AI Agents Are Going to Make Half Your SaaS Stack Redundant - and the Vendors Already Know It

April 1, 2026

I want to talk about something that's been sitting in the back of my mind since I read through our last software audit results and then spent about an hour afterward just staring at the ceiling. Harold came in at some point, asked if I wanted tea, and I said yes without actually registering the question. He came back twelve minutes later with the tea. I still didn't move. That's how much this was bothering me.

The thing is: the headline isn't new. People have been murmuring about AI replacing SaaS tools for at least two years. But at some point in the last six months or so, it stopped being a murmur. The term "SaaSpocalypse" emerged in February 2026 to describe the rapid decline in traditional software-as-a-service company valuations as AI agents began replacing entire product categories - and the shift accelerated after multiple enterprise software companies reported slowing growth and acknowledged AI agent competition directly on earnings calls. These aren't niche blogs saying this anymore. This is earnings-call language.

And what bothers me most - the thing that made me forget about the tea - is that the vendors have known. They've known for a while. The panic you're seeing in stock prices right now isn't surprise. It's delayed acknowledgment.

Let's Talk About What's Actually Happening

Unlike traditional SaaS applications that require users to click through interfaces, fill out forms, and manually execute workflows, AI agents operate as autonomous systems capable of reasoning through problems, making decisions, and taking action without constant human oversight. They understand natural language commands like "analyze our Q2 performance" - eliminating the need for users to learn complex navigation paths through multiple applications.

That's the thing people keep underselling. This isn't about AI making your existing tools smarter. It's about AI replacing the reason you open the tool in the first place. AI agents can now perform tasks that previously required dedicated SaaS tools - project management, CRM updates, report generation, customer support triage, and meeting scheduling. The critical distinction is that these agents don't operate within the SaaS tool. They replace the need for the tool entirely.

I've been watching this with our own stack. We've got tools we're paying for monthly that, when I actually trace back the last quarter of activity, are being used for things an agent could do faster with less friction. Not hypothetically. Actually. Right now.

In three years, any routine, rules-based digital task could move from "human plus app" to "AI agent plus API." SaaS providers know this strategic problem is urgent. That sentence came from Bain. And the key part isn't the three-year prediction. It's the second sentence. They know.

The Seat-Based Model Was Always a Fragile Thing

The whole architecture of modern SaaS was built on a beautifully simple idea: count the humans, charge per head. More employees meant more seats meant more revenue. It scaled elegantly in every direction. This model, known as SaaS 1.0, built empires. It turned software from a lumpy, expensive capital expenditure into a smooth, predictable operational expense. In return, vendors received something even more valuable: Annual Recurring Revenue with massive gross margins.

The problem is that AI agents don't need seats. Multiple SaaS companies reported slowing growth in Q4 2025, not because AI failed to boost productivity as expected, but precisely because it succeeded too well. Customers are reducing software seats rather than adding them, as AI-enhanced workers accomplish more with fewer licenses. That's the irony that I find genuinely darkly funny. The vendors sold AI as a feature to boost productivity. And the productivity gains are now eating their own revenue model.

The market numbers are not subtle about any of this. ServiceNow stock is down 36%, HubSpot stock is down 51%, and Monday stock is down 44%. Atlassian (-35%) and Salesforce (-28%) led declines as their core workflows - task tracking, data entry, customer logging - are exactly what AI agents automate best. We actually have a piece on how we saw some of this coming - HSBC called it the SaaSpocalypse before it had a name, and I think they were more right than even they expected to be.

Nate saw the HubSpot number and said it was just a market rotation thing, that the fundamentals were still fine. I didn't argue with him because it was a Thursday and I had a casserole in the oven at home I needed to get back to. But he was wrong. This isn't a rotation. This isn't a temporary correction driven by rotation into other sectors or macroeconomic fears. It represents a structural shift in how businesses think about software.

A long row of identical ergonomic office chairs in precise perspective, most empty and slightly askew, one near the front occupied by a faint human figure, rendered in thin lines and muted grey tones against a white background — Wanted something that showed the seat problem without having to explain it. This is accurate enough.

Microsoft Said the Quiet Part Out Loud

Here's where I stop being patient with the "this is overhyped" crowd. When Microsoft CEO Satya Nadella first indicated that business SaaS applications are dead last December, it sent shockwaves through the enterprise software world. Now Charles Lamanna, Microsoft's corporate vice president leading business applications and platforms, is doubling down on this vision with a timeline and roadmap that is both ambitious and controversial. He said traditional business applications will become the mainframes of the 2030s: still running, still consuming budgets, but ossified relics of a bygone era.

This is a Microsoft executive. Talking about enterprise software becoming a relic. Microsoft makes enterprise software. When the people selling you the product start describing it as a transitional fossil, you should probably believe them.

And then there's Publicis Sapient - a company that is not some scrappy AI startup with an agenda - which reports actively reducing traditional SaaS licenses by approximately 50%, including major platforms like Adobe, by substituting them with generative AI tools and chatbots. An executive at the firm explains that AI agents are "10x faster, 100x smarter" than junior staff, creating a redundancy that directly cannibalizes the seat-based revenue underpinning commercial SaaS models. Half. They cut half their SaaS spend. And they're a digital transformation consultancy. These are people whose job it is to know what tools work.

There's also Apollo Global Management, which recently revealed they slashed their software exposure in private credit funds from 20% to 10%, effectively betting against the durability of the SaaS model. When a major private equity firm cuts its software bet in half, that's not a bet on hype. That's a bet on fundamentals.

