What's Driving the SaaSpocalypse (techcrunch.com) 69
An anonymous reader quotes a report from TechCrunch: One day not long ago, a founder texted his investor with an update: he was replacing his entire customer service team with Claude Code, an AI tool that can write and deploy software on its own. To Lex Zhao, an investor at One Way Ventures, the message indicated something bigger -- the moment when companies like Salesforce stopped being the automatic default. "The barriers to entry for creating software are so low now thanks to coding agents, that the build versus buy decision is shifting toward build in so many cases," Zhao told TechCrunch.
The build versus buy shift is only part of the problem. The whole idea of using AI agents instead of people to perform work throws into question the SaaS business model itself. SaaS companies currently price their software per seat -- meaning by how many employees log in to use it. "SaaS has long been regarded as one of the most attractive business models due to its highly predictable recurring revenue, immense scalability, and 70-90% gross margins," Abdul Abdirahman, an investor at the venture firm F-Prime, told TechCrunch. When one, or a handful, of AI agents can do that work -- when employees simply ask their AI of choice to pull the data from the system -- that per-seat model starts to break down.
The rapid pace of AI development also means that new tools, like Claude Code or OpenAI's Codex, can replicate not just the core functions of SaaS products but also the add-on tools a SaaS vendor would sell to grow revenue from existing customers. On top of that, customers now have the ultimate contract negotiation tool in their pockets: If they don't like a SaaS vendor's prices, they can, more easily than ever before, build their own alternative. "Even if they do not take the build route, this creates downward pressure on contracts that SaaS vendors can secure during renewals," Abdirahman continued.
We saw this as early as late 2024, when Klarna announced that it had ditched Salesforce's flagship CRM product in favor of its own homegrown AI system. The realization that a growing number of other companies can do the same is spooking public markets, where the stock prices of SaaS giants like Salesforce and Workday have been sliding. In early February, an investor sell-off wiped nearly $1 trillion in market value from software and services stocks, followed by another billion later in the month. Experts are calling it the SaaSpocalypse, with one analyst dubbing it FOBO investing -- or fear of becoming obsolete. Yet the venture investors TechCrunch spoke with believe such fears are only temporary. "This isn't the death of SaaS," Aaron Holiday, a managing partner at 645 Ventures, told TechCrunch. Rather, it's the beginning of an old snake shedding its skin, he said.
The build versus buy shift is only part of the problem. The whole idea of using AI agents instead of people to perform work throws into question the SaaS business model itself. SaaS companies currently price their software per seat -- meaning by how many employees log in to use it. "SaaS has long been regarded as one of the most attractive business models due to its highly predictable recurring revenue, immense scalability, and 70-90% gross margins," Abdul Abdirahman, an investor at the venture firm F-Prime, told TechCrunch. When one, or a handful, of AI agents can do that work -- when employees simply ask their AI of choice to pull the data from the system -- that per-seat model starts to break down.
The rapid pace of AI development also means that new tools, like Claude Code or OpenAI's Codex, can replicate not just the core functions of SaaS products but also the add-on tools a SaaS vendor would sell to grow revenue from existing customers. On top of that, customers now have the ultimate contract negotiation tool in their pockets: If they don't like a SaaS vendor's prices, they can, more easily than ever before, build their own alternative. "Even if they do not take the build route, this creates downward pressure on contracts that SaaS vendors can secure during renewals," Abdirahman continued.
We saw this as early as late 2024, when Klarna announced that it had ditched Salesforce's flagship CRM product in favor of its own homegrown AI system. The realization that a growing number of other companies can do the same is spooking public markets, where the stock prices of SaaS giants like Salesforce and Workday have been sliding. In early February, an investor sell-off wiped nearly $1 trillion in market value from software and services stocks, followed by another billion later in the month. Experts are calling it the SaaSpocalypse, with one analyst dubbing it FOBO investing -- or fear of becoming obsolete. Yet the venture investors TechCrunch spoke with believe such fears are only temporary. "This isn't the death of SaaS," Aaron Holiday, a managing partner at 645 Ventures, told TechCrunch. Rather, it's the beginning of an old snake shedding its skin, he said.
