I thought the internet moved past the "SaaS is dead" takes months ago. Apparently not!
"SaaS is dead" is a good take in that it generates engagement and points to real shifts happening in tech, but at this point it's stale and it's always been wrong.
Here’s why the arguments for “SaaS is dead” don’t hold up:
- The argument is so broad as to be (almost) useless
- The argument is internally inconsistent
- Sufficiently valuable software is very complex
- Comparative Advantage Still Matters
- GenAI ➡️ More Software ➡️ More SaaS, Not Less
- SaaS Companies Are Riding the Wave
- Performant AI Agents Are Neither Simple nor Isolated
Think "SaaS + AI" not "SaaS vs. AI."
The argument is so broad as to be (almost) useless
Before even getting into the substance of the "SaaS is dead" argument, we have to acknowledge that "SaaS"/"Software" are frustratingly vague terms.
They are broad enough to encompass everything from massive enterprise platforms like Salesforce down to the simplest note taking apps.Without precision, the debate is muddled from the start, making substantive discussion a challenge.
I'll admit that this post isn't immune from that problem either.
While I try to engage with the argument on its merits, I’m still using "SaaS" as a catch-all for a vast and diverse category.
But that’s part of the point: any sweeping claim that "SaaS is dead" is inherently flawed because it lacks the nuance needed to say which parts of SaaS are supposedly dying and why.
Alright, onwards!
The argument is internally inconsistent
The second problem with “SaaS is dead” takes? They contradict themselves.
Think about everyone posting "SaaS is Dead" and the tools they are using to create, distribute, and monetize those posts:
- The author probably used ChatGPT. Guess what that is? SaaS.
- Maybe they got a little extra editing helping using Lex.page. Paying $12/mo for that service.
- They may pay for verification on X/Twitter. SaaS again.
- Recorded a short-form video? SaaS.
- Asking you to subscribe to their newsletter/account? SaaS SaaS SaaS.
The entire stack enabling these arguments is, ironically, built on SaaS!!
So unless GenAI obsoletes every single one of these layers, something no credible argument suggests, SaaS isn’t going anywhere.
And that obsoletion won't happen because sufficiently valuable software is very complex, which brings us to the next point.
Sufficiently valuable software is very complex
People arguing that GenAI will replace SaaS misunderstand what makes software valuable.
As Benedict Evans recently put it on Threads, "Writing code is not the hard part of most enterprise software." That point extends to consumer software as well.
GenAI might be able to create the code needed for Hubspot, Monday, AirBnB, Plaid, Stripe, etc... but that's something like 10% of what businesses need to succeed.
The real complexity and value comes from everything else:
- Deeply understanding a problem and having a unique point of view on how to solve it
- Ongoing maintenance, improvements, reactions to market changes
- Regulation and compliance
- Globalization
- Security and risk management
- Third party integrations
- Managing and adapting to human unpredictability
A 2021 post from Slack, Reducing Slack’s memory footprint, illustrates just how much engineering effort goes into a simple decision like whether to notify a user about a message:
Maybe a GenAI tool will eventually handle all of this on its own with no human input.
But even if it could, why would most companies invest the effort to build and maintain it themselves?
Comparative advantage still matters.
Comparative Advantage Still Matters
Even as GenAI lowers the barrier to creating software, most people and businesses won’t build their own tools. That’s not how comparative advantage works.
Consumers
It is obvious that the vast majority of people in the world have no interest in building software solutions for themselves. They don't have the time, energy, desire, or expertise to do so.
Paying $1, $2, or $50 a month for a fully managed solution that works out of the box is a far better deal than spending hours trying to build and maintain a DIY alternative.
It’s why basic, off-the-shelf apps continue to thrive, even though consumers could learn to build their own.
We can take this a step further to make the point and ask
- How many people set up Apple Shortcuts to automatically unsubscribe from political texts? Almost none.
- How many people even change their default web browser or search engine? Again, very few.
If most people won’t take 10 seconds to change a setting, they certainly aren’t going to build their own software from scratch, even with the help of GenAI.
And why would they? SaaS is wildly pro-consumer.
- You get the latest and greatest improvements, fixes, security, customer support into infinity
- You get to cancel on a whim for any reason
- You get access to world-class software for a few bucks a month instead of paying thousands upfront.
That's a screaming deal.
Subscriptions are amazing and have democratized access to world-class software.
Long live subscriptions.
Businesses
If SaaS is a great deal for consumers, it’s an even bigger no-brainer for businesses.
The build vs. buy equation may shift in some cases due to GenAI, but most businesses will still choose buy over build.
Why?
- SaaS lets them offload complexity (maintenance, compliance, security, integrations, etc.) so they can focus on their actual core competencies.
- The cost of buying software is almost always cheaper than dedicating internal teams to build and maintain it.
- SaaS providers specialize in their domain, meaning they move faster and innovate more than an in-house team would.
Even if GenAI lowers the barrier to custom software, why would a payments company want to build its own CRM?
Time/energy is better spent buying a solution that is built/maintained by a separate entity so they can focus on whatever their core thing is.
Businesses aren’t looking to reinvent the wheel. They want best-in-class solutions that just work.
GenAI ➡️ More Software ➡️ More SaaS, Not Less
The friction of creating software has been dropping for decades. Every wave of innovation that made programming easier didn’t reduce the number of developers.
Instead, each innovation led to more software, more ambitious projects, and more innovation.
