Costs You Can't Attribute, You Can't Govern
The Cost-Center Mismatch™ framework applies a long-standing finance concept — cost-center mismatch — to AI consumption and adds a four-question attribution test. The core insight: the party creating the cost isn't the party paying, so nobody has a reason to ask for a limit. Question 1: Can you attribute AI consumption by team and use case, down to the application ID? Question 2: Who pays when the cost doesn't land where it was created? Question 3: How long is the gap between someone consuming and someone seeing the cost? Question 4: What would your business units learn from even a rough chargeback? Includes an illustrative Sev-1 incident cost model (20 engineers, 4 hours, ~$47 on Sonnet 5.5 at current published pricing), the model choice multiplier (~20x per token between Haiku 5.5 and Sonnet 5.5), and the practical stance: attribute first, cap second. Approximate allocation is cheap. An absent one is expensive. Developed from 13 years of enterprise API gateway operations and capacity planning. Keywords: AI cost attribution, cost-center mismatch, AI FinOps, AI spend, enterprise AI economics, AI governance, margin leakage, PE due diligence, AI budget, AI TCO, AI incident cost model, AI model choice multiplier, unowned cost category, CFO AI budget, AI OpEx.
Authors
- DHAVALA DRUHIN (ORCID: https://orcid.org/0009-0009-1335-0353)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-14
- DOI
- https://doi.org/10.5281/zenodo.24228679
- Primary Topic
- Information Technology Governance and Strategy
- Type
- article
- Field-Weighted Citation Impact
- 0.00