Economico for AI startups

Margin for AI startups with moving model costs

Plans, prepaid credits, and token costs on one ledger, so you know margin per customer even as inference prices move. Run by your AI agent.

The week inference gets more expensive, can you still tell if you're profitable?

Your provider changes token pricing on a Tuesday. Your Pro plan sells for a fixed price and bundles a fixed credit allotment — that number didn't move. The cost of serving those credits just did. So the honest answer to "are we still break-even on Pro?" is a shrug and a spreadsheet you haven't rebuilt since the last price change. One founder put the squeeze plainly: "credits just expired and we're not yet break-even." When your revenue is credits and your cost is tokens, and the token side moves weekly, margin per plan is exactly the number you can't see when you most need it.

Credits are revenue; tokens are cost; put them on one ledger and the guess disappears

The reason margin is hard isn't the math — it's that the two halves live in different places. Credits get sold in your billing tool; token bills arrive from your model provider; nobody keeps them on the same books. Economico does. Subscription revenue and the credits customers buy are recorded distinctly, so you can see what's recurring versus what's consumption. Your model-provider bills post as cost of revenue, tagged by model, so the cost of serving each customer is visible next to the revenue it earned. Same ledger, same period — margin per plan becomes a figure your agent reads off the books instead of one you reconstruct after the price changed.

  • Credit packs$9,000
  • Enterprise pilot$3,000
Money in $12,000 one ledger
  • Inference / API$4,200
  • Cloud$800
  • Margin kept$7,000
Credits in, tokens out — one ledger, real margin, even the week prices move.

Prepaid credits sit as deferred revenue until they're actually used

Plans, credits, and the costs behind them, all in one place.

  • Monthly plans that bundle a credit allotment.
  • Prepaid credit top-ups, parked as deferred revenue and recognized as they're consumed.
  • Usage and token metering underneath the credits.
  • Model-inference and hosting costs tracked as cost of revenue.
Credits + plans

Both revenue streams, booked

Subscription revenue and the credits customers buy are recorded distinctly, so you can see what's recurring and what's consumption.

Token costs

Inference as cost of revenue

Your model-provider bills are recorded as cost of revenue, tagged by model, so the cost of serving each customer is visible.

Prepaid handled

Credits as deferred revenue

Credit top-ups are parked as deferred revenue and recognized as they're consumed — the correct treatment, automatically.

Margin per plan

Profit you can read

With credit revenue and inference cost on the same ledger, your agent tells you the gross margin on each plan — even the week prices moved.

An AI startup's first quarter · in journal entries

Credits in, tokens out — margin you can read.

Scroll through Brightforge AI's first quarter — plans and credit packs on the revenue side, inference bills on the other, margin visible the whole way.

01

Incorporate and raise

Selene and Matias found Brightforge and close the YC standard deal. Half a million in — none of it revenue.

We incorporated Brightforge and closed the YC standard deal — set up the books.
4 tool calls
  • Economicocreate_share_class
  • Economicoissue_shares
  • Economicorecord_safe
  • Economicorecord_safe
Founder stock plus both YC SAFEs recorded — $525,000 in, revenue zero.
02

The cost side arrives first

OpenAI for inference, LangSmith for observability — the vendor contracts are already in the inbox.

Add vendors and receipts from email.
4 tool calls
  • Gmailsearch_gmail_messages
  • Gmailread_gmail_message
  • Economicocreate_party
  • Economicocreate_party
OpenAI and LangSmith on file — inference and observability, your cost side. Nothing posts until a bill lands.
03

Plan plus credits, billed and paid

Fern Logistics takes the AI Workflow plan with a credit pack — $3,200, collected within two weeks.

Bill Fern Logistics — the AI Workflow plan plus their credit pack.
3 tool calls
  • Economicocreate_invoice
  • Economicosend_invoice
  • Economicorecord_payment
$3,200 in — plan revenue and credit revenue on separate lines, so consumption stays visible.
04

The token bills land

OpenAI $1,450 and LangSmith $399, charged to the card. Revenue minus the tokens that produced it: real margin.

Find and handle any bills.
4 tool calls
  • Gmailsearch_gmail_messages
  • Gmailread_gmail_message
  • Economicoreceive_bill
  • Economicoreceive_bill
OpenAI $1,450 and LangSmith $399, card-charged. January margin on your first customer: $1,351 — revenue minus the tokens that produced it.
05

February: growth goes red

A second customer, a token overage — and a $5,500 prompt-engineering bill that outruns them both.

Close February — are we still break-even?
2 tool calls
  • Economicoget_income_statement
  • Economicoget_balance_sheet
No: ($998) cumulative. A second customer and a token overage, but the $5,500 PromptOps bill outran them. Exactly the week you need margin per plan.
06

The quarter closes in the black

Three customers, token prices that moved twice — and margin per plan readable the whole way.

Close the quarter — margin per plan after inference.
2 tool calls
  • Economicoget_income_statement
  • Economicosummarize_revenue
Revenue $16,050, inference and tooling $12,677, net $3,373 — back in the black, margin readable per plan even as token prices moved.

Ask whether each plan still clears its cost

Start a plan
Put Pine Street Agency on the $25/month Pro plan with 100 monthly credits.
Record inference cost
Record this month's $92,000 model-inference bill as cost of revenue.
Check your margin
What's my gross margin per plan after inference and hosting costs?

What you'll use

The money loop, tuned for ai startups.

Get started

The finance team you don't have to hire.

Hand your agent the setup guide and it walks through the rest — real books from day one, no dashboard, no finance hire.