manufacturing. Waymo, for example, worries about how many vehicles it can put on the road, and that’s part of the ETA problem. And overall, what does the marketplace look like? Uber definitely has a marketplace; Waymo does not. As we move toward autonomous mobility, I’m seeing two things: where AI can play a role in these operations, and what’s interesting about Uber — the partnerships it’s starting to establish in the autonomous vehicle space. They have over 30 partnerships, and what’s interesting about these partnerships, as it relates to AI, is that they shift the AI responsibility to the partner. Uber has its own autonomous vehicle investments, but they also partner with WeRide from China and with Nuro here in the US, particularly for Uber Eats. They partner with companies that deal with vehicle maintenance, and so on.

Ben Lorica. You might even be able to call a Waymo on the Uber app.

Evangelos Simoudis. Exactly — that’s what’s happening in Atlanta. To me, this is very interesting. Waymo is clearly experimenting and looking at what models work, and that does impact their investments in each of these areas. It impacts where AI will play a role and whether they’ll take on that responsibility or push it to a partner. And ultimately, I’ll go back to what I started with: they’re testing for what part of the customer experience the traveler is willing to pay for. Am I willing to pay for personalization within the cabin? Or is it more important that the ETA is short? Or, as you said, the ability to schedule both pickup and drop-off? These things impact how to approach AI, how to invest, and so on.

Ben Lorica. And before we close this topic, Evangelos, have you learned anything new on what I’d call the most important stat: how many Waymo vehicles does a single human operator oversee?

Evangelos Simoudis. I have not. I’ll say that I’ve been looking a lot at costs. If you look at the Chinese operators — Pony.ai, Wave, WeRide, Apollo — they are laser-focused on bringing down the cost of the vehicle. And tomorrow I’m giving a talk on this. The next area they’ll look at, where AI plays a big role, is the ratio of vehicles per teleoperator. This is something that hasn’t been discussed a lot: what are the actual responsibilities of the teleoperator? We see different models among the various operators and providers.

Ben Lorica. Alright. Topic two is the so-called SaaSpocalypse. As many of you know, SaaS stands for software as a service, and the SaaSpocalypse, broadly speaking, refers to the decline in stock prices of some key SaaS companies. The three I can name off the top of my head are Salesforce, Workday, and ServiceNow — all of which took a hit. The story here is that with the rise of Claude Code, Codex, and similar tools, there’s a growing sense that AI will disrupt lots of workflows. Right now, coding seems to be where it’s happening, and so the SaaSpocalypse is more of a speculation on the part of investors that this could impact the software category in general — not just software for developing software. The trend comes from a few directions. First, the large AI labs themselves — Anthropic, OpenAI — might be able to build AI-native enterprise software tools to replace some of these offerings. Second, there are AI-native startups that might build the next generation of Workday or Salesforce. Third, there’s the notion that large Fortune 500 companies, sick of paying SaaS vendors, will just vibe-code their own Workday or Salesforce — in-house solutions. And then there’s the possibility that the SaaS companies themselves — Salesforce, ServiceNow, Workday — will disrupt themselves by building new offerings that eliminate the need for their own large software installations. A lot of this is speculation, and anyone who has tried to integrate or deploy one of these SaaS solutions knows they don’t work in isolation — they talk to legacy systems within enterprises, which means a lot of integration. Those things are hard to rip out. As Evangelos and I have noted, the dirty secret in AI right now is not really the model — it’s the data integration and plumbing that AI needs to function properly. So: the SaaSpocalypse. If I were to speculate, I think there will be disruption, but I don’t think we should rule out the incumbents.

Evangelos Simoudis. I’ll add a couple of points to that very nice summary. The first is that the frontier labs themselves are trying to find enduring business models, and I think that creates fear among investors. They’re saying: if OpenAI can’t make money with ChatGPT and then moves into, say, workforce management as an application, what does that mean? The uncertainty around the frontier labs themselves is feeding this problem. The second point — and we’re seeing this in startups we’re considering or have invested in — is that because of the power of AI in software engineering, developing what you might call a version 0.8, or generously a version 1.0, of a new product is much easier and requires a much smaller team than it used to. I used to say we had cracked the code on financing startups; now I think we’ve cracked another code — we can create early-stage startups with two or three people and a bunch of agents. We’re looking at two startups right now, one in Sweden and one in the UK. In both cases, it’s two people and hundreds of agents, and their products are genuinely very interesting. Without much effort, they’ve been able to get design partners from notable corporations. How scalable that ultimately is, we’ll have to see, but those two points add to the uncertainty you described. On the technology front, we’re also starting to hear the term “headless AI,” which goes back to your data integration point. A number of incumbents — Salesforce and others — are starting to open up their databases to be accessed by agents developed by their customers. What the business model for that will look like remains to be seen, but the uncertainty around economics is putting pressure on stock prices.

