The End of Easy SaaS

Software stocks lost $285 billion in a single day after Anthropic released a set of plugins for Claude Cowork. Not a new model. Not a chatbot upgrade. Plugins.

A plugin marketplace announcement erased more wealth than most industries generate in a year. Wall Street isn't scared of AI anymore. They're scared of what AI replaces.

But "SaaS is dead" is the wrong framing. What's actually happening is more specific, more interesting, and — if you're building — more useful to understand.

What's actually dying

For the last fifteen years, the SaaS playbook was straightforward: find a common business workflow, build a nice UI around it, add some integrations, charge per seat per month. Defend your position with switching costs and minor product tweaks.

This playbook minted fortunes. But it had a flaw nobody talked about.

Most of the value was never in the software itself. It was in the workflow the software organized. The UI was a middleman. And AI agents just made the middleman optional.

When an AI agent can log into your existing tools — your CRM, your documents, your databases — and perform entire workflows autonomously, the question changes. It's no longer "which dashboard should I use to manage this?" It's "why am I looking at a dashboard at all?"

The implications cascade. If ten agents can do the work of a hundred people, you don't need a hundred software seats. AI doesn't kill the software directly. It kills the headcount that uses the software. Which kills the per-seat revenue model. Which kills the business.

The thin middle squeeze

Picture three layers in the software stack:

Top: the agent layer. The thing that actually executes work. Reads data, makes decisions, takes actions. This layer is new and growing fast.

Middle: the SaaS UI. Dashboards, workflows, buttons you click. This layer has been the center of enterprise software for two decades.

Bottom: systems of record. Databases, CRMs, ERPs that store canonical data. This layer has been quietly essential forever.

Value is getting pulled upward into the agent layer and downward into the data layer. Everything in the thin middle gets squeezed.

That's why forward P/E ratios are compressing across the sector. Not because people don't need what these companies do — but because investors realized the moat around "nice UI plus integrations" is paper-thin when an agent can bypass the UI entirely.

But the data layer isn't safe either

Here's the nuance the "systems of record are safe" crowd misses.

The data in most enterprise systems is small. Your entire Salesforce instance probably fits in a few gigabytes. Every ticket ever created in Jira, maybe a gigabyte. Your HubSpot, Asana, Notion workspace — all of it together might be smaller than a single video file on your phone.

This didn't matter when the only way to interact with the data was through the application's interface. The data was small, but it was trapped. You had to log into Salesforce to see Salesforce data.

AI agents break this assumption.

To be useful, an agent needs API access to your systems. But once it has access, it can pull a complete copy of everything in seconds. The data is small. The API is permissive. And now the agent is sitting on a complete, synchronized copy of your records.

At that point, the original application becomes a write endpoint — a place where records get created before being consumed by the agent. The agent is where the work actually happens. Rate limits won't save you. A sufficiently sophisticated agent facing rate limits will simply build a caching layer, sync periodically, and make the original system irrelevant anyway.

Systems of record survive if they provide something beyond storage: multi-user collaboration, workflow enforcement, compliance governance, network effects. "We're where the data lives" is no longer sufficient when the data can leave in minutes.

Where the money actually goes

SaaS isn't dead. The easy SaaS moat is dead. Massive difference.

Companies will spend more on software this year than ever before. Enterprise AI capital expenditure alone will exceed $470 billion in 2026. The money isn't disappearing. It's repricing.

Usage-based AI platforms. Not per-seat. Consumption-based, like cloud compute. The per-seat model assumed humans were the atomic unit of work. When agents do the work, the pricing model has to change.

Outcome-based pricing. Instead of "$99 per seat per month" — "$5 per contract reviewed." "$2 per support ticket resolved." "$10 per qualified lead enriched." Software priced like labor, because it's replacing labor.

Security, governance, and compliance. When agents act at scale, mistakes happen at scale. Permissioning, audit trails, policy enforcement, monitoring, evaluation. Boring infrastructure that will print money for the next decade.

Vertical systems of record. Not generic CRMs, but authoritative data layers for under-digitized industries — construction, healthcare operations, legal ops. Domains where no good system of record exists yet, and every AI agent will need one to plug into.

Services. When it becomes cheap to build software, companies don't build less — they attempt more customization. Implementation, workflow design, migration, integration work. Vibe coding makes creation easy. Making it work inside a real business is a different story entirely.

What to build

If you're in the thin middle — a nice UI over someone else's data, charging per seat, with no proprietary advantage — the economics of your position are collapsing. The agent layer above is eating your interface. The data layer below is eating your lock-in. You're squeezed from both directions.

Instead:

Build at the agent layer. Tools that don't display information but execute workflows. Don't show the user a dashboard. Do the work for them. Price on outcomes.

Build at the data layer. Own proprietary data in a vertical that doesn't have a good system of record yet. Make yourself the authoritative backend that every agent needs to connect to.

Build the infrastructure. Security. Monitoring. Evaluation. Governance. The tooling that makes agents safe to deploy at scale.

The irony

Anthropic's Cowork — the product that supposedly killed SaaS — is itself a SaaS product. Sold via subscription. To organizations. On the internet.

SaaS as a delivery model is fine. It was always fine.

SaaS as a business strategy built on shallow moats and per-seat pricing for commodity workflows — that's what's over.

The adoption timeline looks like cloud computing: ten to fifteen years, not three to five. The restructuring is real and structural, but it's not instantaneous.

That $285 billion didn't vanish. It's transferring — from companies that captured value by sitting between humans and their tools, to companies that capture value through execution, data, and infrastructure.

The easy SaaS era is over. The hard SaaS era — own the data, execute the workflow, prove the ROI, price on value — is just starting.

For anyone actually building something real, that's the best news in a decade.