Application registers do not die of bad tooling. They die of ownerlessness, and they die in about six months. Anyone who has inherited one knows: last updated by someone who left, with three tabs contradicting each other.
The candidates are always the same, and each fails in its own way if left alone. IT operations already has a register, the CMDB, but that is operations' truth: instances and dependencies, not the business's use of them. Procurement has the contract list, but a contract says what we pay for, not what is used, and certainly not what is used for anything important. Finance has the cost allocation, but a cost centre is not an application. And the model where everyone owns the register together is the fastest route to nobody doing it.
The model that holds has two layers. One named owner of the register itself: the function closest to the architecture decisions, in the midmarket typically the head of IT or whoever actually carries the architect role. The register owner does not own all the data, but the structure, the cadence and the right to chase answers. And beneath that: distributed data ownership, where every application has an application owner who can answer three questions without looking anything up: do we still use it, who here is responsible, and when does the agreement expire.
Cadence matters more than tooling. A quarterly one-hour review that only handles changes beats the annual grand cleanup every time, because the annual one never happens the second year.
The test for whether ownership is in place is simple: ask in a meeting who owns the register. If more than one person points at themselves, or nobody does, you have your answer.
Read on: For the CIO · The application landscape · Agent governance
Atlas is free in Early Access, and you start with what you have. Spreadsheets are a fine beginning. Get early access.
Spekir builds the layer that connects strategy to the IT portfolio. See Atlas