A capability that works in a demonstration has cleared a low bar. It ran once, on a good day, with the people who built it in the room. The bar that matters comes the morning after, when the same capability has to run without them, for an operator who was never consulted, against decisions taken a year earlier by people who have since moved on.
That is the reliance question. Can the organisation depend on this on an ordinary day, when it is tired and short-staffed and the person who understood it has left? Most digital capability is signed off the moment it is shown to work. The harder question is rarely asked until it has already become someone's problem.
By the time you reach go-live the answer is mostly fixed, and it was fixed a long way upstream. Whether an operator can rely on a platform depends on decisions taken before it exists. Someone has to be made accountable for running it once the builders leave. The data it needs has to be written into the contract as a deliverable rather than assumed into existence. Those are procurement and mobilisation questions, and they are decided before a line of the thing is built.
The pattern I see on live programmes is consistent. A team proves a capability, everyone agrees it works, and the first quarter after handover goes on rebuilding the parts nobody was asked to own. It shows up the same way on a metro programme as on a data-centre campus. The proof of concept lands, the room agrees it is impressive, and responsibility for keeping it running was never handed to anyone with the authority to make good on it. The technology was rarely the weak point. What failed was the handover of everything around it that no one had been made responsible for, and that failure stays invisible until an operator falls into it.
So the useful work happens early and it is unglamorous. Name the owner before you name the tool. Write the information you will need to run the asset into the contract as a requirement with an owner and a standard, not as a hope you will sort out later. Then ask, at every gate, one plain question. If the people who built this walked away tomorrow, could we still rely on it? If the answer is no, you have a procurement problem wearing the costume of a technical one.
Whether a capability can be relied on is answered by who owns it and what the contract asked for, and both are settled before anyone switches it on.
A digital twin is one place this plays out in detail, where reliance turns into a test of what the model shows and how current it needs to be for a given decision.