Software, the New Services

Sequoia just published Services: The New Software. The thesis: AI makes it possible to sell outcomes instead of tools. The next trillion-dollar company won’t sell a copilot — it will sell an autopilot.

They’re right about the direction. But we think the arrow points the other way. Services aren’t eating software. Software is eating services.

Sequoia cites a striking ratio: for every dollar spent on software, six are spent on services. That’s the real story. Not whether we need more services or more software — but what happens when the cost of both collapses and the spend redistributes.

The SaaS market contraction everyone’s panicking about? It’s not a collapse. It’s a redistribution. Money is moving between three layers — from legacy platform vendors to frontier model providers, from large-scale services firms to smaller domain-specific builders, and from the services line item into software that eliminates the need for services entirely. The huge scale and platform capture that defined the last era of enterprise tech is losing its grip. Downmarket companies that couldn’t compete before suddenly can.

The Math That Changes Everything

I ran 7,000 automated tests before a production deployment this week. Think about what that represents in the old world. You’d need business analysts writing requirements and acceptance criteria. Engineers translating those into test code. A QA team maintaining and running the suite. That’s five to ten people full-time — easily $1 to $2 million a year in salary alone just to build and maintain the test infrastructure for one application.

With AI-augmented development, one person does it in an afternoon. And every test compounds — it gives information to the coding agent writing the next feature, prevents regression, and makes the next project go faster.

Now scale that across an enterprise. The systems integrators bill $200-400 an hour for engineering talent. When that same work can be done for $10-20 an hour equivalent — one person orchestrating AI agents — the entire services model has to restructure. Not disappear. Restructure.

What Changes for the Big Players

The consultancies aren’t going anywhere. They have relationships with every large enterprise on earth, and those relationships are durable. What changes is where their value lives.

Today, an enterprise implementation means buying a platform and paying a consultancy to configure it. Projects stretch. Customizations multiply. Integrations grow more complex than the original system. Not because the people are bad, but because the economics of building custom were so prohibitive that configuring a platform was the only rational choice.

That constraint is evaporating. The 6:1 services-to-software ratio exists because implementation was hard. When building from scratch costs less than configuring a platform, that ratio doesn’t just shrink — it redistributes. Less money on labor, more on architecture and domain expertise. Less on configuring someone else’s opinions, more on encoding your own. The firms that adapt will thrive — delivering better outcomes, faster, at higher margins. Their value shifts from labor to judgment: knowing what to build, how to architect it, what the business actually needs.

The Real Opportunity

But here’s what excites us: this isn’t just a story about Accenture and Deloitte restructuring their practices. Custom software that only huge companies could afford to build and maintain is now accessible to a ten-person shop. The smaller companies that realize this first will grow faster.

As we wrote in Bespoke Software — mass-market tools are off-the-rack suits. When the tailor charges the same as the department store, why compromise?

This is what we’re building with FabWise. A fab shop owner can’t hand someone a spreadsheet and expect generic software to understand manufacturing. You need a foundation that already speaks the language. So we’re building an onboarding agent that conducts an interview with the customer — in their language, about their business — and configures the entire account for them. No setup wizard. No implementation consultant. Buy it, start using it the same day.

The domain logic lives in a deterministic backend. The front end becomes negotiable — chat interface, mobile app, whatever fits how you work. The opinions are in the architecture, not the UI.

Where This Goes

Software that’s domain-specific enough to work out of the box, smart enough to configure itself, and simple enough that the customer never calls anyone. Not custom-built by a consultancy. Not a generic platform that needs an army to implement. Something that arrives already understanding your world.

The tools got cheaper. The taste didn’t.