Confidence Scoring & Gap Detection

Every answer CoTeam gives — through Ask, the Chrome extension, Forward-to-Answer, or a questionnaire review — carries a confidence score. This is the mechanism that keeps CoTeam from ever quietly guessing.

The three tiers

Tier	Meaning	What to do
🟢 High confidence	Directly sourced from your knowledge base, with a citation to the exact document.	Safe to send as-is.
🟡 Medium confidence	Inferred from related content in your knowledge base — not a direct match, but a reasonable extrapolation.	Review before sending. CoTeam shows its reasoning and the sources it drew from.
🔴 Low confidence	No good source found in your knowledge base.	CoTeam does not fabricate an answer. It flags the gap and names the specific document that would resolve it.

Confidence is always shown alongside its source citation — you can see exactly which document an answer is grounded in, not just a percentage.

Gap detection

When CoTeam can't find a good source, it doesn't guess and it doesn't stay silent. Instead:

• It names the missing document specifically. Not "add more docs" — a concrete recommendation like "Add your SSO/provisioning guide," based on what the question actually asked.
• It never fabricates to fill the silence. A low-confidence answer is always labeled as such. CoTeam will not present a guess as a sourced fact.
• One upload closes it everywhere. Add the missing document once, and every future question and questionnaire on that topic is answered from it going forward — you don't have to fix the same gap twice.

Where gaps surface

Gaps show up at both levels of the product:

• In Ask, inline on the answer card itself, with an "Add document" action right there.
• In a questionnaire, sorted to the top of your review queue automatically, since answers are ordered lowest-to-highest confidence — you see every gap before you see anything else.
• On the Knowledge Base page, as a running gap feed clustered by topic, so you can see patterns across many questions rather than one at a time.

The flywheel

Confidence scoring and gap detection are what make CoTeam's accuracy compound over time, rather than staying static:

• Edit an answer → saved as a knowledge base correction for that specific question type
• Add a document to fill a gap → that topic is answered for all future questionnaires
• Approve a high-confidence answer → reinforces that source in retrieval, making it more likely to surface again for similar questions
• An SME submits an answer → their expertise is captured and reused automatically

None of this requires a separate "training" step. The knowledge base you're already building by using the product is the same thing that improves it.