Michael Brennan, CTO of MagicOrange, on why the architecture decisions enterprises make today about technology cost management will determine whether their AI programmes are actually governable — and what it takes to build a platform that works with real-world data, not idealised assumptions.
AI governance starts with how you build your cost model today
I spend most of my working life thinking about platform architecture. How systems are built, how they scale, how data flows across them, and critically, how the financial model that sits underneath all of that is designed to reflect the way technology is actually consumed rather than the way it was convenient to record it.
That last point — the gap between how technology consumption happens and how it gets recorded — is where most ITFM and FinOps implementations fall over. Not because the technical integration is wrong. But because the cost model is built on assumptions that do not hold under real-world conditions: incomplete tagging, inconsistent source data, organisational structures that do not map cleanly onto the allocation framework. When those assumptions break, the outputs lose credibility with the people who need to act on them. And once credibility is lost, the programme stalls.
This is the architectural challenge that MagicOrange was designed to solve. Not by forcing the data to conform to a rigid framework, but by building a platform flexible enough to work with the data as it actually exists — messy, incomplete, inconsistently tagged — and still produce outputs that are auditable, consistent, and trusted by both IT and finance.
I am proud to share that MagicOrange has joined the FinOps Foundation Governing Board.
From a CTO perspective, this matters for a specific reason that does not always get enough attention: the Foundation is where the technical standards underpinning FinOps practice are developed. Tagging frameworks, cost allocation specifications, data models that define how cloud and AI consumption is attributed — these are technical artefacts, not just financial ones. Having MagicOrange in the room means the platform engineering perspective shapes those standards alongside the practitioner and finance voices.
Our platform is built on Microsoft Fabric and Databricks — modern data architecture that processes billions of rows in minutes rather than hours. That is not a marketing claim. It is the architectural prerequisite for making Technology Economics operationally useful. A cost model that takes twelve hours to run is a model you look at once a month. A model that runs in minutes is one you can iterate on, scenario plan with, and use to make decisions in real time.
AI makes the architecture challenge significantly harder, and I want to be specific about why.
Traditional cloud cost management involves relatively predictable resource consumption that maps reasonably well onto standard allocation frameworks. AI workloads are fundamentally different. Token-based pricing for inference APIs varies with demand in ways that traditional reserved capacity models do not anticipate. Model training runs can consume compute at a scale and speed that overwhelms budget guardrails designed for steady-state infrastructure. And AI is being consumed by business users — not just engineers — which means the organisational scope of the attribution problem extends far beyond the platform team.
None of this fits neatly into cost models built for the previous generation of cloud infrastructure. The platform needs to handle this complexity natively — through flexible data ingestion, configurable allocation logic, and real-time cost visibility — not through post-hoc workarounds that add manual effort and introduce reconciliation errors.
Building that platform is what my team works on every day. The FinOps Foundation Governing Board is where we contribute that engineering perspective to the industry frameworks that will define how the next generation of Technology Economics is implemented across the enterprise.
If you are an architect, a CTO, or a technology leader evaluating platforms for Technology Economics, I would welcome the conversation.
Michael Brennan | CTO, MagicOrange
MagicOrange has joined the FinOps Foundation Governing Board, helping shape the future of AI cost management and Technology Economics.
Read the full announcement: MagicOrange Joins the FinOps Foundation Governing Board
#FinOps #ITFM #TechnologyEconomics #AIcosts #FinOpsFoundation #MagicOrange #CloudFinancialManagement #TBM #EnterpriseIT