Ben Lorica. And if you look historically, these incumbents each started in one area. Salesforce started with sales force automation — as the name suggests — and then expanded.

Evangelos Simoudis. Exactly — sales force automation, as the name suggests, and then they expanded into marketing and other areas.

Ben Lorica. And Workday was HR. So presumably, if you’re a startup, you can focus on one area as they did and come out with something more modern. You would have to replace the functions they provide — the reason their data is valuable to agents is because they have the data, and the reason they have the data is the front-end user-facing functions that generate it. So you’d have to replicate that. My sense is that the rollout would still be slow, though — there are too many critical functions inside a large corporation that depend on this HR SaaS software for it to just be replaced overnight. So the actual transition is going to be gradual.

Evangelos Simoudis. Yes. And putting on my investor hat: the smartest enterprise AI startups are the ones starting from one of two observations — sometimes both. The first is: how should I rethink a given business process, and where does that process sit within the broader workflow? The smart ones think about how AI enables them to completely rethink that process — maybe even rethink the entire workflow. The second category of smart startups are the ones that look at existing workflows and try to identify important white space that existing applications don’t address at all — either because no one thought of it, or customers didn’t think to ask for it — and then build something with an AI core and go to JP Morgan, Citibank, or GM and say: here’s your HR workflow, here’s what’s missing, here’s how we’re covering it. That’s how they get a foot in the door.

Ben Lorica. The classic startup wedge.

Evangelos Simoudis. Exactly. And I think AI not only enables but pushes people to think from first principles, rather than saying, “this is how the process was done — now we’ll sprinkle some AI on it and get a 5% improvement.” The smart ones ask: why is the process done this way in the first place? As part of the research I’ve been doing for what I’ve started calling the AI stakeholder squeeze, I’m looking at the quote-to-cash workflow — all the processes that comprise it and how it manifests across various industries. I think there’s a lot of opportunity there.

Ben Lorica. So for investors — and we’re obviously not providing investment advice here — the question is how quickly will this play out? Almost everything we’ve named could happen quickly in a specific category, just as it happened quickly in coding. For example, I believe some SaaS design software is already being impacted. Though to be fair, there’s more to any piece of software than its one core function. In design, being able to mock something up is one thing, but there’s also asset importing, collaboration, version control, and so on. Having the basic capability to perform the key function might give you a beachhead and allow people to gradually migrate. The key question for me is: will the transition be abrupt or slow? I think it depends on the category. Take HR software at a big company — it touches so many things. Evangelos, “Can I go on vacation?” — that’s HR software. So it may come down to being able to insert yourself into a few critical functions and grow from there.

Evangelos Simoudis. And we’ve seen this in many cycles — the perennial question of best-of-breed versus all-in-one. Different corporations have different philosophies about that.

Ben Lorica. And inside corporations there could be different adoption patterns by division. This division moves fast; this one stays conservative.

Evangelos Simoudis. Right. The prediction I’m gravitating toward is that we’re going to see a series of blips. Today we see the blip in software development — Anthropic’s revenue is skyrocketing because of it. But I’m not certain whether each blip becomes a sustained trend or just fades out. Innovation is always balanced by: what am I getting in return, and what does this cost me? Even in software development, enterprises are starting to ask: for which tasks do I use Claude or another frontier model? For which do I use an open-source model? For which tasks do I rely on a hyperscaler’s data center, and for which do I bring this into my own AI factory and keep it in-house? These decisions will take time. I’m concerned when everyone talks as though AI adoption is just accelerating unchecked. There’s no question it’s impacting all of us and we’re seeing real revenue and workflow effects, but we also need to be patient and observe trends over time, rather than reacting to what happened last week.

Ben Lorica. And a lot of what we’re discussing presumes that the party will keep going. Right now it’s widely accepted that we’re in a bubble, and this bubble could burst at any time. The data center buildouts worry me, as I mentioned before we started recording. And as Evangelos points out, everything in this space is ultimately a business decision. Open weights — that’s a business decision. Meta, for example, is backing away from open weights. Everything is a business decision. Alright, topic three: the AI stakeholder squeeze. Take it away.

Evangelos Simoudis. I’ve been accelerating my research on this topic. I first mentioned it on this podcast a few weeks ago and have written a few pieces about it, but the more I dig, the more interesting I find it. I’ll share a couple of early insights. My focus has been on what I keep calling legacy corporations — I’m not spending as much effort on what startups or digital natives are doing. Uber will figure it out easily, and Netflix is already leading the way.

Ben Lorica. The companies that actually make money, in other words.

