Quick Answer
Activity categorization classifies every AI request by the work it performs, such as meeting prep, code review or account research, then assigns that spend to the team, person, model and provider behind it. It turns a provider invoice that shows only totals by model and API key into a breakdown finance can allocate, charge back and budget from. CloudZero AI Signals does this as the spend happens.
If you’re an enterprise finance leader struggling to make sense of AI spending (and trying to set your budgets for next year) here’s an illustrative scenario that might help.
Say your chief revenue officer decided in July that all 700 sellers across North America, EMEA and APAC should use AI before every customer demo. Reps use it to research the account and review past calls. The CRO didn’t set a specific budget for it, but from the looks of things, the sales team is using AI aggressively.
Then the October invoices arrived from Anthropic, OpenAI and Azure, and AI spend was up seven figures on an annual basis since June. The CEO asks how much of that is the sales program. The CRO doesn’t know. FP&A looks at the invoices and finds totals by model and API key, but no line for Sales, no regions and no names. Nobody in the company can see what the money was used for.
Now run the same scenario with CloudZero in place. You open AI Signals and find a line for demo prep. It cost the sales org $206,000 in October, about $2.5 million a year at that rate. Most reps spent $200 to $350. One 45-person mid-market team in EMEA spent $41,000, three times the average per rep, almost all of it using the most expensive AI model.
The invoice is where most finance teams stop looking. It does not have to be.
The AI invoice doesn’t detail the work
This year, most enterprises gave AI to every team and planned to sort out the cost later. Then the bills caught up. Enterprises now project nearly double the AI spend over the next twelve months that they projected a year ago, according to KPMG’s Q1 2026 survey of more than 2,000 technology leaders.
Enterprise leaders still want AI everywhere, but finance wants to see inside the spend before approving more. In our survey of 260 senior finance professionals, three quarters of teams that cannot measure AI outcomes have already held back on AI investment, and two thirds of boards now condition further funding on proof of return. The questions are the same everywhere: who is using AI, how much, and for what work?
The Anthropic (or OpenAI or Azure) invoice can’t show that. The AI spend has no tags, cost center or owner, so finance either approves a number it can’t break down or caps it.
2026 State of AI Spend Report
Why Finance Caps What It Can’t See
Learn how finance leaders are managing AI spend, and how they’re approaching FY2027
Activity categorization shows the work behind every AI dollar
CloudZero’s activity categorization assigns every AI request to one of more than 35 business activities in plain language, with each team using AI. This is the unlock CFOs and AI champions need not just to explain spending on AI, but how they’re budgeting for next year and trying to determine: what’s worth spending on and what’s worth pulling back on?
Here are some of the categories CloudZero tracks today:
| Admin and back office | Research and development | Revenue generation |
|---|---|---|
| Recruiting | Product planning | Campaign analytics |
| Financial analysis | Feature development | Pipeline analysis |
| Legal compliance | Code review | Account research |
| Security analysis | Debugging | Deal strategy |
| IT support | Testing | Lifecycle comms |
CloudZero classifies each request by the work itself, regardless of job title. When a rep asks a model to summarize a prospect’s last three calls before a demo, CloudZero categorizes the activity as meeting prep and assigns it to that rep, team and region. CloudZero’s allocation engine then assigns the spend across your org by department, team, person, model and provider, plus your own categories such as business unit or cost center.
No other cost platform classifies AI spend by the work being done and ties it to the people doing it, in one view, as the spend happens.
The prompts of your AI users stay private; classification runs inside your own network. CloudZero receives only the activity, the model and the token count.
What you can do once you can see it
First, you can allocate it. The $206,000 now belongs to Sales, broken out by region and team. This means you can charge it back to the right cost centers.
Second, you can judge it. Pull conversion rates for those same teams from your CRM and compare them to what each team spent on demo prep, week by week. That comparison tells you whether the reps who used AI the most are turning more demos into real pipeline. If the heaviest AI users are converting more demos, you can fund the program further. If the EMEA team spends three times as much without converting better, finance and sales leadership can find out why that team uses so much more AI, starting with which model its reps rely on and whether a less expensive one would do the job.
Third, you can plan with it. If you’re building 2027 budgets from this year’s AI totals, there’s a good chance you’re guessing what will happen next year. Now, you can budget by activity, such as 750 reps at about $250 a month for demo prep. You can set limits by business unit and activity instead of one company-wide cap that stifles productive AI work.
Every function has a line like this
We ran activity categorization on ourselves. Customer Success turned out to be CloudZero’s second-largest AI spender behind Engineering, on work ranging from knowledge base authoring to renewal prep. Nobody would have guessed that from the provider bills.
With CloudZero, you can see the work by team and by task before you set next year’s budget.
What CFOs can do with CloudZero beyond the AI bill
Activity categorization tells you who is using AI and what work they’re using it for. CloudZero builds on that so finance leaders can manage AI the way they manage any other major investment.
- Connect AI spend to business outcomes. CloudZero ties each AI dollar to the team, product, feature or customer it supported. You can see which spending drives revenue, faster delivery or better customer service.
- Measure the return on AI. Once you know what work the money paid for, you can compare it with what that work produced. Then you can tell the board what the company got for its AI budget.
- Know your cost to serve. CloudZero calculates unit costs, such as cost per customer, per feature or per AI agent.
- Catch spikes before the invoice arrives. CloudZero monitors AI and cloud spend continuously, at hourly granularity.
- Manage AI and cloud spend in one place. The same allocation engine covers your model spend and the cloud infrastructure underneath it, and reconciles both to the invoice. Finance and engineering work from the same numbers.
See how 250+ finance leaders are approaching FY2027 AI budgets in Why Finance Caps What It Can’t See, CloudZero’s new AI spend research report.
