Every Nova Answer Shows You Why to Trust It.
A Confidence Card appears with every Nova response — showing source strength, knowledge age, owner, risk level, and whether the answer is grounded in approved organizational knowledge.
A confidence percentage has no meaning without context.
Most AI systems express confidence as a probability. That number has no meaning without context: confidence in what? Based on what sources? Reviewed by whom? How old is the underlying information?
AI answers that look confident but are ungrounded
A fluent, confident-sounding AI output with no connection to approved organizational knowledge is indistinguishable from a well-sourced one — unless you have a Confidence Card.
No visibility into source quality or freshness
Without knowing the age and quality of the underlying sources, you cannot know whether the answer was based on a review that happened last week or two years ago.
No link between answer and human-approved knowledge
An AI answer from a general model and an AI answer from approved organizational knowledge look identical unless the system explicitly surfaces the difference.
Six fields. Everything you need to decide whether to act.
01
Answer Status
approved | partial | ungrounded | fast — whether Nova found approved records that match, found partial coverage, found nothing, or recognized an identity question not requiring knowledge retrieval.
02
Source Strength
high | medium | low | none — based on how semantically well the retrieved articles match your question. Computed from cosine-similarity scores at retrieval time.
03
Knowledge Age
Days since the most recent approved source was reviewed. Stale knowledge is flagged visually so you can decide whether to escalate for re-review.
04
Owner
The named knowledge owner of the primary source article. If no owner is assigned, the card shows unowned — which is itself a risk signal.
05
Risk Level
critical | high | moderate | low — computed from article criticality set by the reviewer, combined with staleness and ownership status.
06
Voice Identity Confirmed
Whether the speaker who provided the underlying knowledge was voice-verified at capture time. Only shown when Voice Identity is enabled in the workspace.
Triage AI answers before they become decisions.
Know when to act and when to escalate
Confidence Cards give you the signal to act on a Nova answer immediately or escalate for human review — based on real source quality data, not AI-generated confidence scores.
Surface stale or unowned knowledge
A Confidence Card with a high knowledge age or an unowned source is a signal to review. You catch the gap before it affects a decision, not after.
Brief leadership with evidence
When you take a Nova answer into a leadership review, the Confidence Card is your evidence that the answer came from approved organizational knowledge — not a hallucination.
Questions about Confidence Cards.
What does 'approved' answer status mean?
Nova found at least one approved organizational knowledge article that directly addresses your question, and used it as the primary source.
What does 'ungrounded' mean?
Nova could not find approved knowledge that matched your question. It will tell you what it cannot answer rather than fabricating a response.
What is Risk Level based on?
A combination of article criticality (set by the reviewer), staleness (days since last review), and whether the article has an assigned owner.
Can I act on a partial or ungrounded answer?
That is your decision as a qualified reviewer. Nova shows you the confidence level so you can make an informed judgment about whether to escalate.
Start briefing with confidence.
Every Nova answer in DebriefCore includes a Confidence Card. No additional setup required.