Exactly what every IT dept needs (Score:5, Insightful)
Re:Exactly what every IT dept needs (Score:5, Interesting)
It tells you something that AI slop apps are a viable alternative to SaaS outfits.
Re: Exactly what every IT dept needs (Score:5, Interesting)
Where I work, we generally frown on SaaS, especially for the stuff that we use the most. The last thing we need is massively overpriced vendor lock-in shit that gives other people access to all of our data.
Atlassian recently decided to discontinue all of their on-prem products, and their sales derp was just having none of it when our IT director told him, point blank, that we're not moving anything into the cloud, under any circumstances. It's like they have no concept that the only reason they were an option for us to begin with was the fact that they didn't require SaaS crap at the time.
Re: (Score:2)
The sales derp is right and your enterprise architect derps should be fired. Having a couple of neckbeards who can write a bash script does not qualify your company to homebrew critical applications any more than Salesforce people building their own HQ tower
He didn't say any of that. He specifically said his enterprise architect did not want to move applications to the cloud that used to be on-prem. Specifically Atlassian does not offer an on-prem variant anymore. There are still many reasons for keeping things off the cloud. The main reason is cost for the cloud is often not as cheap as it claims to be. The second reason is the reliability of cloud environments is theoretically high but in actuality there have been many outages of AWS, Azure, Cloudflare, etc
Re: Exactly what every IT dept needs (Score:1)
Excuse me sir, they're greybeards, not neckbeards. Besides, the last time one of you Salesforce derps made this kind of comment to me, it was something to the effect of "Salesforce spends $5 billion a year on R&D". The very next day, a major Salesforce data breach came to light. You may have heard of it. I mean, it impacted a lot of huge companies after all. And in the last month, they've had a few major outages.
You know how many of all of the above we've had in the time I've been here? Exactly zero.
I g
Re: (Score:2)
I *really* wish Redmine would modernise and catch up with the times. Back in the day it was the god-mode alternative to Atlassian and Github. But it really does look like a 2010s rails app these days. Hard to sell the boss on.
Its a shame, it really did solve so many problems.
Re: (Score:2)
It gets worse. The real lesson from AI?
Your job is so mindless and dumb that you can be replaced by a stupid statistical algorithm.
Re: (Score:2)
Re:Exactly what every IT dept needs (Score:5, Insightful)
Indeed - and actually, the running of the thing is probably harder than building it - not least because you run it 24x7, forever, whereas you may mostly build it for (say) 6 months of 8x5, and then can drop to minimal development operations after that.
The "saaspocaypse" isn't really real. What is happening is that people are realising that a lot of the "big boy" SaaS products are pretty crappy, cost a fortune and actually aren't necessarily run all that well. They also need a load of full-time "devs" to make them do anything useful, and so the ROI on them isn't nearly as good as it should/could be.
If you're a small or medium sized business, you definitely do not want Salesforce, Servicenow, etc - you want a far, far simpler system which feels like you're going to outgrow it in the next year or so. In a year or so, you won't have outgrown it at all - but you will have saved a tonne of money. The really big guys might well build their own, and it might be a way for them to (finally!?) properly personalise what they do for their customers, but they sure aren't going to be "vibe coding it in a weekend" - it'll take longer than that, but it is perhaps more accessible now than ever before.
I've said it before, but I predict the opposite of the SaaS-pocalypse - actually, I see an explosion of SaaS apps coming - with the barrier to entry reduced significantly, a load of new products will come along. Sadly, 80% of them will be PoCs dressed as finished products, and will likely fail spectacularly within a year or two. The remainder though will likely be small going-concerns, but service a loyal customer base, "do one thing, do it really well" and hopefully properly nail the customer service to deliver some real value.
Someone on Linkedin was rattling on about how there are going to be a rash of single-person billion dollar companies because we're all going to be making apps now - I seriously doubt that. If it's easy to do, then it's not going to be worth the billion, but it could be a very nice living for a decent number of people, and maybe some "a few million" sales and IPOs for a very small number, followed by the requisite enshitification and exodus of the loyal users that made it, etc etc.
