A platform is a foundation, not a project

Teams keep funding “the platform” the way they fund a marketing site: a start, a finish, a demo, a handoff. Then the first product team that has to live on it files a ticket into a backlog nobody owns. Six months later the paved path is a wiki, and every squad has a special-case pipeline they are slightly ashamed of.

A platform is an operating stance. It has an owner, a changelog, and a person who can say no. The work is deciding which paved paths exist, which exceptions you will carry, and what you will not automate because the judgment does not belong in YAML. Those decisions decay. If nobody is paid to renew them, they become folklore.

Treat that as a project and you will get a pile of pipelines and a retirement party for the contractor who remembered why the staging account looks like that. Treat it as a line of business — smaller than the products it serves, staffed like something that can fail on a Tuesday — and you might get a platform.

The test is simple. When a product team asks for a new path, is there a human who can answer this week, with authority, and live with the consequence? If the answer is a steering committee or a backlog named “platform phase two,” you have a project with residual risk, not a foundation.

We take this work when a company already feels the cost of the alternative: three ways to ship, two ways to observe, and an on-call rotation that pages the same three people. We do not take it as a six-week transformation. Transformations end. Foundations have to be there on the Monday after.