Somewhere in your company, a team is three weeks into the AI project that’s going to matter. Somewhere else, a support pilot from the spring is still summarizing every ticket with a frontier model, and nobody has looked at it since it started working. On the invoice they’re identical, and the company has two moves: leave everything open, which funds the waste, or cap everyone, which kills the bet.
Telling them apart is your job. You’re the AI champion, the one a CFO or CIO trusts with this budget. When it doubles with nothing under it but a list of people and API keys, finance reaches for the sentence you’re trying to get ahead of: people or AI, pick one. A blanket cap is what a company does about spend it can’t value, and it lands on your best people first.
Why you can’t tell them apart yet
Engineering can mostly account for its AI spend: a key per service, an account per team, tooling that counts every call. Everywhere else in the company, none of that scaffolding exists. Finance is running the close through Claude and Gemini at the same time. Sales is doing account prep in ChatGPT. Marketing is generating campaign images and video. Each one shows up as a person or a key, and a total. None of it is misuse, and some of it may 10X the company. The invoice can’t tell you which.
Signals, not statements
A bill is a statement: weeks late, everyone totaled together, and the only move it leaves you is reacting to the number. A signal arrives while there’s still a decision to make, and it goes to the person who can make it. The manager whose team’s spend doubled yesterday hears about it today, with the job that did it named, while letting it run or stopping it is still a choice.
I know how a screen with everyone’s AI spend on it sounds to the people on it. Your teams don’t need a policeman. They need someone outside the work who can see the whole picture and get word to them in time. The engineer with the biggest bill in the company may be doing work for three teams, on a model that costs more per call and saves hours of correction afterward. She knows that. Her manager and her CFO need to know it too, so they can back her instead of capping her. You are not watching your team. You are watching out for them.
What you can do now
Catch the overspend before it forces the cap. AI Signals is where AI spend and usage live in CloudZero: all of it, across providers, in one place. Overview, the new console, shows it as it’s happening. Monitors, the new alerting capability, watches it for you: each team’s spend against its own normal pattern, and you tell it what you care about. When a team’s spend breaks its pattern, you find out with the team, the people, and the work attached, and you can open an incident with all of that already in place. The agent somebody left in a retry loop gets caught on Saturday, and it’s a conversation with one engineer, not a budget problem.
Apply the smarter limits you already know you want. When a person or an API key is the finest grain you can see, it’s also the finest grain you can manage. Overview shows AI spend by department, by activity type, by model and project, with the work behind every dollar. Click a user and their spend breaks down by what they were doing, prospecting or code review or content drafting, not just by which model they called. Once you can see what the ticket summaries are running on, you can move them without touching anyone else.
Scale the workflows that are working. AI Signals tells you what’s happening. Deciding where to put more money is a different question, and it runs on the rest of CloudZero: Explorer, and the allocation engine underneath it. Every AI dollar gets assigned through the dimensions your business runs on (customer, product, team, project, whatever you’ve defined), with totals that match the invoice and always foot to the same sum. That isn’t the live view. It’s the reconciled one, and it’s the one you take to finance. Two people on the same enterprise plan, calling the same model on the same day, land in different products and customers. The engine doing that has run CloudZero customers’ cloud spend in production for ten years, and a dollar from an engineer’s API key and a dollar from a marketer’s seat go through it the same way. When a project is paying off, you can see what it cost across everyone and every model that touched it, and that’s what you need to put more behind it.
I get asked whether your provider’s console can do this. It knows which key spent what, and it’s fresh. It’s also one of five, and nobody is reading all five. What none of them know is which repo the key belongs to, what project the work was for, or that the same project ran on another provider last week. That context lives where the work happens, and that’s where CloudZero collects it.
Your CFO sees this through you. You walk into the budget conversation with the split already made and tied to the invoice: this is the spend that’s on the right things and what it’s for, and this is the waste we already cut. People or AI, pick one gets a third answer. All of it is in CloudZero today, no waitlist, no alpha. You’ve been doing this job on instinct and a bill that arrives too late to help. As of today you can see which is which. Cap the accident, not the bet.
Come look at it
Visit our website to request a demo, read more, or see a short video highlighting how AI Signals works.