Re: (Score:2)
Zhao forgets what quality, supply chain, consistency, scalability, reliability and lifecycle worth mean. His remarks are self-serving, and facile at best-- and a symptom of a deeper disease that permeates capitalism: Cheap Tricks That Don't Work.
This is Darwinism at work.
Re: (Score:3, Interesting)
Re: (Score:2)
FWIW, that's roughly my experience too. I've yet to see "I vibe coded it over the weekend" actually turn into something, but in the meantime as an accelerator, it's been pretty good.
Re: Exactly what every IT dept needs (Score:1)
That's mostly what I view AI as -- a replacement for what Google used to be back in 2015. Now that Google sucks, LLMs are a suitable replacement.
Though I still prefer stack overflow, and the reason why is because some people chime in with useful tidbits that you should probably consider, where LLMs really don't even understand what they're telling you.
Re: Exactly what every IT dept needs (Score:2)
Lame
Re: (Score:2)
My employer runs mostly it's own in-house coded system that's been built over 25 years. I'm effectively one of two developers maintaining it and for a long time that was patching critical bugs and adding a few key features each year. I had the question raised a year ago if we could added asset tracking capabilities - since we already had records for employees and a location based inventory system it seemed like a really natural fit but we
Re: (Score:2)
Applications no one knows how to maintain and have zero support. There is a reason enterprises BUY software vs build, support. Access Database apps version 2.0
I agree with these sentiments to an extent yet one thing I've seen in my area is what larger enterprises buy tend to end up looking more like SDKs than finished product. There may be a core that resembles a product yet basically to be usable requires a substantial amount of programming activity with all of the follow on maintenance / lifecycle concerns anyway. I could see cases where better generic tooling and access to open supporting stacks results in reduced reliance on SaaS.
Off the shelf software/SaaS
Re: (Score:2)
If you're believing the AI hype, apparently the AI is going to maintain and support it for you. Just give Claude Code your AWS credentials, and it will take care of it. What could go wrong?
Sure, you're laughing at that statement right now. BUT, don't forget... Even if YOU don't believe that statement is true, you're going to get sent down this rabbit hole anyway if your boss does or your bosses boss does.
Re: (Score:3, Insightful)
Often companies have to learn the hard way. Many companies who outsourced everything to India-based IT firms during that fad's heyday often regretted it later when managing miscommunications costed more than the initial labor savings. Embedded domain knowledge matters.
Some projects did stay in India, but it wasn't the silver bullet of "cheap IT" that the hype nebula implied.
When the side-effects of "AI slop-ware" start adding up, I expect a similar pattern will follow. Yes, AI will take over some "traditio
Re: (Score:2)
Re:Exactly what every IT dept needs (Score:4, Informative)
Yeah, I get it, I spent years running a software dev team that rebuilt LOB abomination built in Excel. That said, if you can spend 15 mins prompting an AI and it builds a ticketing or CRM system for you that just works and meets your needs, isn't that a valid choice over the insane costs of the available third party systems?
That's not reality. And even if it were how do you know it's secure? How do you know it's accurate? How do you know it's properly storing data? "15 minutes of prompting" to get this isn't a reality so why both asking about it?
Re: (Score:3)
Bullshit software for bullshit jobs ... (Score:3)
... going the way of bullshit. How predictable. And how poetic.
Re: (Score:1, Troll)
This idea that there are all these middle managers doing nothing hasn't been true in almost 20 years but I don't think people want to let it go.
By the time I was doing actual work instead of just dumbl
Re: (Score:1)
Re: Bullshit software for bullshit jobs ... (Score:2)
Iâ(TM)ve seen plenty of union stuff, and I have to say this is a pretty interesting and useful tank. Thanks for the awareness. That would make sense why as workers become managers in unions, and become executives, they all seem to be picked to be anti union.
Oh hurray.... (Score:2)
Already dealing with going from one company to another or, god forbid, partnering with another company tended to be a pain in the ass as they decided to apply stupid customizations or even write bespoke software for utterly common stuff is a pain...
A world where no two companies have anything vaguely resembling a common frame of reference sounds.... interesting...
