Y Combinator changes the rules: agencies are back

For years, the playbook was clear. If you wanted to build something scalable, you avoided agencies. Low margins, growth tied to hiring, heavy dependence on people. It didn’t fit the venture model.

In the Spring 2026 Request for Startups, Y Combinator introduces a shift worth paying attention to. Among AI, fintech, and industrial ideas, one category stands out: agencies.

Not in their traditional form.

What YC is actually asking for

YC is pointing to AI-native agencies. Teams that don’t sell hours or manual execution. They build systems that do the work and sell the outcome.

Their examples are very concrete:

• A design agency generating full creative before the contract is signed

• An ad agency producing high-quality video without physical production

• A legal firm drafting documents in minutes

This is a shift in the unit of value.

The structural change

Until now, the split was simple:

• SaaS sells tools

• Agencies sell services

That model is breaking.

Software is no longer supporting execution. It is execution. Teams design and refine the system, while the customer buys the result directly.

The logic flips:

• You don’t help the customer do the work

• You do the work for them

That changes how value is captured.

Why this works now

The traditional constraints of agencies were well known:

• Work scaled with people

• Hiring was the growth lever

• Margins were under constant pressure

AI changes those constraints:

• Execution becomes automated

• Growth depends on systems

• Margins improve as the system scales

What emerges is a hybrid model:

• Outcome-driven pricing from services

• Scalability and efficiency from software

This is bigger than agencies

Looking across YC’s other startup ideas, a broader pattern appears.

• Systems that decide what to build

• Agents executing complex workflows

• Software replacing manual processes in traditional industries

• AI guiding physical work in real time

The direction is consistent. Software is moving from tool to operator.

Why agencies are the entry point

This is where YC’s thinking becomes practical.

Agencies already have:

• Outcome-based value

• High ticket sizes

• Repeatable workflows

That combination allows something SaaS struggles with early on:

• Deliver value immediately

• Capture revenue upfront

• Iterate while operating

You don’t need adoption at scale to start. You start by solving a real problem and progressively automate it.

What this changes for product teams

The question is no longer what software to build.

The question becomes:

What work can you take completely off the customer’s plate?

From there:

• Design the system that executes it

• Automate as much as possible

• Package the outcome

• Price it as a service

Teams are no longer competing on features. They compete on owning the problem end to end.

The product perspective

This is not about building an agency with AI.

It’s about understanding:

• Which parts of the work should disappear

• Which decisions actually matter

• Where the real value sits for the customer

That’s the product.

The code comes after.