SaaS Isn't Dying. The UI Layer Is. Here's What Comes Next.

The SaaS panic is real. The conclusion most people are drawing from it is wrong.

Software stocks are down 25%. Everyone's calling it the death of SaaS. They're half right. SaaS as a UI business is dying. SaaS as a business logic business is just getting started.

SaaS Isn't Dying. The UI Layer Is. Here's What Comes Next.

Software stocks are down 25% from their highs. The Goldman Sachs Software Index fell 30% from October 2025 peaks. Salesforce dropped 26%. Monday.com fell 20% in a single session. Freshworks is down 37% year-to-date — its IPO investors have lost 85%.

Jefferies coined the term "SaaSpocalypse." InvestorPlace called it "SaaSmageddon." Every analyst, pundit, and LinkedIn thought leader is calling it the death of SaaS.

I think they're half right.

SaaS as a UI business is dying. SaaS as a business logic business is just getting started.

And the difference between those two things is the biggest opportunity most people are missing right now.

---

Someone just proved the thesis in public

John Rush — a founder who runs what he calls "the most automated org on earth" — posted something on LinkedIn recently that should terrify every SaaS CEO and excite every service business owner.

Over the past 30 days, he replaced his logins to Stripe, his bank, Gmail, and 30 other SaaS tools. Instead, he connected an AI agent to all of them via MCP servers. Now he wakes up, opens one chat, and says: "Show me yesterday's revenue, any failed payments, and all questions and bug reports from my customers." Done. No 14 tabs. No context switching. No remembering which dashboard has which chart.

His key insight was brutal: "The key reason I chose the products I did was their UX. On the database level, all products are pretty much the same, or even if different, they cover the same use cases."

Read that again. He chose products for their interfaces. Now those interfaces are irrelevant. The traditional UI was just a layer between him and the database. AI chat does that job better.

It gets worse for SaaS incumbents. He only uses about 5% of any given tool. So he picked the 5% from each product and vibe-coded his own unified dashboard. All in one place. Only the things he needs.

His conclusion: "The UX moat is dead. Brand loyalty is gone. Switching costs are gone. Tech has no defensibility anymore."

---

Where John's analysis stops and the opportunity starts

John is completely right about the UI layer. If the primary value of your software is that it presents data in a nice dashboard, you're in serious trouble. AI agents will generate better, more personalised dashboards on the fly. The interface was never the product — it just felt like it was.

But here's what his analysis misses.

Not all SaaS is a UI layer on top of a commodity database. Some SaaS products have genuine intelligence baked in. A compliance platform doesn't just store risk data — it has five years of regulatory logic encoded in its rules engine. A recruitment SaaS doesn't just hold CVs — it has candidate scoring models trained on millions of placements. A marketing platform doesn't just track campaigns — it has attribution models refined through billions of data points.

An AI agent can't hallucinate that intelligence. Claude doesn't have your proprietary data. GPT doesn't know your domain-specific logic. These models are extraordinary at general reasoning. They're terrible at knowing the specific thing that makes your business valuable.

The value didn't disappear. It's just no longer accessible through the right channel.

Right now, most SaaS products lock their intelligence behind a login screen and a React dashboard. In an agent-first world, that's like having a gold mine with no road to it.

---

The real split: interface vs intelligence

The SaaSpocalypse selloff wasn't random. It targeted specific categories. Customer support tools, project management, SMB CRM, marketing automation, document management — these were hit hardest. What do they have in common? They compete primarily on UI and feature count.

The categories holding their value tell the opposite story. Cybersecurity is actually growing — more AI agents means more attack surface. ERP systems are too deeply embedded to replace — AI agents use ERPs, they don't displace them. Accounting software has regulatory moats around tax filing and compliance.

Bain put it clearly: the "unbundling of interaction from the underlying platform" is the most profound shift since SaaS unbundled features from on-premise software. Your data is your moat. The real value lies in proprietary data, usage patterns, domain-specific content, and transaction history.

Jason Lemkin at SaaStr added the pricing dimension: "AI isn't eating the product. It's eating the budget." If AI agents reduce a 10-person sales team to 2 people, that company needs 2 CRM seats, not 10. Per-seat pricing collapses when agents don't need seats.

