Trust & audit
Two read-only monitoring views for AI autonomy — the Trust Dashboard of per-agent accuracy and the Audit Log with a full decision trace of every action.
Two read-only screens let you watch how your AI agents are performing and account for every decision they make. Both sit under Autonomy Settings.
Trust Dashboard
Trust Dashboard (/settings/autonomy/trust) reports "Per-agent accuracy metrics and trust
progression." Each agent card shows its total decisions and an accuracy percentage — color
coded so 95%+ reads healthy — with a per-action-category breakdown beneath it. A time-window select
scopes the numbers to Last 7 days, Last 30 days, or Last 90 days. Accuracy is what earns
an agent higher autonomy, so this is where you check whether one is ready to be trusted with more (or
has slipped). With no history yet it reads "No trust metrics recorded yet. Metrics will appear once
agents start processing actions."
Audit Log
Audit Log (/settings/autonomy/audit) is the "Complete decision trace for every action." Each
row records the timestamp, the agent, the action category, the confidence, and the outcome —
outcomes range across approved (no edit, minor edit, or major edit), rejected, rolled back,
downstream failure, and SLA miss. Expand a row for the action ID, exact confidence, any human
modification, and notes. Three filters narrow the trace — Agent, Category, and Outcome —
and long logs paginate with Previous and Next. Nothing here is editable; it is the
system of record for what the agents did and why.