AI Tokens & Limits

AI usage in DocMgt is measured in tokens. A token is a small unit of AI work — roughly a fragment of a word — counted on everything the AI reads (input) and everything it writes (output). DocMgt tracks tokens so usage can be capped and reported.

How tokens are counted

The token figure you see in reports and gauges is a weighted number, not a raw count. Weighting reflects real cost and value:

• The AI's output counts for more than its input, because generating text is more expensive than reading it.
• Content the AI can re-use from cache counts for much less.
• The three feature groups are weighted differently. Configuration AI work (creating Record Types, E-forms, workflows, etc.) counts for more per token than End User work, because each Configuration result produces far more configuration value. Automation AI — the AI actions that run inside your workflows — is weighted lowest of the three, because it is priced for processing documents in quantity.

You don't need to do this math yourself — the reports show the final weighted totals. Just know that the token numbers are value-weighted, so a few Configuration operations can use more of the limit than the same number of raw tokens spent on, say, an AI Search.

Three monthly pools

Each pool is counted and capped separately, so heavy use in one can never starve another. All three are configured under Admin > AI Configuration > AI Settings Settings and reset at the start of each calendar month. Months are measured in UTC, the international time standard, so the reset happens at UTC midnight — during the evening in US time zones. Every AI usage figure, limit and report in DocMgt uses UTC, so they always agree with each other and with billing.

What the limit values mean

A limit of 0 means the pool is disabled, not unlimited. Nothing in that group will run until you enter a value. This is the default for the Configuration and Automation pools, so both start switched off on a new site.

The Workflow AI Limit additionally offers No monthly limit, set with the tickbox beside the field rather than by typing a value. This is the only "unlimited" setting in DocMgt, and it exists because Automation tokens are consumed by processes rather than by people: a large batch or a misconfigured workflow can spend a great deal without anyone watching. Turning the ceiling off is therefore a deliberate, confirmed choice, and while it is set a red warning stays on the AI Settings page showing what the site has spent so far this month.

Because the Automation pool starts at 0, the AI workflow actions cannot be added to a workflow until you set a limit. They still appear in the action list, greyed out, with a note pointing back to AI Settings — so a missing action is never a mystery.

What happens when a limit is reached

The limit is checked at the start of each request or action, before any work is done.

For the End User and Configuration pools, the request is refused with a message like "AI token limit reached. You have a monthly limit of N tokens and have used N this month."

For the Automation pool, the workflow action fails — the same as any other action failure, so it follows whatever error handling that action already has. The message names the limit and what has been used. Raise the limit (or wait for the new month) and the work item can be run again.

Because the check happens up front, a request that starts just under the limit is allowed to finish even if it ends slightly over. The Automation pool stops at 99% of its limit rather than 100%, which leaves room for the action that is about to run. On a small limit a single large document can still carry you past the number you set, so if you must not exceed a figure, set the limit a little below it.

Watching usage

• Usage gauges appear next to many AI features and on the AI reports, turning from green through amber to red as you approach the limit.
• Usage reports — the AI usage report (AIProcessing.html) breaks the End User and Configuration pools down by user, and the AI workflow report (AIWorkflow.html) covers the Automation pool and shows a month-to-date gauge against the Workflow AI Limit.
• Token variables — [TOKENLIMITUSER], [TOKENUSAGEUSER(From|To)], [TOKENLIMITAUTOMATION], [TOKENUSAGEAUTOMATION(From|To)] and their relatives let you show usage inside E-forms and documents, and test it inside workflow rules. See AI Variables.

Building your own limit alerts

DocMgt does not email you when a pool gets close to its limit. That is deliberate: you already have a better tool for it. Because the limit and the usage are both available as variables, you can build exactly the alert you want with an ordinary Send Email action and a rule — and you can decide the threshold, the recipients, the wording and the frequency, rather than accepting ours.

The usual shape is a scheduled workflow that runs daily, with a rule comparing [TOKENUSAGEAUTOMATION] against a percentage of [TOKENLIMITAUTOMATION], sending mail when it crosses. AI Variables shows the pattern in full, including how to avoid alerting on a site that is uncapped or switched off.

Quality levels affect cost

Several AI features offer a quality choice (for example Economy / Balanced / Maximum on the extraction actions, or Balanced / Maximum on Auto-Annotate). Higher quality reads more carefully but costs more tokens, and the weighting reflects that — a row processed at Economy is discounted against your limit. Each action's own help page describes its quality options.