AI Settings: Answer Assist
Guide to setting up and configuring Answer Assist
Enable Answer Assist
Enable Answer Assist when you are ready for AI answers (or expert suggestions) to appear in selected boards. Answer Assist relies on the CLM knowledge base, so first ensure that the CLM ingestion is complete.
Review Mode
Review Mode: community-wide default toggle plus board and category overrides.Review Mode lets you run Answer Assist without posting anything to members, generated answers are held privately in the Answer Queue (and in an admin-only preview on the post) so you can evaluate answer quality before switching a board to Live Mode.
For full setup steps, behavior details, and best practices, see the Review Mode guide.
Target Boards (where Answer Assist answers)
Answer Assist: select Target Boards where AA should generate answers for initial posts.Target Boards determine where Answer Assist runs. Answer Assist watches for new posts in these boards and decides whether to respond.
|
Recommendation: Start with 1-2 target boards(can also be hidden boards that only authorized personnel can access). Expanding to many boards too early can create noisy results and make troubleshooting harder. |
Persona (how the AI appears)
Answer Assist Persona: display name, tone selection, and preview of how answers will read.Set a Display Name and Tone to match your community style. This will create a new profile for the Answer Assist, the name & tone of which will be reflected in the AI answers displayed to community members. Use the Preview to confirm the voice feels appropriate.
Answer Quality
Answer Assist Persona and Answer Quality settings.|
Setting |
What it controls |
Recommended for initial setup |
|
Answer Confidence Threshold |
How confident Answer Assist must be before posting an answer publicly. A three-level dropdown: Moderate (Recommended), High, or Extreme. Higher levels mean fewer answers but higher trust. |
Start with Moderate (Recommended). Move to High or Extreme for strict/official support communities where accuracy matters more than coverage. |
|
Answer Length |
Concise vs Medium vs Detailed answer output. |
Medium is recommended (clear but not overwhelming). |
|
Answer Delay (hours) |
How long Answer Assist waits before answering after a question is posted. |
Set to 0 to see the answers generate immediately; use 24 hours to simulate human-first workflows. |
|
Note: the Post Relevance Threshold is no longer an admin-facing setting — how closely a source must match the question is now handled automatically by the system. |
Citations and Expert Suggestions
Citations and Expert Suggestions: minimum citations, citation style, expert suggestion settings.Use citations to make answers verifiable, and expert suggestions to avoid low-confidence guessing.
|
Setting |
What it controls |
Recommended for setup |
|
Minimum Number of Citations |
Minimum citations required in each answer. |
Use 2 for an authoritative approach ; 1 is acceptable for lightweight answer gen. |
|
Citation Style |
Footnotes vs Inline links. |
Footnotes are usually easier to scan; use Inline links if your users prefer sources embedded in the text. |
|
Enable Expert Suggestions |
Shows expert recommendations when confidence is below your threshold. (Banned members are automatically excluded) |
On (recommended). |
|
Maximum Number of Experts to Suggest |
How many experts to list. |
3-5 is a good range for testing. |
|
Notify Suggested Experts |
Whether suggested experts are notified when they are recommended for a question. |
On (recommended), can be paired with the frequency caps in Expert Routing Notifications (below) so individual experts are not overloaded. |
In subsequent releases, there will be an option to opt out of Expert notifications entirely.
Expert Routing Notifications
Expert Routing Notifications add a load-balancing layer to expert routing: suggested experts can choose whether they receive routing notifications, and you can cap how often any one expert is contacted — so the same experts do not get pinged over and over.
Expert Routing Notifications in the Answer Assist settings.|
Setting |
What it controls |
Recommended for setup |
|---|---|---|
|
Enable Expert Routing Opt-Out |
When off, expert routing keeps working as it does today, every identified expert gets pinged. When on, the opt-out controls and frequency caps take effect, and members see a new Expert Routing section in their notification settings. |
Off by default. Turn it on once expert routing is actively in use, so experts can manage their own notifications. |
|
Daily routing cap per expert |
How many routing pings a single expert can receive in one day. After this cap, the routing event picks the next-best available expert. |
Can be set as per requirement |
|
Weekly routing cap per expert |
How many routing pings a single expert can receive in one week. Whichever cap (daily or weekly) is hit first stops the routing for that expert. |
Tune both caps to the size of your expert pool. |
Editing the Expert Routing Notifications settings.
Coming next: in an upcoming release, this expands into a multi-tier expert routing model — questions route first to auto-qualified community contributors, then escalate to an admin-designated organization team, and finally to community managers, with configurable escalation windows and full routing history for admins.
