The "SaaS is dead" crowd is confusing two very different things.
Figma's stock is down 85% from its peak. Designers on Threads are calling it a "SaaSpocalypse." Satya Nadella says AI agents will "collapse" business applications. A16z is telling founders to stop building traditional SaaS.
If you only read headlines, you'd think every software company should shut down and let ChatGPT handle it.
Here's the problem with that take: it confuses two completely different categories of software, and the distinction matters if you run a creative team.
What's actually dying
Generic, horizontal SaaS - the kind that does one broad thing for everyone - is under real pressure. Content generation tools, basic chatbots, simple analytics dashboards, commodity project management. If an LLM can replicate your core function in a single prompt, you have a problem.
The numbers back this up. HubSpot is down 51%. Monday.com down 36%. Atlassian down 34%. These are tools where AI legitimately reduces the number of seats you need. If one AI agent does the work of ten marketers, you don't need ten HubSpot licenses.
This is real. We're not going to pretend it isn't.
What's actually thriving
Vertical software - tools built for specific industries with deep domain knowledge - is growing 2-3x faster than horizontal SaaS. The global SaaS market hit $300 billion in 2025 and is projected to reach $1.5 trillion by 2034. That's not a dying market.
The winners share a pattern: they solve coordination problems that require deterministic accuracy, domain expertise, and multi-stakeholder workflows. Snowflake is up 47%. CrowdStrike up 51%. MongoDB up 70%. Palantir up 142%.
AI agents need infrastructure to run on. They need structured data to work with. They need someone to have already mapped out the workflows, permissions, and business rules. That "someone" is vertical software.
The vibe coding reality check
"But anyone can build an app now with AI."
Sort of. You can vibe-code a prototype over a weekend. You can build an internal dashboard. You can automate personal workflows. This is genuinely powerful.
What you can't do: build production software that handles security, compliance, multi-user permissions, relational data propagation, and long-term maintenance - by describing it in a chat window. NYU research found 40-62% of AI-generated code contains security flaws. Non-technical users regenerate code until it works without understanding why it broke. That's fine for a side project. It's not fine for a system your team relies on every day.
Rapid creation is getting commoditized. The judgment to build something that actually works at scale is getting more valuable.
The real problem AI doesn't solve
Here's what we see every day working with creative teams at brands like Estee Lauder, Amazon, and Zara:
Production coordination is still run on spreadsheets.
Not because people are behind the times. Because the problem is genuinely hard. A single campaign touches production, post-production, marketing, and external partners. Each team maintains their own tracker. Those trackers constantly fall out of sync. Someone updates a delivery status in one spreadsheet, but the other three don't know about it. Assets get lost between Dropbox folders. Feedback lives in email threads nobody can find.
This is a coordination problem, not a content generation problem. AI can write your ad copy. It cannot reconcile four spreadsheets maintained by four different teams across three time zones with different access levels and approval chains.
That's the job. That's what vertical creative operations software does. Update a brief once - sizes, status, assets, and assignments propagate everywhere. One source of truth that everyone works from.
You're not going to vibe-code that over a weekend. And even if you could, who maintains it when a field type changes, a new stakeholder joins, or your compliance requirements shift?
The line between what AI replaces and what it can't
The useful framework is deterministic vs. probabilistic:
-
Probabilistic systems (chatbots, content generation, recommendations) - AI replicates these at a fraction of the cost. If your software is "correct 6 out of 10 times and that's fine," AI eats it.
-
Deterministic systems (production tracking, asset management, approval workflows, financial reconciliation) - these require 100% accuracy. "The brief says 5 sizes were delivered" is either true or it isn't. There's no room for hallucination.
Creative operations lives in the deterministic world. The brief is either approved or it isn't. The assets are either delivered or they aren't. The status is either synced or it's out of date. You need software that gets this right every time, not a language model that's pretty sure.
What this means for creative teams
If you're running production at a brand doing high-volume content, here's the honest assessment:
AI will help you with: Generating copy variations, creating rough concepts, transcribing feedback, summarizing meeting notes, automating repetitive formatting tasks.
AI won't help you with: Knowing which of your 47 active briefs is blocked, which assets were approved vs. still in review, whether the photographer delivered all five size variants, or why the marketing team's tracker says "complete" while production's says "pending."
That second list is where creative teams actually lose hours every week. It's the coordination tax - the invisible overhead of keeping everyone aligned when the work spans multiple people, tools, and timelines.
SaaS isn't dead. The wrong kind of SaaS is.
The software that survives - and grows - will be built by people who've actually run the operations they're building for. Who know where the handoffs break because they've been the ones chasing down missing assets at midnight before a launch.
Generic tools that try to be everything for everyone are getting squeezed from both sides: AI from below, and specialized vertical software from above.
If you're still coordinating creative production across spreadsheets, email threads, and shared folders - the tools that replace that aren't getting killed by AI. They're the ones AI can't replicate.
Spark Projects builds production coordination software for creative teams. Our founders have 20+ years of experience running content operations for brands including Amazon, Estee Lauder, Armani, and Zara. We build this because we lived it.