The market is repricing. It's saying: we no longer pay a premium for interfaces. We pay for intelligence, data, and business logic that AI agents can't replicate.

---

What this means in practice

The fix is simpler than most people think. Not easy — but simpler than the scale of the panic suggests.

If your software has proprietary business logic, domain-specific data, or expert methodology encoded in it, you don't need to panic. You need to expose it.

MCP servers — Model Context Protocol — are the mechanism. Think of MCP as a universal adapter that lets any AI assistant plug into any data source or service. Anthropic created it. OpenAI adopted it. Google, Microsoft, and Amazon followed. It was donated to the Linux Foundation. 97 million monthly SDK downloads. Over 17,000 servers listed. This isn't speculative infrastructure — it's the emerging standard.

When you build an MCP server for your business logic, you're making your intelligence accessible to AI agents. Your scoring engine, your compliance checks, your recommendation models — all callable directly by Claude, ChatGPT, or any MCP-compatible tool.

The result: you keep the UI for humans who want it. You add agent access as a new revenue stream from the same codebase. You become more essential, not less — because in an agentic world, the platforms with the best data and logic become the foundation everything else is built on.

---

The SaaS products that survive will look different

Here's what I think the next generation of valuable software looks like:

Intelligence layer first, interface optional. The business logic is the product. The dashboard is one way to access it. MCP is another. APIs are another. The intelligence doesn't care how it's consumed.

Usage-based pricing replaces per-seat. When AI agents are the primary consumers of your logic, per-seat doesn't make sense. You charge for what gets used — API calls, decisions made, data processed. This actually aligns incentives better than seat-based pricing ever did.

Data compounds. Every interaction through your MCP server generates more data. More data makes the logic better. Better logic attracts more usage. This is the flywheel that pure UI products never had.

Methodology becomes infrastructure. The specific way your business scores candidates, assesses risk, evaluates campaigns, or matches supply to demand — that methodology is the moat. Not because it's secret. Because it's been refined through thousands of real decisions and outcomes.

---

This is especially true for service businesses

I've spent 18 years building products, and a growing chunk of that time building for service businesses specifically. The pattern I keep seeing: service businesses have extraordinary intelligence locked inside their operations. A recruitment firm's matching methodology. A compliance consultancy's risk framework. A training provider's assessment model. A marketing agency's campaign evaluation system.

This intelligence currently lives in spreadsheets, in people's heads, and in manual processes. It's the reason clients pay premium rates. And it's exactly the type of intelligence that becomes more valuable as generic software gets commoditised.

When I built SetWise — a football card market intelligence platform — the entire system was designed to be MCP-native from day one. 13 custom AI tools, all exposed via MCP server. Any AI agent can query the market data, check prices, analyse trends, and compare players directly. The dashboard exists for collectors who want it. But the real value is in the intelligence layer that any AI tool can consume.

That's what the next generation looks like. Not SaaS or nothing. SaaS-as-intelligence, accessible through whatever interface — human or agent — makes sense for the moment.

---

Three things every SaaS founder and service business owner should be asking

If you run a SaaS product: Which parts of your product are pure UI (and therefore vulnerable), and which parts contain genuine proprietary logic? The UI parts need radical rethinking. The logic parts need MCP exposure.

If you run a service business: Is the methodology that makes your business valuable currently trapped in people's heads and spreadsheets? Because someone in your space is going to structure theirs first. The awareness gap between what MCP enables and what most business owners know about is still enormous — but it's shrinking fast.

If you're building something new: Don't build a dashboard with business logic behind it. Build business logic with an optional dashboard in front of it. Design for agent consumption first, human consumption second. The economics favour it and the infrastructure supports it.

---

The era of selling dashboards is ending

John Rush cancelled 30 SaaS subscriptions and replaced them with one chat window. He's not alone. The trend is directional and accelerating.

But the intelligence those tools sat on top of? The scoring algorithms, the compliance logic, the matching engines, the recommendation models? That didn't go anywhere. It just needs a new front door.

The SaaSpocalypse isn't the end of software. It's a repricing. The market is finally distinguishing between software that presents information and software that contains intelligence.

If your value is in the presentation layer, the market has spoken. If your value is in the intelligence layer, you've never been more relevant.

The question is whether that intelligence is accessible to the agents that are about to do all the work — or locked behind a login screen nobody opens.

---