As one Meta engineer recently put it:
For the last 50 years we've been making programming easier, and it has never meant fewer devs. It always means we build even more stuff. More code. More ambitious projects. More wild ideas come true.
Small projects became trivial, complex projects become cheap, and impossible projects begin
Engineering productivity drives costs down, which boosts the entire economy. It doesn't reduce the demand for software.
GenAI is going to contribute to this trend, and we're seeing the early iterations with Cursor, Replit, and so on.
By making it easier and cheaper to build software, it won’t shrink the SaaS market, it will expand it. Lower development costs mean more SaaS businesses, not fewer.
In 2017, Alex Danco posed a simple but powerful question in his essay series Understanding Abundance :
What happens when friction goes away?
Alex's answer is that we get two types of businesses:
- Pointy Businesses - that focus purely on differentiation, relying on utilities for everything else.
- Utility Businesses - that provide core infrastructure (e.g., AWS, Stripe).
I think that GenAI will accelerate this trend.
By making software creation easier, it will lead to an explosion of Pointy Businesses, niche SaaS products that focus on solving hyper-specific problems while outsourcing infrastructure to existing platforms.
In practical terms, may 1,000,000 new note taking apps flourish!
SaaS Companies Are Riding the Wave
I stand by Sustaining Today, Disruptive Tomorrow and think that GenAI / Agents aren't going to replace SaaS, though I do think they'll extend it.
The most effective AI agents won’t exist in isolation. They’ll be deeply integrated into existing SaaS platforms. Given that assumption, current SaaS platforms are positioned well to take advantage of the GenAI wave.
They have
- Users actively engaging with their products
- Underlying data that provide the necessary context for GenAI tools.
- Human input loops that keep the data fresh, ensuring GenAI inputs stay relevant and accurate.
a16z has a great blog post, Pricing and Packaging Your B2B or Prosumer Generative AI Feature, showing that this is the case today.
I also think that any SaaS built from ground up to make an outcome-based agent effective could have massive advantage.
If I were launching a startup today, this is where I’d focus. It's a worthy bet.
Performant AI Agents Are Neither Simple nor Isolated
Okay, so now we're at the point in the argument when someone says, "But AI agents are going to do all the work for us!"
I don't think that's the case, at least in the near term. Performant AI agents are not simple.
On a recent Oct. 2024 Stratechery Interview, Marc Benioff of Salesforce described the what he thinks is needed for such a product: - LLMs - "You want an LLM, but that’s just like sugar. That’s going to make it great,"
- RAG- "you’re going to have to have a pretty advanced RAG technique or an ensemble RAG technique"
- Data - "you’re also going to need data because LLMs by themselves don’t know anything, they’re just an algorithm"
- Metadata - "you’re also going to want metadata because the data is better if it’s explained to the LLM what the data and metadata are"
- Hyperscaler approach - "You’re going to want some kind of hyperscaler approach so that you can scale up and down as you need to go.
- UI - "You need to have some kind of user interface with the customer where your employee is participating because while all this is great, no one’s LLM, no one’s capability...is 100% accurate yet, it just isn’t...you’re going to want humans to be able to mitigate and be part of the process or keep the humans in the loop and that means the apps have to be integrated as part of the process."
I think AI agents will do some work for us (i.e. Deep Research, Coding), but the most complex tasks are going to need humans & AI working together.
In other words, the very SaaS platforms that AI is supposedly going to replace are the ones that provide the data, infrastructure, and human oversight that AI needs to function accurately.
AI-driven automation is happening, but it’s unfolding in a way that extends SaaS rather than eliminates it.
Instead of making SaaS obsolete, GenAI is making SaaS platforms even more valuable because such products will be the gatekeepers of data, workflows, and user input.
Okay, so now we're at the point in the argument when someone says, "But AI agents are going to do all the work for us!"
I don't think that's the case, at least in the near term. Performant AI agents are not simple.
On a recent Oct. 2024 Stratechery Interview, Marc Benioff of Salesforce described the what he thinks is needed for such a product:
- LLMs - "You want an LLM, but that’s just like sugar. That’s going to make it great,"
- RAG- "you’re going to have to have a pretty advanced RAG technique or an ensemble RAG technique"
- Data - "you’re also going to need data because LLMs by themselves don’t know anything, they’re just an algorithm"
- Metadata - "you’re also going to want metadata because the data is better if it’s explained to the LLM what the data and metadata are"
- Hyperscaler approach - "You’re going to want some kind of hyperscaler approach so that you can scale up and down as you need to go.
- UI - "You need to have some kind of user interface with the customer where your employee is participating because while all this is great, no one’s LLM, no one’s capability...is 100% accurate yet, it just isn’t...you’re going to want humans to be able to mitigate and be part of the process or keep the humans in the loop and that means the apps have to be integrated as part of the process."
I think AI agents will do some work for us (i.e. Deep Research, Coding), but the most complex tasks are going to need humans & AI working together.
AI-driven automation is happening, but it’s unfolding in a way that extends SaaS rather than eliminates it.
Instead of making SaaS obsolete, GenAI is making SaaS platforms even more valuable because such products will be the gatekeepers of data, workflows, and user input.
Conclusion
SaaS isn't dead, but it is evolving.
Think "SaaS + AI" not "SaaS vs. AI."
Just like every past innovation that lowered the cost of software creation, GenAI will lead to more software, more businesses, and therefore more demand for SaaS, not less
What’s changing is how SaaS delivers value.
.jpg)
