The price is the only proof

Every software provider says artificial intelligence has changed the way they work. There is a very simple way to check who is telling the truth.

You ask for a change in your ERP. Something modest: an automatic alert when an order has been stuck for five days.

A quote comes back. Eighteen hours. And underneath, the rate.

There is nothing odd about that document. It is exactly the one you would have been sent in 2015, in 2010 and in 2004. The amount changes, the structure does not. And that continuity, which looks like an administrative detail, is the actual news: the world has moved and the price has not noticed.

Nobody ever wanted hours

It is worth starting by dropping the accusation, because the hourly model was not a scam. It was an accurate description of the world.

For forty years, software was paid for by what it cost to make. Not because the hour was the valuable part, nobody has ever wanted to buy hours, but because the hour was the scarce part. For an improvement to exist, somebody had to sit down and write it, and that person was the bottleneck. Pricing hours was pricing what actually constrained the work.

That was the deal, and it was an honest one. What has happened is that one piece of the deal has fallen away.

One of three things collapsed, not all three

This is where most of the industry overshoots, and where saying less gets you closer to right.

For an improvement to exist in an ERP, three different things are needed: writing it, knowing which one to write and answering for it when it touches invoices, stock or payroll.

Of the three, only one has collapsed. Writing code tends to zero, which is true and enormous. But knowing which change is worth making costs exactly what it cost ten years ago. And answering for a mistake that throws your accounts out costs more than ever, because things can now be built much faster and therefore broken much faster.

The cost of writing software has collapsed. The cost of knowing which software to write has not. What is scarce is no longer hands: it is judgement.

That sentence is what separates a change of era from a slogan. Anyone can say they are faster now with AI. Being faster means nothing if what got faster is the part that was already cheap.

The question that settles it in ten seconds

And now the uncomfortable consequence, which is the only part of this article you need to remember.

If what is scarce changes, the price has to change. A price is a claim about what is expensive. If a provider tells you that artificial intelligence has made them three times faster and still bills you the same eighteen hours, then the gain exists but they have kept it.

No bad faith needs to be assumed. Price is the last thing a company touches, because touching it is frightening and it works untouched. But that is exactly why it is the proof: the price is the one thing that cannot be claimed for free.

Hence the question you can ask tomorrow, in any meeting, with no way around it:

Ask your provider whether artificial intelligence has changed the way they work. They will say yes. Then ask whether it has changed the invoice.

Their face during the second silence tells you everything you need to know.

The damage is not on the invoice

But the amount is not the worst part of the hourly model. Not by a distance.

The worst part is what it teaches you to do. When every change has to be requested, scoped, quoted and approved, small improvements stop being worth the paperwork. So you stop asking for them. And you write them down nowhere, because they never become a request: they stay as a "this ought to be automated" said out loud on a Tuesday that nobody picks up.

We call that the invisible list, and it is the real cost of the model. It is on no invoice. It is in everything you stopped asking for.

"If code is nearly free, what do I need you for?"

It is the right objection, and it deserves an answer before you get there yourself.

Because the bottleneck was never the code. An ERP is the one system in your company where a mistake does not cost an apology: it costs a wrongly issued invoice, stock that does not match or a payslip with an error in it. On that ground you do not want cheap code. You want somebody who decides what is worth doing, reviews it before it touches anything real, and answers for it when it goes wrong.

Writing got cheaper. Deciding and answering did not. And those are exactly the two things worth paying for.

What can be promised now, and could not before

Here is the turn that makes everything above matter.

Under the hourly model, proposing work a customer has not asked for means giving hours away. That is why almost nobody does it: it is not a lack of generosity, it is arithmetic. The incentive pushes the other way, because the longer each thing takes, the better the month looks.

With a flat fee, that same gesture stops being generosity and becomes the business model. We only get paid next month if you stay. And you only stay if something you did not know you needed shows up each month. Our only way of making money is to look at your business and tell you what it should be doing and is not. There is more detail in consulting hours and flat fees.

It is not that we are better people now. It is that the incentives finally point the same way you do.

And yes, the obvious objection: won't you raise the price once I depend on you? The answer cannot be a promise, because promises are worth nothing. It has to be a structure: the price published on the site, no minimum term, and your data and your code are yours. Leaving has to be cheap for staying to mean anything.

---

For forty years software was paid for by what it cost to make, because hands were the scarce part. That is over. What has not got cheaper is knowing what to do, or answering for it when it touches your invoices.

What is scarce has moved. Almost no price has noticed yet.

When you find one that has, you have found somebody who believes their own story.