March 16, 2026
AI-powered, fully automated FinOps that optimizes IT spending sounds like a great idea. Set it and forget it — and say good-bye to your worries about talent shortages in cybersecurity or cloud management while you watch the savings roll in!
But the problem is that the algorithms would be optimizing without context, and that means they wouldn’t be optimizing at all. They are simply executing rules at scale, and sometimes those rules are spectacularly wrong for your business.
According to the State of FinOps survey, automation has risen in importance, particularly for organizations with small to medium cloud spend. The efficiency gains are real. Machine learning can analyze usage patterns across vast infrastructures at speeds and scales impossible for human analysts.
But most organizations use automation primarily to gather data and detect the need for action, with humans still taking actions manually. Full automation, where actions occur without human approval, remains rare, especially among large organizations in regulated industries.
This isn't technological conservatism or fear of change. It's hard-earned wisdom about what algorithms can and cannot do. For all their performance value, machine learning models lack something critical: business context.
For instance, your AI system identifies an "underutilized" database instance running at 20% capacity. Optimization logic says downsize it. But that database supports your compliance reporting system that spikes to 100% capacity for three days each quarter. Downsizing it saves money fifty weeks a year and creates a catastrophic failure during audit season. Or your development team spins up expensive GPU instances that the algorithm flags as anomalous spending. Should they be stopped? Not if that team is experimenting with AI models that could transform your customer experience, and the cost of those instances is trivial compared to the competitive advantage they might unlock.
The fact is that AI will have a significant impact on operations by helping business better understand trends and optimize estates, but it won’t replace all of the human elements. Those elements are context, judgment and strategic thinking, and they aren't optional add-ons. They're essential to optimization that genuinely serves business objectives.
For organizations in regulated industries, the stakes of blind automation are even higher. Financial services, healthcare and government entities must adhere to various data sovereignty, audit trails and change management requirements.
Automated systems might optimize costs by migrating workloads to different regions where compute is cheaper, inadvertently violating data residency requirements. They might consolidate databases for efficiency, breaking segregation of duties controls that auditors require. They might scale down resources during perceived low-usage periods that actually represent critical backup windows or compliance processing.
These aren't hypothetical risks. They're real failures that occur when optimization algorithms operate without understanding regulatory constraints.
Lack of trust in fully automated systems persists for good reason. The fundamental limitation isn't the sophistication of the AI, it's that algorithms apply one-size-fits-all logic that can collapse in the face of enterprise complexity.
DXC Total FinOps embraces a more sophisticated approach: true human-AI collaboration where each amplifies the strengths of the other.
This collaborative model delivers benefits that neither approach achieves alone:
DXC's Total FinOps solution, which can draw on our strategic alliances with partners IBM Apptio and Flexera to maximize IT investment, operationalizes the human-AI partnership through an integrated platform and expert team:
The most sophisticated organizations recognize that the future isn't about choosing between human expertise and artificial intelligence. It's about orchestrating them together in ways that amplify the strengths of both while mitigating their respective limitations.
The best FinOps teams aren't fully automated because the best optimization isn't purely algorithmic. It's contextual, strategic, and nuanced. It requires both the scale of machine intelligence and the wisdom of human expertise.