Re: (Score:2)
Already dealing with going from one company to another or, god forbid, partnering with another company tended to be a pain in the ass as they decided to apply stupid customizations or even write bespoke software for utterly common stuff is a pain...
A world where no two companies have anything vaguely resembling a common frame of reference sounds.... interesting...
This is being sold as a good idea by the companies providing the AI agents. Because if *EVERY* company is custom code made by AI, the only solution to any problem will be an AI agent to continue the project. No need for an humans, you can just pay way, way more to keep the AI made code sort of limping along between AI revisions.
I hope I get to retire before this infestation completely infects the business I work for. Luckily, we're backwards and slow, though I do get a lot of push to use AI to "help" me rig
Re: (Score:2)
Recent had a non-technical executive give me "ok, but I got an LLM to spit out a 'hello world' that manages to execute the simplest, most basic operation, so maybe it's ready to displace people now".
It's *really* rough how much software development has people who don't understand the work at *all* in charge... Giving a very reasonable and open assessment of LLM gets them to nod sagely, but it's clear they are thinking we are underselling LLM because they assume we are just covering for our jobs...
Re: (Score:2)
Recent had a non-technical executive give me "ok, but I got an LLM to spit out a 'hello world' that manages to execute the simplest, most basic operation, so maybe it's ready to displace people now".
It's *really* rough how much software development has people who don't understand the work at *all* in charge... Giving a very reasonable and open assessment of LLM gets them to nod sagely, but it's clear they are thinking we are underselling LLM because they assume we are just covering for our jobs...
Management has to relearn every lesson the hard way. AI is just going to have them relearning the same lessons on a faster repeating schedule.
shit replaces shit (Score:5, Insightful)
the moment when companies like Salesforce stopped being the automatic default. "The barriers to entry for creating software are so low now thanks to coding agents,
I can believe that coding agents would produce a result equivalent or superior to using salesforce. But what about compared to competent developers not using a platform which is total garbage?
Re:shit replaces shit (Score:4, Interesting)
the moment when companies like Salesforce stopped being the automatic default. "The barriers to entry for creating software are so low now thanks to coding agents,
I can believe that coding agents would produce a result equivalent or superior to using salesforce. But what about compared to competent developers not using a platform which is total garbage?
Let us know when you find that.
Re: (Score:3)
But I have it on good authority from many slashdotters that they are that! /runlegs
Re: (Score:3)
Competent developers using a decent platform have better stuff to do than the stuff you would use salesforce for/how much people would be willing to pay for something like salesforce.
Re: (Score:2)
Competent developers using a decent platform have better stuff to do than the stuff you would use salesforce for
You mean underpinning someone's entire business?
how much people would be willing to pay for something like salesforce.
Most people using salesforce don't actually need something like salesforce. For one thing, they need something that works.
Re: (Score:2)
For smaller companies using Salesforce was overkill. Can AI be a viable alternative? In the short term and for simple requirements, yes. From what I have seen, if the code works, it is very fragile. Making small changes can break them. That is because they do not understand details. That is not including all the examples where the code looks fine but does not work.
However, long term and complicated installations will struggle under AI. I can see years from now that there is a spike in work to replace thes
Re: shit replaces shit (Score:2)
They struggle under Salesforce, too
Re: (Score:2)
They struggle under Salesforce, too,
Salesforce is known. The issue is costs money to customize and change Salesforce. With these AI agents, it may be far easier and cheaper to replace everything than modify the code as no one understands it.
Re: (Score:2)
I agree with all of those points. My real comment about Salesforce here is that they are counting on AI to get them out of their mess, with their apparently already unmaintainable codebase.
Proof in point (Score:3)
I'm 90% sure that summary was itself written by AI.
Re: (Score:3)
Unclear how much the bubble is floating the market versus sinking others as it goes.
The 25% functional unemployment number is actually pretty low by the standards of the last 30 years. In the late 90s it was low 30s, getting to about 27% before 2008 hit pushing it back up to low 30s, and then it managed to touch 25% again as COVID spiked it, and yes, after COVID it went all the way down to 23% and so it's a *hair* higher than then, but 25% is actually, over the course of the last 30 years, a pretty typical
non-issue (Score:4, Insightful)
Re: (Score:2)
-- interns still have to learn X86 assembler. .
