← All field notes

What a 1700s Steam Engine Salesman Knew About Pricing Automation

By Caleb Crane

Before he was a unit of measurement, James Watt had a problem. He'd built a steam engine that could replace the horses mine owners used to pump water and haul loads, and he had to figure out what to charge for it.

He didn't sell the engine outright. He rented it, and priced the rental against what the mine owner was already paying to keep horses: feed, housing, and the people needed to manage them. His pitch was roughly this: replace this many horses with one engine this size, and pay me a share of what you would have spent keeping them. Ongoing, not a one-time fee.

It's the same shape of deal I'd make automating a task today, three hundred years later, and I don't think that's a coincidence.

Why the horsepower pricing model still works

The mine owner didn't have to trust a number Watt made up. He already knew what horses cost him: feed, stable space, a person to manage them. Watt's engine only had to beat a cost the owner could already see and already believed.

That's the part worth copying. When you price automation as a flat number picked in advance, the prospect has no way to check it against anything. When you price it as a share of what the task already costs them, in their own hours and their own rate, the number is self-evidently fair, because they already knew the input.

Watt actually undersold himself slightly, since the engine could run continuously and horses need rest and rotation. But the pricing structure, not the exact math, is what made the deal easy to say yes to.

The other half of the lesson: find the actual cost first

There's a second story, from a completely different conversation, about a business that was still doing something the hard way for no real reason.

An organization was spending two full business days every month manually typing every credit card transaction into QuickBooks from paper statements. Two entire days, every month, for years. When someone finally sat with them and asked what their process actually was, it turned out QuickBooks had a bank feed feature that would have done it automatically. The bookkeeper knew the feature existed. She had just never gotten around to looking into it, because she was too busy doing it the slow way to stop and check.

That's the pattern behind almost every automation win worth having: not a clever new capability, but someone finally asking what a task costs today, in plain hours, and discovering the gap was bigger than anyone had bothered to measure.

Putting the two together

Before you can price automation the way Watt priced his engine, you need the number Watt had: what does the current way actually cost.

That's arithmetic, not magic. How often does the task happen. How long does it take. What is the person doing it worth per hour. Multiply those out and you have the real cost of the horses, in today's terms, and the fair price is a share of what replacing them saves.

The part that's easy to skip is asking at all. The bookkeeper doing two days of manual entry a month knew a better way existed and never checked. Most automation opportunities are hiding in exactly that gap: not a task nobody knew was possible, but a task nobody stopped long enough to actually cost out.

Why this beats a flat rate

A flat number invites the question "why does this cost that much," and there's no good answer because the number wasn't built from anything the prospect can verify.

A share of the actual, measured savings answers the question before it's asked. It's the reason the 9-minute UseCase Assessment works the way it does: it starts by finding your horses, in hours and dollars, before it ever gets to a price. Three hundred years later, that's still the only pricing model that doesn't require anyone to just trust you.