This article was adapted from its original version on TDWI.
Enterprise AI budgets are expanding at a pace that is making CFOs pay close attention. According to a 2026 survey of 2,360 senior executives, companies expect to spend approximately 1.7% of their revenue on AI this year, more than double the 0.8% average from 2025. Measuring ROI from AI is still a nascent practice. CFOs and IT leaders will need to manage AI related spend with a fine-toothed comb, especially since 60% of IT organizations are not increasing their budgets for AI.
The Meter Is Always Running
When deploying AI in production, the dominant cost is not model training but inferencing: the compute cost incurred every time a model generates a response. The FinOps community has developed frameworks for managing these inferencing costs: model routing, prompt compression, caching, batching, and tighter context window management. These are legitimate levers and the best engineering teams are pulling them hard.
But there is a dimension of inferencing cost that the FinOps conversation has largely skipped over: the quality of the unstructured data being fed into AI pipelines.
Unstructured Data: Unknown and Unexamined
Enterprises sit on enormous volumes of unstructured data: documents, images, medical files, contracts, research archives, email threads, and more, the raw material for many valuable AI use cases. Yet it remains largely unclassified, unsegmented, and unknown. Metadata enrichment, a crucial piece of unstructured data classification, must be automated and ongoing. A hospital, for example, might extract DICOM header data to flag body part and study type for AI-driven diagnostics.
Read more here on how Komprise helps with custom metadata extraction such as DICOM.
Why Metadata Management Matters to the CFO
Inferencing is billed by the token, and token count per request is the most controllable cost lever. Feeding an AI pipeline data that is poorly curated, redundant, or irrelevant is the fastest way to inflate it. IT teams that manage data quality before the inference request, not compute costs after, take the more efficient posture.
What Metadata Enrichment Does for the Balance Sheet
Metadata-enriched data can be searched and filtered by keyword, so only the right data goes into an AI pipeline, not everything available. The payoff is not marginal: enrichment can cut AI compute and storage costs 80% or more, a figure that matters when infrastructure runs $10 or $20 million annually. It is the same FinOps logic as matching workloads to the right compute tier, one layer down: matching AI workloads to the right data curated precisely for the task. Doing so avoids the waste of processing irrelevant, duplicate, or inappropriate content at model inference time.
A Healthcare Case That Quantifies the Gap
A leading healthcare system’s digital pathology AI workflow was delivering faster, more accurate diagnostics, but cloud storage costs were becoming a barrier. A curated, metadata-driven approach sent only the most relevant, recent files to cloud storage and automatically removed copies after 30 days.
The results:
• 96% reduction in cloud storage costs
• 10x faster AI data ingestion
• Storage down from 1 petabyte to a rolling 33 terabytes
Tagging data accurately and routing only relevant files through the pipeline produced the economics; deleting data after jobs complete was a major contributor.
The Governance Risk of Unstructured Data Quality Problems
Feeding a model outdated documents, duplicates, drafts, or regulated information carries its own cost, from degraded accuracy to libelous outputs, plus PII or IP exposure that brings fines, competitive loss, and customer defection. These costs are harder to measure than token spend but scale with data volume. A metadata-enriched approach mitigates the risk:
- Discover and exclude PII from AI pipelines by policy, before data reaches a model
- Scan file shares, directories, or sites for sensitive keywords unique to the organization
- Delete files by date or owner, such as ex-employees or the C-suite, to purge old or protected data
- Delete duplicate files based on metadata
The Reframe Finance Leaders Need
AI FinOps is an emerging practice that covers GPU usage, model routing, token budgeting, and inference unit economics. But sustainable AI economics need one more discipline: treating data preparation and metadata enrichment as a cost management function, since data determines the compute and storage a pipeline consumes. Organizations that apply the same rigor to unstructured data that they apply to cloud infrastructure gain a durable advantage: lower, predictable AI inferencing costs, better model performance, and a clean answer when the board asks what the AI spend bought.
The meter is running. What feeds it matters as much as how fast it runs.
