Govern RITA

01In a nutshell #sec_7dccec99

You decide for the entire subscription whether RITA exists, in which modes workspaces can use it, if your data can go to an AI model, and how many assisted questions the month can consume — each workspace can then only go below that.

02Where #sec_7a445843

Settings › Integrations & AI › RITA Governance. A single screen, from the broadest to the finest: the circuit breaker, the authorized modes, the default mode, telemetry, the month's allocation, deployment by workspace.

03The three modes #sec_695a156a

Two facts structure the rest. Guided mode is the engine's fallback when the provider is unavailable: as long as AI is authorized, it remains locked in the authorized position. And the choice of mode decides whether the workspace data is sent to an AI model — not the assistant's capacity. RITA answers in all modes.

04Authorize AI: the addendum #sec_70d2a05c

AI-assisted mode is not authorized with a click. You must first accept the AI processing addendum: the authorization will remain locked as long as the addendum is not accepted by a subscription administrator.

Accepting the addendum is an auditable act: it is recorded with your name and the timestamp, and unlocks the authorization. It is valid for the entire subscription. The screen then displays AI Processing Addendum accepted by … · date, and View addendum reopens it.

The AI provider is displayed next — (platform default) or the selected provider — with the phrase Data processed by: … This is the answer to the question you will be asked: where does the data go?

A subscription that authorized AI before the existence of this governance carries an inherited authorization: cutting it will require the addendum to reopen it.

05The default mode and deployment #sec_0ce9cf54

Default mode of new workspaces — among the authorized modes only. Applied at creation — each workspace can then derive from it.

Deployment by workspace lists each workspace with its current mode, its last change and by whom — subscription default when it has not chosen anything. Forcing a mode imposes a mode on a workspace: same grammar as the change on the workspace side: the workspace administrator will see the new mode, its author and the timestamp. Effective in less than a minute.

06The month's allocation #sec_2a261b78

The allocation is contractual and is settled with the publisher. It is shared by all workspaces and is never carried over from one month to the next. Only questions processed by AI consume it.

The counter tells you where you stand: there are N questions left before the end of the period, or allocation reached: the generative route is closed until the start of the next period. RITA continues to answer in guided mode.

Authorizing overage in credits opens a margin, bounded by a monthly cap in questions: a maximum of N questions beyond the included, or N credit(s) per month. One credit covers 100 questions initiated. A cap at zero prohibits any overage even if the option is checked; without overage, no credit is consumed. In overage, the screen counts: N question(s) in overage on a cap of N — N credit(s) debited this month.

07Telemetry #sec_34000972

Authorize the logging of unanswered questions: anonymized (hash category and length, never the content), logged solely to enrich the intent catalog. None of the question text is sent. If unchecked, the catalog learns nothing from your missed questions.

08The circuit breaker #sec_e942dd84

At the top, RITA activated for this subscription. Unchecking it opens Disable RITA across the entire subscription?: N workspaces · N members lose the conversational interface in less than a minute. The settings and prompts of each workspace are preserved — reactivation restores them. Each workspace screen will display Deactivated at the subscription level, with your name and timestamp. Recorded in the subscription audit.

This is the action for an incident — a suspected leak, a supplier to be removed — and it is done in a minute without losing anything. Reactivation restores everything as it was.

09What RITA never does, regardless of mode #sec_44abedf5

It does not decide. No setting on this page gives it a verdict to render; the reviewer's journey states it on its side, and it is true on both sides.

10To remember #sec_4950bcdf

Authorizing AI requires an addendum signed by you; guided mode is always there; each workspace goes down, never up; the month has an allocation and, if you wish, an overrun ceiling in credits. And a circuit breaker, for the day it is needed.