Not Everything Is Going to Die - But the Middle Is

I want to be precise here because I think the "SaaS is dead" framing, while satisfying, gets the details wrong in a way that matters for actual business decisions.

Some categories are genuinely resilient. Things like payment processing and other core infrastructure are pretty safe. You're not going to replace Stripe and all their engineering work on core payments easily with an agent. Compliance tools, security infrastructure, systems with serious regulatory overhead - these aren't going anywhere fast. Security, compliance, and infrastructure tools remain less immediately vulnerable. These categories require human oversight, regulatory compliance verification, and real-time threat response where the consequences of AI errors are too severe for full automation.

What's getting hollowed out is the middle. The point solutions. The single-workflow tools. The things you bought because they did one thing and did it well enough that paying $89 a month felt reasonable. For the majority of "point-product" SaaS companies, the challenge is survival.

Think about how many tools in your stack exist purely as interfaces between a human and a database. The tool's value was never really the logic - it was the UI. The buttons and the dashboards and the onboarding that made the underlying data accessible. Before, a SaaS product's competitive edge was directly proportional to the quality of its user interface, ease of use, specialized features, and brand reputation. Yet, if AI now handles all interactions, SaaS could lose its front-end relevance, reducing its differentiation to data storage and process execution.

When I looked at how we were actually using some of our cold outreach tooling last year - the way I described it in our piece on cold email software - I kept noticing that the "smart" features were things an agent wrapper could do more flexibly. The tool wasn't providing unique intelligence. It was providing a box to put the intelligence in. That distinction is about to matter a great deal.

The Vendors' Response Tells You Everything

Watch what the vendors are actually doing, not what they're saying on investor calls. ServiceNow announced it would acquire AI agent platform Moveworks for $2.85 billion in March 2025, signaling that legacy software incumbents are making massive strategic bets on agent technology as core to their future platform strategy. Salesforce spent $1.2 billion acquiring AI startups in the past six months, racing to transform from a database company into an AI platform.

You do not spend $1.2 billion acquiring AI companies because you think your existing model is fine. You do it because you've run the numbers on what happens if you don't, and those numbers scared you.

Vendors are already starting to rethink pricing for an AI-first world. Seat-based pricing may not fit when AI is doing the work. If an agent replaces a human task, customers will expect to pay based on outcomes, not log-ons. Start experimenting with pricing tied to results: tasks completed, tickets resolved, AI outputs generated. Leaders such as Intercom and Salesforce are already shifting in this direction. The fundamental shift is to stop charging for access and start charging for work done.

HubSpot is actually the most interesting case study here. While everyone else was getting demolished, it posted 20% year-over-year revenue growth, fueled by its aggressive pivot to an "agentic customer platform." This performance stands in stark contrast to legacy competitors who are struggling to justify their seat-based pricing models in an era where AI increasingly performs the work of human employees. How? HubSpot introduced "HubSpot Credits," a consumption-based model that charges customers based on the volume of work performed by agents rather than the number of human logins. They stopped charging for access. They started charging for work done. And it worked.

Cal was very enthusiastic about this when I mentioned it. He's going through a lot right now - the divorce was finalized, and apparently the car situation still isn't resolved - but he lit up talking about consumption-based models like they personally offered him some kind of comfort. I didn't push it. Sometimes you let people have the thing that's working for them.

What This Actually Means If You Run a Business

My honest read: you're currently paying for interfaces you don't need, and the vendors know it. The scramble you're watching in the stock market isn't investors suddenly discovering a new risk - it's SaaS companies having spent two decades building moats around user habits, data lock-in, and workflow integration, while AI agents are now eroding each of these advantages simultaneously.

This is why we ran a software audit last year and why it went sideways almost immediately. Everyone has tools they've built habits around, and habits feel like necessity until they don't. The audit forced us to ask which tools we were actually using for their logic versus which ones we were using because they had a button we'd gotten used to clicking. More of the latter than I expected.

I'm not going to tell you to cancel everything and rebuild on agents tomorrow. Organisations with no real software knowledge are not going to suddenly replace their entire SaaS suite. That's true and I'd be annoyed if someone told me otherwise. But I do think it's worth being honest with yourself about which tools in your stack would survive the question: "if an agent could do this, what would I actually be paying for?"

For a lot of point solutions, the answer is: the familiarity. The login screen. The onboarding doc nobody read except me.

Harold asked me at dinner a few weeks ago why I keep talking about software companies like they're people going through something. I told him it's because in a way they are - they built something that worked for twenty years and now the ground is shifting under them and some of them are going to handle it with grace and some aren't. He thought about that for a moment and then asked if there was more chicken. There was. I could tell he was thinking about what I said though.

The vendors who are going to make it through this are the ones charging for outcomes, not access. The ones that own the data, the compliance layer, the logic that can't just be swapped out. Everyone else - the interface-heavy, workflow-rigid, seat-counting middle of the market - is in a more precarious position than their current pricing suggests.

We've been watching the agentic AI shift move from boardroom buzzword to actual strategy for a while now. The part that nobody wanted to say plainly is that this transition doesn't just affect how you work. It affects which software companies survive. And the vendors have been sitting with that knowledge for longer than they let on. The acquisitions give it away. The pricing model pivots give it away. The earnings call language gives it away.

Half your SaaS stack being redundant isn't a 2030 prediction. It's a description of what's already starting to happen. The only question is whether you notice before or after the renewal invoice arrives.