Your Workforce Intelligence account: billing, keys and spend limit
Create a Workforce Intelligence account, see how it is billed (free, invoiced or card), create and rotate the API keys you paste into Settings, set a monthly spend limit, read your usage statements, see whether HIPAA mode is on and reset your password.
A Workforce Intelligence account lets your installation use models without you holding any model provider keys. You create an account, we set up how it is billed, you create an API key, and you paste the key into Settings, Administration, Model providers (see Add a provider). You are billed once a month for what you use, at the downstream model maker's published API rates plus a small markup that covers infrastructure, bandwidth and routing costs.
Create your account#
- Open the Workforce Intelligence account portal. On the sign-in page, select Create an account.
- Enter your email address and a password of at least 12 characters.
- Open the email we send and select the link, then select Confirm email. The link works once. If it expires or you lose it, signing in shows Send a new confirmation link.
- Sign in.
A new account can sign in, but its requests wait until billing is set up. What you do next depends on how your Workforce Intelligence service is set up.
- Invoiced billing only. Overview and Billing show "Your account is waiting for billing to be set up. To set up invoiced billing, contact sales at" the sales address, and there is no card form. Requests from your installation are refused with a message saying the same.
- Card billing available. Billing and Overview offer Add a card (see Add a card), with the line "Prefer to be invoiced? Contact sales at" the sales address. Requests are refused with a message that tells you to add a card in your account or contact sales to be invoiced.
In both cases API keys and Spend limit show a message in place of their forms, so you cannot create or rotate a key or change the limit until billing is set up; you can still see and revoke any keys you already have. Billing is set up when a card is on file, or when we choose how your account is billed, and then the message goes away, the forms return, and requests work. See How your account is billed.
If we cannot deliver email to your address, a notice at the top of the portal says we have stopped sending to it and asks you to contact us. See Undeliverable email addresses.
When your installation sends a request through Workforce Intelligence, we process the prompt and the response only to produce the response and do not store them. To produce the response we send the request to the AI model provider that answers it, and that provider handles what it receives under its own terms and privacy policy. We keep usage details (the model name you chose, token counts, cost and time), your email address and your billing records so we can bill you accurately and show you a statement. Your card is entered on a page hosted by Stripe and we never see the number. The Privacy Policy and Data handling and security list everything we keep, who else receives it and how long it is kept, and the Terms of Service cover charges, spend limits and acceptable use. To have your account data corrected or deleted, write to the contact address in the Privacy Policy.
How your account is billed#
We choose how each account is billed. Billing and Overview show only what applies to yours.
- Invoiced. Each month we create an invoice for what you used, with an invoice number and a due date 30 days after the month ends, and we send the invoice to you separately. Your portal shows the amount, number, due date and status; it does not take payment and nothing pauses your account because an invoice is late. There is no card form.
- Card. You add a card and we charge it each month. See Add a card.
- Free. You are not billed for usage while your account is free, and no card is needed.
A month with no charge shows No charge. An account with no billing choice waits, as described under Create your account.
Add a card#
Accounts billed by card, free accounts, and (when card billing is available) accounts with no billing set up have this. Billing has a Card section for them; Overview shows Card with "On file", or a link to add one. When you add a card to an account with no billing set up, the account is billed by card from then on.
- Select Billing, then Add a card.
- Enter your card on the secure page you are taken to. Workforce Intelligence never sees or stores your card number.
- You return to Billing, which shows "Card on file" with the card's brand, last four digits and expiry.
Until a card is on file, requests are refused with a message asking you to add one. If the card page cannot be started because of a problem on our side, it says "Card setup isn't available right now. Your account and usage aren't affected." and gives the sales address so you can be invoiced instead. Replace card changes the card. If a monthly invoice cannot be collected, the account is paused and requests stop; update your card, and usage resumes once the invoice is paid. Invoices appear on the Billing page after each month ends.
Create and rotate API keys#
- Select API keys, optionally type a name, and select Create key.
- Select Copy key to copy it now. It is shown once and cannot be recovered.
- Paste it into the Workforce Intelligence provider in your installation.
The list shows each key's name, an identifier, when it was created and when it was last used.
To replace a key with no downtime:
- Select Rotate on the key. A new key with the same name is shown once.
- Paste the new key into your installation.
- Select Revoke on the old key, then select Revoke key to confirm. It stops working within a minute.
Revoke a key at once if it may have been exposed. You can have up to 20 active keys.
Model names#
Once the Workforce Intelligence provider is added, the product lists the model names it offers. There are two kinds.