After 40 years in IT, I have yet to meet anyone who codes X86 assembler. We must live in different worlds.
Re: (Score:2)
Hey, just because your company does all it's work on Z80 assembler doesn't mean the rest of us can ignore x86...
But seriously, other than an academic exercise, I agree that virtually no one does assembly. Especially misguided if someone thinks their hand-written assembly likely consistently beats modern compilers for a lot of code.
SaaSpocolypse is about data & digital sovereig (Score:5, Insightful)
The bigger issue with SaaS is data and digital sovereignty, not AI.
This story still promotes the fallacy that AI can generate production-quality code or anyone can use AI to build a SalesForce alternative or some other nonsense.
The real issue is the erosion in trust in the big companies and hosting countries that data is safe, secure, controlled, not being used to train AI or shared with US government entities, ...
Digital and data sovereignty are causing folks to turn away from SaaS more than people rolling their own software solutions using AI...
Pay Per play (Score:2)
A friend told me that pay-per-play is where the money is, as opposed to buying.
Software has been moving towards subscription-based models because they generate more profit. Just like PC Lint (after Jim Gimbel retired and sold it to Vector Informatik -- I'm still using version 9.0), which has turned into a subscription service. People who do C/C++ programming on a non-regular basis and do not need the newest version, why upgrade? The same thing for Boundchecker or Timeslips. Timeslips is fully SaaS, as oppos
The barriers to entry... (Score:2)
..for creating crappy software are so low
Real progress is being made, but current tools do not create excellent software
I suspect that future tools will, but early adopters of immature tools will discover the depths of software hell
I mean, I guess it is a better choice than Salesfo (Score:2)
I mean, I guess it is a better choice than Salesforce, SAP et al. but that really is not saying much, and the AI slop is coming with a hefty maintenance disaster.
People are probably just tired of all the subscriptions, and getting nickled and dimed for every basic feature.
SaaS was classic rent-seeking behavior (Score:3)
The companies started seeing sales of software go down, because people had what they wanted, and didn't need "new features"*, So sell it as a service, so you pay rent on the software, and can lose everything if you don't, so you're locked in.
It isn't buy vs build, it's *OWN*, not cloud-based crap.
* In '95, in PC Mag, they did a review of then-current word processors, and noted that 90% of the users didn't use 90% of the features *then*, and of the 10% that did, they only used them 10% of the time. But "we've got new features (that you'll never use), but you need to buy the new version!!"
Re: (Score:2)
* In '95, in PC Mag, they did a review of then-current word processors, and noted that 90% of the users didn't use 90% of the features *then*, and of the 10% that did, they only used them 10% of the time.
This was and always will be meaningless because 1) the features people use vary from person to person and job to job and 2) that 10% of the time is critical to their business.
Re: (Score:2)
The subscription model is hardly a recent development. IBM was leasing software to customers in 1969. And they were leasing equipment to customers before 1920. The idea that you could *own* software was the new thing.
Inevitable... (Score:2)
Silver lining? (Score:2)
IMO, the damage this trend will inevitably cause will be partly justified if it ends up killing companies such as Adobe. It might be worth living with buggy, spastic, unmaintainable shit just to see those fuckers go down in flames.
AI, or margins? (Score:2)
Is it AI driving this? Or is it just the fact that if your gross margins are 70-90% there's a lot of difference between the value you're providing and the price you're charging and your customers are recognizing that? Because I've done tests with Claude and while it might be acceptable for boilerplate code it still falls on it's face with complex business requirements and logic unique to specific internal processes that don't appear elsewhere.
Has any of these people actually used an AI agent (Score:2)
for customer service? Because it looks good on paper to your boss. But all the AI agents I've seen are mostly unusable to get anything done, and are basically a shield for customers to get caught in so the company doesn't have to do any real customer service because they can't access the company. All of the customer facing AI agents I've interacted with result in me pressing '0' on the phone or asking for a human agent, not one has fixed my issue or answered my question.
But hey, this is good for business fo
On prem coming back? (Score:1)
Regulated business (Score:2)