# AI spend you can see, cap and explain.

> Budgets for every team, person, API key and agent, alerts before a budget runs out, the cost of every agent run, and pricing by consumption, not per seat.

- Canonical: https://secondstack.ai/for/finance/

---

SecondStack is an AI agent platform. This view is written for finance: AI cost management with budgets by team, person, API key and agent, and the cost of every run.

Priced by consumption, not per seat. Every team, person, key and agent has a budget, and every run shows what it cost.

Illustrative example. A budget alert in ControlTower: on October 21 the Support team passed 80 percent of its October budget, $860 of $1,000, and the alert went out. Its three biggest lines: the named agent @support-specialist, Lena R. and the API key helpdesk-sync.

## AI cost management that follows your org chart.

Budgets go from organization to team to person. API keys, which your own apps use to call AI models, and agents get budgets of their own. Every call is charged to one of them, so every cost has an owner.

An admin or team lead can top up a person's budget for the month.

## Every run says what it cost, and who paid.

A named agent runs on its sponsor's budget (the person or team that funds it) or on the budget of whoever calls it. A personal agent runs on its person's budget.

## Pay for use, not for seats.

A per-seat license charges for every person, whether they open the assistant or not. With SecondStack you pay for what people and agents use, and model usage stays on your own provider accounts.

Illustration with the same forty people on both sides. Per seat: all forty seats paid every month, twelve of them active. By consumption: each person costs what they used, so the twelve active people carry the spend and the other twenty-eight cost little.

You get two bills: your AI providers bill model usage on your own accounts, and SecondStack bills by consumption. Pricing on request: ask us for the current terms.

What a model costs before you allow it: [Model prices](https://secondstack.ai/llm-pricing/)

## Questions finance teams ask.

### Do we get two bills?

Two, yes: your providers bill model usage on your accounts; SecondStack bills by consumption, not per seat.

### Is there a per-seat fee?

No. Pricing is by consumption, not per seat, and model usage stays on your own provider accounts. A person who rarely uses it costs little; a team that runs agents all week shows up on its own line.

### Who pays when an agent runs?

Its sponsor, the person or team that funds it, or whoever called it. A personal agent's runs go on its person's budget. Every run's cost is on its record.

### Can we give each team a budget?

Yes. Budgets go from organization to team to person, with budgets for API keys and agents. They take effect the moment they are saved, and alerts go out when spend passes a threshold you set.

### How do we see what a model costs before we allow it?

Model prices shows the list price of each model per cloud. In ControlTower, the admin console, a model has to be on a team's allow-list before anyone on that team can call it, so the price is known before the spend starts.

## Further reading

- [LLM cost management: budgets, alerts, per-team visibility](https://secondstack.ai/blog/llm-cost-management/)

## Bring last month's AI invoices.

We show how that spend would map onto budgets by team, and which line a single agent run would land on.

Book a demo: hello@secondstack.ai