- Workforce Intelligence's own names (
reasoning-low,reasoning-mid,reasoning-high, the same withvision-,embedding-andimage-) pick a level of quality and cost. We choose the model that serves each request and can move a request to another model family if one is busy. Theembedding-names are the exception: each one always uses the same embedding model, so the vectors your Knowledge Bases store stay comparable. If that model cannot answer, the request can be served by the same model through another of our providers, and otherwise it fails and can be retried. - Model family names (for example
claude-haiku,claude-sonnet,claude-opus,gpt-luna,gpt-terra,gpt-sol,gemini-flash-lite,gemini-flash,gemini-argonandgrok) always use the newest version of that family. Each belongs to a level (low,midorhigh;grokismid). If the maker of that model cannot answer, a request can use the same model through another of our providers (Claude, GPT and Grok through Amazon Bedrock), or the equivalent model of the same level from a different maker. A model family name is listed only while at least one model that can answer it is available.
Every request is billed at the published API rate of the model that answered it, plus the same small markup for infrastructure, bandwidth and routing. When an equivalent model from a different maker answers, the price of that request follows the model that answered it, so it can be higher or lower than it would have been, and the statement marks such requests as answered by an equivalent model. When the same model answers through Amazon Bedrock, the request is billed at Bedrock's rate for that model, which can be higher than its maker's own rate; the statement does not mark it, because the model is the one you asked for.
Read your usage statements#
Select Usage to see one row for each month: the number of requests, the amount billed and, once the month has ended, the state of its invoice. For an invoiced account the invoice column shows "Invoiced", the amount, the invoice number and its status (Open, Sent, Paid or Void). Select a month to open its statement.
A statement shows the Total billed for the month, the invoice for that month (for an invoiced account, Invoiced with the amount, number, due date and status, and a note that the number is a reference for your records and the invoice itself is sent separately), and two tables: one By model (requests, input tokens, output tokens, and the amount billed for each model name) and one By day (UTC). When an equivalent model answered any request, the statement also shows how many were Answered by an equivalent model. The total is the amount we bill: it equals the sum of every request listed, and the invoice is that total in whole cents, with any part of a cent carried to the next month's invoice.
Select Download every request (CSV) for one row per request: its id, time (UTC), model name,
status, token counts, amount billed in US dollars and a fallback column (true when an equivalent model
answered the request). The file never contains prompts or responses.
A month with a very large number of requests shows a summary of the first requests only; the total
and the download always cover the whole month.
HIPAA mode#
HIPAA mode limits your account to HIPAA-eligible models. Workforce Intelligence turns it on for an account once a signed business associate agreement (BAA) between your organization and Workforce Intelligence Inc is in place. To have HIPAA mode turned on or off, contact Workforce Intelligence support or your Workforce Intelligence account contact.
While HIPAA mode is on, Overview shows a HIPAA mode card that says "This account runs only on HIPAA-eligible models." The card has no controls: you cannot change HIPAA mode from your account. While it is on:
- Every request from your account is served only through HIPAA-eligible services. A request fails when none is available for it.
- A Knowledge Base built with a search model that has no HIPAA-eligible service stops searching.
While HIPAA mode is off, Overview shows no HIPAA mode card.
Reset your password#
- On the sign-in page, select Forgot your password?.
- Enter your email address and select Send reset link. The page gives the same answer for every address; if the address belongs to a confirmed account, we send a link.
- Open the link within an hour, choose a new password of at least 12 characters and select Save new password. The link works once.
Saving a new password signs you out on every device, and we email you a notice that it changed.
Sessions and signing out everywhere#
You are signed out after an hour without using the portal, and after 12 hours in any case. To end every session on every device now, open Overview and select Sign out everywhere.
Set a monthly spend limit#
- Select Spend limit.
- Enter the most you want to spend in a month, in US dollars, and select Save limit.
The limit applies to your very next request and resets at the start of each month (UTC). Select Remove limit to run without one.
With a limit, spending is spread evenly across the month. If you use your budget faster than that
pace, Workforce Intelligence slows requests to the pace your remaining budget allows. A wait of up
to 90 seconds is retried automatically and the reply simply arrives a little later. A longer wait
is not held open: the reply says so and states how long until the next request can run, in about
minutes, hours or days. Near the end of the month's budget, when the pace is almost used up, that
can be days. Chat shows this as "This workspace is pacing replies to stay within its monthly
limit" with the time to wait, and suggests an admin can raise the limit. When you reach the limit,
requests are refused (error spend_ceiling_reached, HTTP 402) and chat says the workspace has
reached its usage limit; requests resume next month or when you raise the limit. A request that starts while you are under the limit always finishes and its
full cost counts, so one long request, or several at once, can take the month's total above the
limit by their whole cost.
If you lower the limit to an amount you have already spent this month, the page asks you to confirm, because requests stop immediately. Without a limit there is no pacing and no stop.
For example, with a limit of $100 you can spend about $10 in a quick burst. Beyond that, requests are slowed to the rate your remaining budget allows over the rest of the month, so heavy early use slows later requests more. The amount shown as Spent this month on Overview is the same total the month's statement shows.