Evangelos Simoudis. Exactly. I’m far more interested in legacy companies in manufacturing, healthcare, insurance, and financial services. The first insight is that when the CEO personally owns the AI initiative, I’m starting to see a very different approach than when the CEO delegates it to the CTO, CIO, or CAIO. The reason is that AI, to me, is an opportunity for a corporation to rethink its culture, its workflows and processes, and its technology — not just to incorporate a new technology the same way you’d incorporate any other. Others have drawn the analogy to the internet in the late 1990s: companies initially saw it as a technology that let them communicate faster. But the ones who ultimately benefited were the ones who saw entirely new business processes being enabled by it, new personnel roles, and a restructuring of existing processes. I’m starting to see a big difference between companies looking at AI one way versus the other.

Ben Lorica. So among the CEOs in your preferred category — the ones personally engaging with AI rather than delegating it — I assume there are subcategories. There are those who are more knowledgeable about AI’s limitations and capabilities, and those who are less so. One metric that many people, including me, have settled on — maybe half-jokingly — is that the executive most likely to really “get” AI is the one who is closest to the actual work. If you’re in an ivory tower, so far removed that you don’t know what people are doing day to day, your ability to direct how AI should be used won’t be great — because everything is an abstraction to you.

Evangelos Simoudis. That’s a good way to frame it. And this is where my work on the squeeze comes in — I’ve been working to identify the forces driving it. But first—

Ben Lorica. Maybe define what the squeeze is for our listeners, in a sentence or two.

Evangelos Simoudis. It’s a phenomenon resulting from two different pressures: on one side, how the CEO feels pressure to use AI, and on the other, how employees are reacting to AI. I’m still refining the definition — it’s a work in progress — but people can go to my blog for more detail. The point I’ll make here is: if you look at someone like Jamie Dimon of JP Morgan — a huge corporation — I think he understands very well what impact AI can have across his portfolio of businesses and how to approach it, recognizing that it’s not only about the technology. I’ve also mentioned Intuit as another example.

Ben Lorica. With Jamie Dimon, he seems to have a genuine understanding of what people actually do in various parts of the bank. He’s the kind of CEO who walks the floor.

Evangelos Simoudis. Exactly. I’ve been spending time with a hospital chain I can’t name yet, and I’m seeing very encouraging results — strong collaboration, real ownership, no delegation. Going back to costs: I think this is important enough that the CEO, the board, and the technical leadership need to be able to determine what they want to own. What will our AI factory look like? Will we have one, and if not, why not? Versus what we’ll outsource to partners — whether that’s a hyperscaler with frontier model capabilities like Google or Microsoft, or some other combination of infrastructure providers. These are starting to emerge as early insights. It may also be that we’ll find meaningful differences even within the category of digital natives. For example, I consider Airbnb a digital native, but I don’t see them as a particularly innovative user of AI. We’ll see how they evolve.

Ben Lorica. Quick question on legacy corporations: as you know, the hot new role in Silicon Valley is no longer data scientist — it’s forward-deployed engineer, and to me that hearkens back to consulting. So, are you seeing the AI stakeholder squeeze as an opportunity for the big consulting firms?

Evangelos Simoudis. Two comments. First, I’m very bothered by the “forward-deployed engineer” framing, and I wrote about this on LinkedIn: IBM had systems engineers doing exactly this several decades ago.

Ben Lorica. Everything is a relabel.

Evangelos Simoudis. Exactly — people can read about that on my LinkedIn. Now, as for management consulting: it’s changing dramatically and will be one of the industries impacted by AI quickly. Their business model is under pressure. We’re hearing of corporations asking for outcome-based pricing rather than time-and-materials billing.

Ben Lorica. And the stereotype is that a firm like McKinsey hires many people straight from Ivy League schools, and as a C-suite executive you might find yourself talking to someone who graduated a few months ago. How is that different from talking to Claude or ChatGPT?

Evangelos Simoudis. Exactly. The second thing we’re seeing is corporations demanding that the senior partner be much more personally engaged with the account. And third — you may have noticed this — within the last month, several new companies have been funded with significant capital, with McKinsey, Capgemini, and others as investors, specifically to create new models for management consulting and IT consulting centered around AI. One last point: I’m also starting to see a significant impact of all this on Indian IT outsourcers.

Ben Lorica. It’s not just the outsourcers — multinationals are also cutting back on their India teams.

Evangelos Simoudis. Both are happening — cutting back on India headcount and pushing back on the rates outsourcers charge. It actually started for me last year after CES ’25, when I was asked to present to the CEOs of large Indian IT service providers. The presentation was about AI’s impact on their business, and the discussion afterward was very lively. Since then, I’m seeing many more of them coming back to our firm for additional conversations — now with concrete, and concerning, data driving those conversations. I think we’re going to see a big impact on that model, not just management consulting.

Ben Lorica. Here’s the thought experiment for you: if Y2K were today, I could do everything I needed using Claude Code. And Y2K, by the way, is what gave rise to the Indian outsourcing movement in the first place.

Evangelos Simoudis. Yes.

Ben Lorica. And with that, thank you again, Evangelos.

Evangelos Simoudis. Thank you. Bye.