Main idea

Faced with a large account that wants to impose its email gateway, its SMS provider or its language-model contract, one question settles it: if I replace that provider with theirs, does the business value of my service change?

For email, no. The message still goes out at the right moment, because it is the vendor's engine that decided it should. For SMS, no either. For the model, no more so — if the value lies in the business workflow, the context, the rules, the data and the orchestration, the execution model can change without moving the heart of the product.

What survives the replacement can be handed to the customer. What doesn't survive it sits too close to the core to be freely substitutable.

Why it matters

The test replaces an architecture discussion with a decidable criterion, and it applies block by block with no prior theory.

It also guards against the opposite slope — handing everything to the customer — by giving an explicit reason to refuse: that particular replacement destroys what the product is bought for.

Nuances and limits

Providers are not perfectly interchangeable: one model can outperform another, one email gateway can have better deliverability. The test bears on business value, not on quality of execution.

And a block that is substitutable from the standpoint of value can remain very expensive to substitute in practice.

Open questions

• What do you do with a block whose replacement doesn't change the value but clearly degrades perceived quality?