My daughter is in college. She’s sharp, skeptical, and has developed a healthy distrust of AI—not because it’s always wrong, but because she can’t tell when it’s wrong.

I've been thinking about that framing a lot this week, and not just because its AI, but because of what I’m building with the FinOps Directory.

Mentioned in a LinkedIn post last week, a research team at Amplifying.ai recently published a systematic study of how AI assistants handle product recommendations. They asked Google AI Mode and ChatGPT 792 questions across 33 consumer categories — things like "best laptop under $1,000" — and tracked what each model recommended across multiple runs and multiple phrasings.

ChatGPT changed its top recommendation on 94.7% of repeated queries. The same question, asked hours apart, returned different answers. Google AI Mode was more consistent, but largely because it optimizes for available inventory, not product quality. Across platforms, the two disagreed more than half the time.My daughter would call this exactly what it is: total BS and why I don’t trust it.

This isn’t just a consumer problem—it’s the exact same failure mode showing up inside enterprise AI. The same characteristic that makes AI an unreliable product recommender — confident outputs that vary based on hidden inputs, training data, and timing — is operating inside enterprise AI deployments right now. Models are driving decisions and workflows that teams can price, but not reliably evaluate. The question my daughter asks about a chatbot recommendation — how do I know when it's wrong? — is the same question CFOs are now asking about millions of dollars in AI spend.

For the past 2–3 years, AI budgets have operated like a blank check. That era is ending.

Only 14% of CFOs surveyed by RGP say they've seen clear, measurable impact from their AI investments. 61% of CEOs report increasing pressure from boards to demonstrate returns. One CFO quoted in a recent Fortune piece described the directive from her largest investor in plain terms: "We need to see AI pay for itself or we need to see AI go away."

This is not a niche signal. It is the defining enterprise technology conversation of 2026.

And it lands directly in FinOps territory — because AI spend is cloud spend. Every inference call, every fine-tuning job, every embedding pipeline runs on a cloud invoice. Gartner put worldwide public cloud spending at $723 billion in 2025 — and that was before the current wave of enterprise AI deployment hit full stride. The AI portion is growing faster than any other category, and it is the portion that most FinOps teams are least equipped to understand, attribute, and help govern.

The C-Suite is looking for answers. The question is whether FinOps is positioned to provide them — or whether it will produce another impressive dashboard that explains the bill without changing it.

Last week I introduced the concept of Visibility Theater: the organizational condition where FinOps teams produce strong insights that nobody acts on, because ownership of execution lives somewhere else.

The AI ROI crisis is Visibility Theater at executive scale.

Organizations have spent billions instrumenting AI deployments. There are cost dashboards, token usage reports, model comparison analyses. What they frequently cannot do is connect those numbers to business outcomes — or find a clear owner responsible for improving the ratio.

The visibility is there. The accountability isn't.

This is not a technology problem. FinOps tools have gotten genuinely good at surfacing AI cost data. The problem is organizational: AI spending crosses teams, models, and use cases in ways that existing cost ownership structures weren't designed to handle. A single generative AI product feature might involve API calls from three teams, inference on two cloud providers, and fine-tuning jobs nobody remembered were still running.

When the CFO asks "what are we getting for our AI spend," the honest answer in most organizations is: we can show you what we spent. We cannot tell you why it was worth it or who is responsible for making sure the next dollar is better spent than the last.

That gap is now a board-level problem. But with the release of The FinOps Foundation Framework 2026 last week, it reads like a direct response to this problem. The headline addition — a new capability called Executive Strategy Alignment — is the discipline's formal attempt to close the gap between FinOps insights and the executives who need to help ensure action takes place.

The framing is explicit: FinOps needs to "shift up" — moving from operational reporting to strategic partnership with executive leadership.

The Framework describes FinOps earning influence through trusted data, connecting technology investment to business outcomes, and supporting multi-year planning at the C-suite level. For organizations still fighting for a seat in engineering reviews, this is meaningful progression.

But there is a gap between the Framework’s ambition and the execution problem most practitioners are actually living with.

Executive alignment without execution doesn’t solve the problem, it just moves Visibility Theater into the boardroom. It still doesn’t answer who owns the remediation: who turns off idle compute, who removes unused licenses, who is accountable for AI spend that isn’t delivering ROI.

The Framework maps where FinOps needs to go. The harder problem is building the organizational model that actually gets it there.

The Flexera 2026 State of the Cloud Report recently came out with a finding that deserves more attention than it will probably get.

63% of organizations now have a dedicated FinOps team — up from 51% just two years ago. 71% have a cloud center of excellence. Governance infrastructure is being built at real pace and scale.

