From people-led to service-led

An earlier version of this appeared on rework.com.

What if every piece of work in a company had clear inputs, outputs, and standards? What if people and AI could hand off tasks as smoothly as relay runners? Building Rework, I’ve seen this work.

In the beginning, a startup runs on a handful of key people. They drag the company from 0 to 1, sometimes from −1. It works, but it’s risky. Not every company has enough of those people. When everything depends on a few, growth feels like gambling.

As you scale, the natural move is to split into departments: sales, marketing, HR, operations. Each has its own role. This works for a while. But the bigger you get, the more layers you add. Layers become silos. Each department keeps its own data, its own software, its own language. Work that crosses departments breaks, and everything slows down.

That’s when I think a company needs to become service-led.

Any piece of work, whatever its size, wherever it comes from, can be described as a service. It has inputs, outputs, and rules. Services make work easy to combine, track, and run. They give both people and AI the context to work accurately.

It’s still the same business. It’s just organized into units that are easier to understand, measure, and improve. Not replacement. Evolution.

In practice, three things change.

Work has inputs and outputs. Outreach becomes a service: contact details go in, a qualified lead comes out. A support request becomes a service: a customer’s problem goes in, a solved problem and a satisfied customer come out. Once work looks like this, tracking it gets much simpler.

The company speaks one language. Department borders start to blur. Instead of a set of departments, the company is a set of services. You can look at anyone and see which services they provide. The company’s performance becomes the sum of every person’s, and every AI agent’s.

AI gets the context it needs. The hard part of using AI isn’t the model. It’s the business context. In most companies, data is trapped in separate tools. When context lives inside each service, an agent gets exactly what it needs, and it follows the same rules people do.

That’s what I mean when I say we should service-ize our work.