Knowledge Review and Reuse

Updated 9 September 2026

99 Venture owns knowledge management. Tenant staff see approved answers in Teams and can suggest improvements through case nomination or feedback. They cannot create, revise, approve or retire library answers through the support interface. Knowledge approval does not establish consent or authorise model training.

Use an Approved Answer in Teams

1. Open a ticket assigned to you, then open your name menu → Knowledge Library.
2. Find and read an approved answer and its review date.
3. Choose Insert into Reply Draft. Existing draft text is retained.
4. Check the wording and customer details, then explicitly choose the channel's Send button.

Insertion does not send. One linked approved answer can be inserted per draft; send or clear it before inserting another. Reply windows and size limits apply. Source messages and publication history stay in Administration; tenant support staff receive the approved answer, not the underlying management evidence.

Approval, source permission, revision and expiry are checked when a linked reply is queued and before provider dispatch. Expired, revised, retired or withdrawn knowledge blocks its reuse. A dispatched message cannot be recalled. Staff may compose ordinary independent responses, but should not remove a blocked reference just to bypass a review decision.

Suggest an Improvement

Use Case Context → Nominate as a Knowledge Candidate, or Add Feedback with the Knowledge category. Explain what could be reused or corrected. The nomination is a reminder for review; it does not generate or publish an answer automatically.

Manage Discovery Knowledge

1. In 99 Venture Administration, open Studio → Discovery.
2. Review a candidate against its supporting evidence. Remove identifiers, unsupported claims and promises limited to one case.
3. Record Review Reason, then Approve Knowledge or Reject Candidate.
4. In Agents → Build, select the approved knowledge and save and publish the agent revision before using it.

Discovery knowledge belongs to the separate Studio collection. An approved candidate is available for agent configuration; it is not automatically inserted into the older support Knowledge Library. See the Studio guide for source permissions, process approval, publication and rehearsal steps.

Manage Existing Support Answers

1. Open Studio → Knowledge → Support Knowledge Library.
2. Choose Manage Answer, then inspect the source and answer.
3. Enter any revised wording, a Review Date and Decision Reason.
4. Choose Save Revision, Approve, Reject or Retire.

Revision returns an answer to review. Approval requires an available source and a future review date within the permitted range; use the displayed UTC date. Retirement blocks future linked reuse and keeps history. Stale changes are rejected; refresh and review the current item before making another decision.

Management requires 99 Venture platform-administrator access. An old support reviewer role does not grant the management API. The single-user pilot's explicitly configured self-review exception remains a pilot exception; independent production approval needs separate people and agreed policy.

Withdrawal and Records

Studio source permission changes invalidate affected candidate/process versions and agent publications. Re-allowing evidence requires fresh review and publication; old approvals do not become current automatically. Pause agent runs while a broader issue is investigated. These actions do not train a model, erase historical evidence or recall replies already sent.

Support answer revisions, approval reasons and linked reuse remain recorded in the support database. Studio source snapshots, publications and runs are in its separate database and evidence export. Apply retention to both stores and their exports; a support-case deletion does not automatically delete a Studio source pack.

Keep each deployed workspace isolated. The support library follows the support pool's visibility; do not assume document-level or department-level partitions that have not been configured and tested. Production policy and real-user acceptance remain separate from the fictional demonstration.