Learning suggestions can come from real conversations, agent answers, teaching through Yuna, or explicit AI corrections. A suggestion is only a proposal. It does not affect AI customer service until you adopt it.

When you give Yuna a durable rule, Yuna first asks whether it should continue organizing the instruction. Confirming only authorizes a reviewable proposal; it does not publish or activate the rule. Correct AI also carries the customer question and original answer, but both entry points use the same learning-review flow.

A proactive-outreach suggestion must first include its audience, trigger, channel, content, and frequency limits, with a complete preview shown before confirmation. If no safe proposal can be produced, Yuna says that nothing was created; the result is not treated as success and nothing is enabled or sent.

What a suggestion should contain#

A reviewable suggestion normally includes the customer question, proposed answer, learning source, scope, risk level, and the result created after adoption. Proactive outreach and suggestions that need live business information should also explain the trigger, action, and safeguards.

Make a decision#

• Save to learning: Confirm that the content is correct and create or update the related content.
• View suggestion details: Review the source, scope, and risk before deciding.
• Ignore: Keep the suggestion inactive and remove it from To confirm.

If saving fails, the suggestion remains in To confirm with an explicit failed state and Retry save action. It does not disappear when processing stops.

Ignored suggestions remain available in their status view. YundaDesk should deduplicate identical or highly similar suggestions so that the same card does not immediately return after being ignored.

Review high-risk content carefully#

Price commitments, refund policies, health or safety guidance, sensitive conclusions shared across customers, and rules that send proactive messages require careful scope review. Do not adopt an incomplete suggestion simply to clear the queue.

After confirmation#

Different suggestions become knowledge, AI skills, customer memory, or proactive outreach rules. Confirmation does not always mean immediate use. An AI skill remains a draft until you review its customer scenarios, complete simulation, and publish it. High-risk and proactive actions may require stricter confirmation.

The After adoption section explains what to do next. Knowledge becomes available to AI after it is saved and prepared for search. AI skills must be simulated and published. Temporary skills expire on schedule. Customer memory applies only to that customer. Proactive outreach still requires testing and approval.

Open Workspace → Improvement suggestions to search customer questions or proposed answers, then filter by To confirm, Adopted, or Ignored. Lists show the latest update time and are ordered newest first. Adopted suggestions link to the related content. Skill testing and publication still happen from the skill detail.