Reserved instances will not save an organisation that cannot say who owns a resource. Start with ownership, then negotiate.
Every cloud cost review we run finds the same three things: environments nobody has logged into for months, storage tiers that were never revisited after migration, and a long tail of resources with no owner recorded anywhere.
Commitment discounts are applied on top of this, which locks in the waste for a year. The order of operations matters. Tag and attribute first, decommission second, right-size third, and only then commit.
The durable fix is procedural. Every resource gets an owner and a review date at creation. Every environment above development has a documented decommissioning trigger. Finance receives a monthly attribution by team, not a single line labelled infrastructure.
Organisations that do this typically find twenty to thirty-five per cent of spend is recoverable in the first pass, and, more importantly, that the saving does not creep back.