Every experiment deserves a budget and an exit.

Walk through any organization more than a few years old and you will find them: the pilots that never ended. A chatbot serving three users. A dashboard nobody has opened since its champion changed roles. An AI proof-of-concept, eighteen months old, still 'showing promise.' Nobody killed them. Nobody harvested them. They simply drift — consuming licenses, attention, and a little credibility every quarter.

This is the real failure mode of corporate innovation. Not a shortage of ideas — organizations overflow with ideas — but a shortage of endings. Experiments are started with enthusiasm and no destination, so they cannot succeed or fail. They can only continue.

The three-part contract

The fix is a contract, agreed before the experiment starts, with three clauses. A sponsor: one named person who wants the answer and will act on it — not a committee, not 'the business.' A budget: real money and real time, finite by design, because an experiment that cannot exhaust its funding cannot conclude. And an exit: the conditions, written down in advance, under which this experiment will either be harvested into the operating model or retired without drama.

The exit clause is the one that gets skipped, and it is the one that matters most. Deciding the kill criteria before you are emotionally invested is cheap. Deciding them eighteen months in — after the demos, the internal newsletter feature, the sponsor's reputation quietly attaching itself to the outcome — is nearly impossible. Sunk cost is not a character flaw; it is physics. The exit clause is how you build around it.

Endings are what make experiments safe

Here is the paradox leaders miss: strict endings make experimentation safer, and therefore make more of it possible. When every pilot has a budget and an exit, saying yes to a new idea is a bounded risk instead of an open-ended liability. Teams propose more, because proposing is no longer a life sentence. And retiring an experiment stops being an admission of failure — it becomes the system working exactly as designed.

Harvested or retired, never drifting. That is the whole discipline. An experiment that ends is knowledge. An experiment that drifts is just cost wearing a lab coat.