Cloud cost visibility was already broken BEFORE AI
Cloud cost visibility was already broken BEFORE AI. IE - shared infrastructure, tagging drift, cross-team ownership disputes, costs that don't connect to customers or features or outcomes. Most teams were already flying blind on a bill they couldn't explain.
AI doesn't JUST add to that problem. It MULTIPLIES it because the unit of cost is no longer something you can track at the infrastructure level.
Every other cost variable drifts in ways you can at least investigate. AI token consumption is different. It scales with what each user does, not how many users you have. A user who pastes a 50-page contract into your AI feature can cost 100 times more than a user who asks a simple question. Same subscription price. Same customer tier. Completely different cost to serve and almost nothing in the current landscape traces that back to the customer, the feature, or the workflow that caused it.
If you thought attribution was hard before AI, you haven't seen anything yet.
Curious how others are approaching this. Are you treating AI spend as its own attribution problem or folding it into general cloud cost management?