# The SaaS Endgame

Every technology wave follows the same pattern: what used to require a team, a building, and a budget gets compressed into a function call. Then that function call gets compressed into a sentence. Then the sentence becomes implicit.

We're watching this happen to entire businesses in real time.

This is not speculation. The pattern has repeated with every major technology shift. When websites became trivial to build, value moved to SEO. When cloud made infrastructure free, value moved to data. When LLMs make code trivial, value moves to distribution.

The question is never "will this happen?" The question is "what becomes scarce next?"

Historically, holding companies existed because operating businesses required teams of humans. Berkshire Hathaway, Constellation Software, LVMH—they all succeed because they buy or launch many companies and compound them. But the constraint was always: each company needed a management team.

Agents remove that constraint.

Instead of betting on one startup, you run hundreds of experiments. Launch 100 businesses. 60 fail. 30 break even. 9 make good money. 1 becomes massive. That one pays for everything. This is venture capital inside a single company.

And the portfolio doesn't have to be software. Everything above assumes digital businesses—SEO tools, APIs, newsletters. But the same arbitrage applies to physical, local businesses, and that market is far larger and far less contested. A flower shop. An electrician. A plumber. The constraint was never the trade—it was the operating overhead around the trade.

The play: an agent scans tens of thousands of local businesses, scores each one on SEO difficulty, search-volume-to-ad-cost ratio, and wholesale input margin, and spins up a fully commercial operation—site, CRM, Google Ads, Meta Ads, booking, billing—for a few hundred dollars in under a day. Seasonal demand spikes like Mother's Day or Valentine's Day flowers become timed arbitrage windows you can enter and exit deliberately.

Labour works the same way—you don't replace the tradesperson, you wrap them. Hire one electrician fresh out of an apprenticeship; the agent layer handles dispatch, scheduling, quoting, marketing, and invoicing around them—enough of an edge to undercut every other operator in the suburb. When humanoid robots cross the line, that same person stops pulling wire and starts supervising the robots. The headcount doesn't get cut; its job changes.

The whole physical transition is gated on a single number: the return on investment of one humanoid robot. The day that number clears a dollar, physical labour becomes capital-allocatable—put a million in, get more than a million out—because growth, operations, and the company OS are already solved problems. And it won't take a decade. Markets move at the speed of the fastest mover, not the slowest. Three to five years.

Ask yourself: why do we need 50 different CRMs? 30 outreach tools? 20 email marketing platforms? The answer is not that each one is uniquely valuable. It's that the supply chain for matching buyers and sellers is inefficient, and each tool exists to help a different vendor compete.

Agents collapse this supply chain.

When the buyer has an agent actively seeking the best option—with enough context to know what to pay, what quality to expect, and what alternatives exist—and the seller has an agent that can negotiate, customize, and deliver, the entire middle layer evaporates.

Every SaaS tool that exists to facilitate the competition between vendors—CRMs, outreach platforms, proposal tools, analytics dashboards—becomes unnecessary. Agents find the most capital-efficient and time-efficient routes to match buyer and seller directly.

The biggest mistake people make is thinking "everything disappears." What actually happens is: abstraction collapses create new bottlenecks. When something becomes abundant, value shifts to the next constraint.

If millions of agents can build products, the bottleneck is no longer "can you build it?" It's "can you get it in front of the right people?" The scarce resources in the agent economy are not code. They are:

Imagine a world where millions of agents run businesses. They all need the same things: customers, reputation, suppliers, data, payment rails, legal compliance. Agents will compete for these scarce resources.

The most powerful position is owning the infrastructure agents must use. Thousands of trains will run. The rails collect tolls from all of them.

This is why the biggest winners of the agent economy might not be AI companies. They might be owners of scarce real-world things that agents compete over. The ad market alone could explode in scale when millions of agents bid for attention in real-time, fully automated.

There's a reason this paradigm shift keeps stacking instead of plateauing. The people who figure it out first don't just have a slight edge. They have a 100x head start—and that head start creates a flywheel that accelerates the entire shift.