Show your work,
or don't expect to be trusted.
Five things the platform does that most monitoring tools don't. Not features — disciplines. Each one is engineered into a surface you can audit.
Nothing here is inferred.
Carol's memory contains only what you've explicitly marked — pinned messages and summaries of conversations you've had. She doesn't synthesize a profile of you out of patterns. Export it as JSON. Purge it whenever. The thing that gives her continuity is the thing you can audit and erase.
Settings → Memory · "Carol cites real artifacts — pinned messages you marked, or summaries of conversations you had. Nothing here is inferred. Unpin or delete anything at any time."
Every assessment gets reviewed against what actually happened.
When a situation resolves, the platform pulls the original Council determination back up next to the outcome and asks: was the read correct? The accuracy log is part of the Post-Situation Debrief. Repeated misreads get flagged. Carol's confidence calibrates from her track record, not her self-image.
Debrief → Council Accuracy · "Assessment: Within expected range. Accuracy tracking improves as the platform learns your entity's baseline over time."
Sometimes Carol's best advice is to say nothing.
Every opportunity brief has a What Not to Do callout: the move that would backfire if you made it. Don't run a press push when amplification would wake a settlement story. Don't engage a refuted claim that's losing oxygen. The platform is in your corner — even when that means telling you it's not the moment.
Opportunity Brief → What Not to Do · "Overproduction risk — amplifying too aggressively can undermine authenticity. Monitor sentiment for the first 24h after engagement."
When the words get legal, you find out separately.
Sentiment tools blend legal-language into the general sentiment score. We don't. A dedicated detector pulls out language that names a lawsuit, threatens one, or could create securities exposure — surfaced as a discrete alert with specific article citations. Your counsel gets the right material; your comms team gets a clear handoff.
Entity Overview → Legal Content Detected · banner alert listing specific article titles + Override-with-Note + Request Legal Review.
What's normal for your brand isn't ours to assume.
The Reputation Standing band — what counts as your brand's normal — is calibrated from mentions you reviewed yourself in onboarding. You marked the routine ones routine, the threats threatening, the irrelevant ones irrelevant. The Council uses that to detect deviations during live scanning. The baseline is auditable because you wrote it.
Baseline Training · "Use Add to Memory for normal content, Flag as Threat for concerning content, and Ignore for irrelevant items. The Council uses this baseline to identify deviations during live scanning."
We provide the map.
You determine the course.
Every claim on this page is engineered into the product. Take a look — the audit trail is there.