Over $1 trillion in market cap was wiped from software stocks in the first week of February. Price-to-sales ratios for SaaS companies compressed from 9x to 6x, levels we haven't seen since the mid-2010s. Twitter (or whatever we're calling it) declared SaaS dead. Forrester coined "SaaSpocalypse." Every AI influencer posted some version of the same take: agents will replace your software stack.
I build enterprise AI tools for a living, and I find this argument completely unconvincing. Not because SaaS isn't changing. It absolutely is. But the "AI replaces SaaS" narrative gets the relationship backwards.
SaaS didn't die. It got demoted. From the thing you interact with to the thing your AI interacts with. And that demotion actually makes it more important, not less.
Think of It Like a Kitchen
Bain published a framework recently that I think nails what's actually happening. The easiest way to understand it: think about a restaurant kitchen.
At the bottom, you have the pantry. The walk-in fridge. The ingredients. That's your SaaS layer. BuildingConnected with 700,000+ subcontractor contacts. Procore with your project history. Your CRM with every client interaction. The raw material that makes everything else possible.
In the middle, you have the chef. That's the AI agent. It takes ingredients from the pantry, follows a recipe, combines things in the right order, and produces a finished dish. An agent that automates your bid outreach is pulling contacts from BuildingConnected, drafting emails, scheduling follow-ups, tracking responses. It's doing the work. But it's working with ingredients that already exist.
On top, you have the plate that lands on the table. That's the interface. The thing the customer (or in our case, the user) actually sees and interacts with. That part is changing. Instead of clicking through dashboards and menus, you might just tell the chef what you want.
The "SaaS is dead" crowd is looking at the plate changing and concluding that the pantry is irrelevant. But fire the chef and swap out the plates all you want. If the pantry is empty, nobody eats.
The Stock Market Is Panicking. The Buyers Aren't.
Here's the part that gets buried under the doom scrolling.
Gartner's February 2026 forecast projects worldwide software spending will grow 14.7% this year to more than $1.4 trillion. That's roughly $180 billion in net new software spending in a single year. Global SaaS spending specifically is projected to rise from $318 billion in 2025 to $576 billion by 2029.
So Wall Street erased a trillion in value while actual enterprise buyers are increasing their software budgets by double digits. Those two things can't both be right. My money is on the people writing the checks, not the people writing the headlines.
Where AI Actually Replaces Software (And Where It Doesn't)
I'll grant the SaaS-is-dead crowd one thing. There are categories where AI will eat significant market share. The tools that never specialized in anything specific are vulnerable. Canva is a good example. Claude can generate a solid graphic now. If your use case was "I need a quick social media image and I don't want to learn Photoshop," AI handles that. The core value proposition of "design tools for non-designers" gets weaker when the AI is the non-designer.
But Photoshop and Illustrator aren't going anywhere. Real creatives need real tools with precise control over layers, masking, color profiles, export settings. AI makes those tools more powerful (look at Adobe's generative fill). It doesn't replace them.
The same logic applies to the generative side of document work. Yes, Claude can build me a presentation. But I still want PowerPoint (or a tool purpose-built for presentations) because I need to open it up and make the tweaks. Move a chart. Adjust the spacing. Fix the one slide where the formatting is slightly off. AI generates the first draft. The specialized tool handles the last mile.
And yes, Claude can produce an advanced CSV. But I still need Excel because I need to interact with the data. Filter it. Pivot it. Build formulas that reference other tabs. My estimator has 47 interconnected sheets with conditional formatting that flags when a sub's bid comes in 20% below the next lowest, with macros that auto-populate the bid summary. That's years of institutional knowledge encoded in a tool. AI makes it more powerful. It doesn't replace it.
The Agentic Side Is Where the Argument Falls Apart
The generative side is at least a real debate. You can make the case that for simple use cases, AI replaces the tool. Fine.
But on the agentic side, the "SaaS is dead" argument completely falls apart. Because agentic AI only works when the data already exists somewhere.
Every AI agent that's done anything worth talking about pulls data from existing SaaS tools. That's the whole architecture. The agent connects to your systems, reads the data, takes action, and writes back. Without the system, there's no data to read and nowhere to write.
Take the construction world. I'm building tools that automate subcontractor outreach for bid collection. To do that, you need a platform like BuildingConnected where the contact data exists. You need the subcontractor profiles, the trade classifications, the prequalification records, the project history. You need the database. The agent orchestrates on top of it, but it can't conjure contacts out of thin air.
This is true across every industry. An AI agent managing your sales pipeline needs your CRM data to exist first. One automating project workflows needs your project management platform. The agent is the orchestration layer. The SaaS tool is the data layer. One doesn't work without the other.
MCP and the New Integration Arms Race
There's a technical shift happening right now that makes this even more concrete. Anthropic released something called the Model Context Protocol (MCP), an open standard for how AI agents connect to external data sources. It's becoming the de facto way agents plug into software.
The adoption numbers are significant. Over half of integration platform vendors are expected to adopt MCP by end of 2026. Three quarters of API gateway vendors will integrate MCP features. SDK downloads are at 97 million+ per month across Python and TypeScript.
What this means practically: every SaaS tool that wants to survive in an AI world needs to become an MCP server or expose clean APIs that agents can work with. The tools that do this become more valuable, not less. They become the connective tissue that AI agents depend on.
The tools that don't? The ones with closed ecosystems and no integration points? Those are the ones that are actually dead.
This creates a new competitive dynamic. The question for a GC evaluating software used to be "which tool has the best features?" Now it's "which tool can my AI agent plug into?" The SaaS products with the best integrations and the most open architectures will win. The ones that treat their data as a walled garden will lose to competitors that let AI work on top of them.
What This Actually Means for Buyers
If you're a GC or any mid-sized company trying to figure out where AI fits into your operations, here's the practical takeaway.
You should be investing in good SaaS tools right now. Not fewer tools, better tools. Specifically, tools with strong APIs, integration ecosystems, and (increasingly) MCP server support. Because when you're ready to layer AI agents on top of your workflows, the companies with clean, connected data across good platforms will be the ones that get results. The companies that waited for AI to replace their software will be sitting there with no data for the AI to work with.
The construction industry already has a massive data fragmentation problem. 96% of captured construction data goes unused. Only 5.5% of contractors have full integration across their tools. Nearly 30% have no software integration whatsoever. If AI is going to unlock value in this industry, it needs something to unlock. The SaaS layer is what gives it something to work with.
SaaS Isn't Dead. The Interface Is Changing.
The way you interact with software is changing. Instead of clicking through dashboards and filling out forms, you'll increasingly talk to an agent that does it for you. The interface gets replaced. The data layer underneath it stays.
If anything, the AI revolution is making a case for spending more on your software stack, not less. Better data in means better AI output. Cleaner systems mean smoother automation. More integration points mean more surface area for agents to do useful work.
SaaS got demoted from the thing you look at to the thing that powers everything behind the scenes. Back to the kitchen analogy: nobody goes to a restaurant because of the pantry. But every great meal starts there.
The companies scrambling to rip out their software and replace it with AI are emptying the pantry and wondering why the chef can't cook. The companies that will win are the ones stocking better ingredients, organizing them well, and letting the chef do what it does best.
SaaS isn't dead. It just stopped being the thing you see and started being the thing that makes everything else work.