And yet estimated wasted cloud spend increased to 29%, reversing a five-year downward trend. Managing costs ranked as the number one challenge for the fourth consecutive year according to 85% of the respondents.

Read that combination slowly: more FinOps teams than ever, and more waste than last year.

Flexera attributes the increase to AI and growing PaaS/SaaS complexity. That’s directionally correct. But it doesn’t explain why more governance is producing worse outcomes.

Only 12% of respondents cited lack of AI-specific FinOps skills or frameworks as a challenge. Organizations believe they are ready. The waste numbers say otherwise.

This is Visibility Theater at scale. The teams exist. The tools exist. The frameworks exist. What’s missing is the organizational wiring that turns insight into action.

More FinOps teams do not guarantee more execution. They often just produce more reporting.

FinOps is being asked to do something genuinely new in 2026: govern AI spend at a moment when CFOs are demanding ROI accountability, when the Framework is pushing toward executive alignment, and when the tools to do the job are still catching up to the scope of the problem.

The organizations that solve this won’t do it with better dashboards.

They’ll do it by building ownership models that force action—clear accountability and actionability, economic signals upstream, and feedback loops that actually close.

63% of organizations now have a FinOps team. 71% have a cloud center of excellence. Governance infrastructure is expanding year over year. Wasted cloud spend just went up anyway.

That’s the Flexera finding that should be on every FinOps leader’s radar this week. It also makes the next number more uncomfortable: only 12% of organizations say AI-specific FinOps skills and frameworks are a challenge.

The confidence is high. The outcomes are not following.

Organizations that have built FinOps practices for public cloud are applying that same muscle to AI workloads and assuming it transfers. In some ways it does. In the ways that matter most such as attribution across teams, models, and consumption patterns that don’t map cleanly to existing cost structures, it doesn’t.

Flexera describes AI cost management challenges as "reminiscent of the early days of cloud." The organizations that navigated that era best weren’t the most confident. They treated it as a new problem requiring new accountability structures, not an existing problem with a familiar solution.

Signal: The 12% who acknowledge the gap may be better positioned than the 88% who don’t.

Released Thursday, March 19, the FinOps Framework 2026 introduces a new capability called Executive Strategy Alignment — formalizing FinOps as a strategic partner to leadership rather than a reporting function. Six existing capabilities were renamed and broadened to cover all technology categories, not just public cloud. AI, SaaS, Data Center, and Data Cloud Platforms now each have dedicated framework guidance.

The "shift up" framing with the Executive Strategy Alignment is the Foundation's answer to the ownership gap this newsletter covered last week.

But read closely: Executive Strategy Alignment is still structured around FinOps earning influence through trusted data — escalating the advisory role to the C-suite through the CTO or CFO, rather than resolving who actually owns execution. The boardroom gets better visibility. The idle compute is still waiting for someone to turn it off. C-level sponsorship helps — but sponsorship and ownership aren't the same thing.

The Framework is expanding its scope faster than most FinOps tools are covering it. Worth watching what the vendor landscape does next.

Featured from the FinOps Directory, built and maintained by CloudXray AI.

Company: ProsperOps (acquired by Flexera)

Category: Automation & Optimization

What They Do: ProsperOps autonomously manages your commitment portfolio, whether convertible RIs, Savings Plans, and/or term optimization, continuously rebalancing coverage and lock-in risk faster than any internal team can operate manually

Why It Matters: Execution matters more than visibility theater — even when it comes to commitments

One of the clearest examples of the execution gap in FinOps is tagging. In most organizations, cost allocation still breaks down into “Unknown” and “Other.” Not because the tooling is missing, but because tagging is treated as optional. And optional work doesn’t get done.

When engineers choose between shipping a feature and adding metadata, the feature wins. Every time. This is how Visibility Theater starts. You cannot hold a team accountable for a bill you cannot trace back to them.

Most organizations approach tagging as a policy problem. Documentation, Slack reminders, periodic audits. That approach does not scale. If attribution is required for accountability, then it cannot be optional. It has to be enforced at the point of creation.

The most direct way to do this in AWS is through Service Control Policies (SCPs). (Note, for Microsoft its Azure Policy. For Google Cloud Platform, its a combination of Org Policy Service and VPC Service Controls. In any situation, it can be done.)

Define a deny rule: if a resource is created without required tags such as Cost Center or Project ID, the API call fails.

No tag, no resource.

You move from manual enforcement to system-level integrity.

Cost allocation is no longer a best effort process. It becomes a property of the system. Attribution approaches 100% because untagged resources never exist in the first place.

This is the difference between insight and action.

Tagging policies produce reports. Enforcement produces outcomes.

If FinOps is going to meet the moment with AI and increasing cost complexity, this is the model. Build systems where the correct behavior is the only possible behavior. Start taking action.