Copilot or autopilot: where the money is and where the margin is

In March 2026 Julien Bek, a partner at Sequoia, published “Services: The New Software”. Here is his thesis, in my words.

Companies spend far more on services than on software. The next large AI companies will not sell tools to professionals. They will sell the work those professionals do. The best place to start is work that is already outsourced and follows rules. Each better AI model makes that service cheaper to deliver.

Foundation Capital had named the idea in 2024: Service-as-Software. Sequoia gave it a sharper line: “A copilot sells the tool. An autopilot sells the work.”

The buyer does not want a tool. A sales director does not want a lead scoring app. They want qualified leads on Monday morning.

A tool also has to be used. Someone learns it, someone feeds it, someone checks what it says. In most companies that someone does not exist, so the tool sits there. An autopilot removes that step.

And the budget for work is already there. A company that pays an agency or a consultancy for a task has decided the task is worth money. Nobody has to invent a new budget line.

The result has to be defined before you start. If you charge per qualified lead, both sides must agree on what a qualified lead is before the first one. Otherwise every invoice turns into an argument.

Someone answers for the result. A tool that fails is the user's problem. An autopilot that fails is the seller's. That means a named person accountable for it, insurance where it makes sense, and a signature where the law asks for one.

Running it costs money. Each task pays for the AI models it uses and the tools it relies on. Those costs grow with every task, unlike most of the cost of software.

So the margin sits in between: an autopilot earns more than an agency and less than a software company. Better models lower that cost, but they do not remove it. I would not sell anyone software margins on day one.

Five questions. If a task gets five yeses, it is a good first autopilot.

Do you already pay someone else to do it? Then the budget and the idea of “done” exist.

Does it follow rules? Sequoia calls this intelligence, as opposed to judgement. Filling a grant form is intelligence. Deciding which grant to go for is judgement.

Can anyone check the result? A lead, a report or a checked case file can be checked. “Better positioning” cannot.

Is there a clear unit to charge for? Per task, per case, per qualified lead.

Is the liability low? If a mistake can end in court, start with another task.

Look at what you already outsource before you buy another AI tool. The first autopilot is usually hiding in an invoice you pay every month.

Ask any provider who answers when it fails and what each task costs to run. If they cannot answer, they are selling you a copilot with a new name.

That is how I work: a diagnosis to pick the task, then the autopilot and, every month, a report with the cost of each task.