Not long ago, cloud cost was an engineering problem. FinOps teams owned it, finance leaned in occasionally, and everyone else stayed out of it. Now, that’s changed.
AI changed who has skin in the game. CFOs get asked about it in board meetings. CEOs field questions on earnings calls. The audience for cloud cost management has exploded — and that means the conversation CloudZero is built to enable isn’t only a technical one, it’s a business one.
Based on recent conversations with customers, here are five problems driving that conversation right now.
1. ‘We’re spending a fortune on AI and nobody can tell me if it’s working.’
When only 14% of CFOs report seeing clear, measurable ROI from AI investments (and research shows 95% of AI pilots generate zero return) executives start asking harder questions.
The problem usually isn’t that AI isn’t delivering value. It’s that the value can’t be proven because the cost can’t be attributed. AI spend is fragmented — it’s buried in shared accounts, invisible to the people being asked to justify it. You can see the total bill but can’t connect it to anything that matters.
One customer, a global SaaS platform running more than 50 LLMs, had exactly this problem. By allocating costs per model, per region, and per customer segment in CloudZero, they unlocked cost per token and cost per user and drove over $1M in savings. They cut compute spend by more than half. They didn’t spend less on AI, they just spent it smarter, and they could prove it.
That’s the answer boards are asking for. It’s the answer only unit economics can give.
2. ‘Engineering keeps getting pulled into financial conversations they shouldn’t own.’
Every month, finance needs to understand what was spent and why. So they ask engineering. Engineering stops building to pull data, construct reports, translate infrastructure decisions into financial language and defend choices they made six months ago.
Half (52%) of engineering leaders say the disconnect betweenFinOps and developers is driving wasted spend on cloud infrastructure costs, according to Harness’s FinOps in Focus 2025 report. But the real cost isn’t invoice-related, it’s the interrupted focus and context-switching that has nothing to do with why you hired them.
Drift’s engineering team knew this feeling well. Once they had cost per customer visibility through CloudZero, their framing changed:
“Our job is to make our customers more successful. What CloudZero helps us do is figure out how we can encourage customers to use more of our expensive features – while we make those features more affordable to operate.”
That’s engineering thinking about business outcomes instead of being up to their necks in cost reports. It helped them reduce annual cloud costs by $2.4M, while making better product decisions along the way.
When finance gets self-service answers and engineering gets back to building, the dynamic changes entirely.
3. ‘Our AI and cloud costs are unpredictable, and we keep getting surprised by the bill.’
Traditional infrastructure scales in ways companies can model, but not AI. Token-based pricing shifts fast, based on prompt complexity, context length, usage patterns nobody anticipated when the feature shipped. A minor change in how an AI product works can double inference costs — companies that launched AI features in 2024 are seeing bills five times their original estimates. Month-end arrives and then what?
NinjaCat lived a version of this before AI even entered the picture. Costs kept spiking and they had no idea why – was it a power user? A pricing tier issue? Natural fluctuation? CloudZero gave them visibility at the unit level and they could identify exactly which customers were driving increases and why. They restructured their pricing tiers, stopped the bleed, and had honest conversations with customers about actual consumption. The bill stopped being a surprise and became a management tool.
That’s a significant shift from reactive to continuously informed.
4. ‘We’re running AI agents at scale, and we have no idea what they actually cost.’
This is the newest pain and it’s moving faster than any other. AI cost management has gone from a concern held by 31% of FinOps teams in 2024 to 98% today. Agents are running in production and the economics are almost entirely hidden.
There are easy explanations but they’re not easy to solve. A single agent interaction can trigger dozens of LLM calls, each one adding cost that compounds invisibly while the demo looks great. CTOs and engineering leaders get asked whether agent investments are viable at scale. Most can’t answer yet because the unit economics don’t exist in their current tooling.
One neobank valued in the tens of billions tracks more than 40 unit cost metrics and 45+ cost allocations inside CloudZero to manage this complexity. They don’t wait for the bill to find out what happened. They know cost per workflow, cost per interaction, cost per outcome on an ongoing basis.
Want to run agents responsibly, without flying blind? There’s an answer.
5. ‘We can’t connect cloud and AI spending to business outcomes, so every budget conversation turns into a fight.’
Only 43% of organizations track cloud costs at the unit level. For the rest, the same conversation plays out over and over. Finance sees a massive, growing spend line. Engineering sees the products, features, and customers that spending is enabling. Neither side can prove their point because there’s no shared language.
One exec described it to us perfectly: “A CFO wants to look at it from a balance sheet perspective. A CTO wants to confirm they’re innovating. A business leader wants to confirm the business model is changing. These three are on completely different pages.” Sound familiar?
Upstart broke out of this cycle using cost per product and cost per organization to create engineering accountability that translated directly to $20M in cost reductions, not by cutting, but by understanding. SmartBear did something similar. Unit cost data gave their senior VP visibility into cost per customer by segment, letting them improve margins, manage EBITDA, and reprice intelligently.
In both cases, the budget conversation stopped being adversarial. Everyone was working from the same, trusted number. When that bridge actually exists, the budget conversation stops being a negotiation and starts being a discussion.
These five problems are showing up in every budget meeting, board presentation, and FinOps backlog right now — increasingly raised by people who never thought of themselves as CloudZero’s audience before.
That’s actually the point. This isn’t just an engineering or FinOps conversation anymore. It’s a business conversation, and it’s one CloudZero is built to have.
See it in action at CloudZero.
The Cloud Cost Playbook
The step-by-step guide to cost maturity
