DPO Radio

Reconciling Usage With Your Provider's Invoice

Overview Image

Why Reconciliation Matters

A country compliance manager presenting an AI budget to finance has to defend the numbers, not just report them. "Where does that figure come from, and how do you know it is right?" is the question finance asks every quarter. An invoice total that nobody can go behind does not survive the audit committee.

Your organisation pays its AI provider directly, so the provider's invoice is the bill. What the invoice cannot show is which work produced it. Forseti keeps that side of the record. Every request is appended to a usage ledger with the provider, model and effort it ran on and the tokens it used. The Usage page reads that ledger for any period you choose.

Reconciliation is setting the two side by side. Choose the invoice's dates, read Forseti's totals by provider, model and effort, and compare. The provider's figures remain what you pay. Forseti's figures explain how its requests made them up.

How It Works

A Ledger Row per Request

Every Forseti request is written to an append-only usage ledger. Each row carries the feature, the provider, the model, the effort, the input, output and cached input tokens, and the time. The ledger service can add a row and has no way to change or delete one.

The Invoice's Own Dates

The Usage page shows the current month by default. Choose a start and an end date to match the period on your provider's invoice.

Totals by Provider, Model and Effort

For that period the page shows the number of requests and the token totals, then a table by provider, model and effort. This is the table to compare with the invoice's line items.

Totals by Feature

A second table splits the same period by feature: chat, workflows, tabular reviews, document redlines and titles. It explains a month that cost more than the last.

When the Two Differ

Your provider's figures are what you are billed for. A difference can have ordinary causes: other systems using the same provider key, the connection test, or requests made outside the chosen dates. Use a key that only Forseti uses if you want the two to match line for line.

An Internal Check Behind the Numbers

Forseti also keeps a fast running total of each month's consumption beside the ledger. A background check compares the two every five minutes. A difference of more than 1% and more than 1,000 units raises an alert to AesirX operations, because it would mean a platform defect, not a question for you to answer.

The Difference

Graphic Image

Without a Request-Level Record

Graphic Image

With Forseti Usage Reconciliation

IconThe provider's invoice is the only figure, with nothing to go behind it.
IconEvery request is its own append-only ledger row with provider, model, effort and tokens.
IconFinance cannot see which features or models produced the month's cost.
IconThe Usage page totals any period you choose, including the invoice's dates.
IconA change of model or effort cannot be traced to a change on the bill.
IconTotals by provider, model and effort sit beside the invoice's line items.
IconA key shared with other systems mixes Forseti's use with everything else.
IconTotals by feature explain where the month's tokens went.
IconDefending the number is a matter of trusting whoever reports it.
IconA background check keeps the running total and the ledger in agreement.

Built for Auditable AI

Build For Image

The usage ledger is append-only by construction: the service that owns it can add a row and has no operation that updates or deletes one.

Build For Image

The internal check reads both of Forseti's records and writes to neither total. It does not move ledger rows to match the running total, or the total to match the ledger. A difference is raised, never absorbed.

Build For Image

The page states its limits. It counts Forseti's requests; it does not price them. Prices and the invoice come from your provider, under your organisation's agreement with it.

Overview Image

Include Delegated Reading in the Review

Delegated reading is tracked separately in Forseti Usage. Include it when reconciling the requests made under your configured model service and compare matching dates, models and provider totals.

Background Image

See Reconciliation in Action

Ready to see a month of Forseti requests by provider, model and effort, set beside a provider invoice for the same dates? Request a personalised demo.

Demo Image
Ronni K. Gothard Christiansen

Ronni K. Gothard Christiansen - Technical Privacy Engineer & CEO

Technical Compliance Expert, 32+ Years Open Source Advocate, X-BoD Open Source Matters Inc.

Or contact via

ronni@aesirx.io+84 909 500 760

Next Steps

Icon Image

Start a Compliance Pilot

Run a month on your own provider account, then set the Usage page's totals beside the provider's invoice for the same dates.

Icon Image

Discuss Your Compliance Needs

Talk to our team about how Forseti's usage record supports your budget conversations with finance.

Frequently Asked Questions About Reconciliation

Your AI provider's invoice against Forseti's own usage ledger for the same dates. The provider counts and prices the tokens; Forseti records each request it sent with the provider, model, effort and tokens. Compare the Usage page's table by provider, model and effort with the invoice.

The provider's invoice covers everything sent with that key, while the Usage page covers only Forseti's requests in the chosen dates. Other systems sharing the key, the connection test, or a difference in the period's boundaries can all explain a gap. A key used only by Forseti makes the comparison simplest.

Not from the Usage page today; it has no export. The figures are read on screen for the period you choose. What your organisation paid is on the provider's invoice; what Forseti sent is on the Usage page.

No. Forseti's figures are its own record of what it sent and are never adjusted. The internal check compares Forseti's running total with its ledger and raises a difference to AesirX operations instead of absorbing it.

Yes. The Usage page and the internal check are the same on Light, Standard and Enterprise. A person who can configure the AI provider sees the whole organisation's requests; other people with access see their own.