Skip to content
IntellobotsTechnologies

Set up · How-to

AI models and governance

Choose which AI providers may be used for what, set budgets, approve skills, and switch AI off where needed.

Updated 11 Nov 20261 min read

All Tracewright pages

Tracewright works with the AI providers your organisation approves, including self-hosted models, and works fully without AI.

AI Control Centre

AreaWhat you set
ModelsWhich providers and models are available, with your organisation's own keys
BudgetsMonthly token budgets per organisation and project (0 means no AI at all)
SkillsThe prompts agents use; each version is evaluated and approved by a different administrator
ConnectorsTools agents may read through, with read-only limits
AssistantSwitch the assistant on or off
ActivityEvery AI call, with who, what and cost

AI governance

Administration → AI governance decides who may use which AI for what: data classes (which models may see which kind of information), approved models, human review rules and the AI risk register.

Per project

Project → Settings → AI sets which agents run (for example the meeting transcriber and drafter), which approved models they use, and whether the assistant is on. A project can switch AI off entirely.

When something looks wrong

If you suspect an agent was manipulated by something in a source, the source shows flagged statements. Governance can switch AI off for the project at once, while people keep working by hand. Every AI run keeps its full trace for investigation.

Cost

AI cost shows tokens and cost by person, project, agent and model over time, against the monthly budget.