Configuration Manager - Channels

The "Channels" menu is used to define settings for channel content display and AI-based comment moderation.

Setting

What does the setting do?

Number of days for new flag

Defines how many days a channel content item is displayed with a "new" badge after publication. Set to 0 to show the flag indefinitely.

Show publication/update date on the channel content card

When enabled, the publication or last update date is shown on the content card within the channel. If the content has been updated in the respective channel, the date of the most recent update is displayed instead of the original publication date.

Enable AI Review Moderation

When enabled, comments are automatically checked for netiquette violations using AI. Flagged comments are highlighted in the Review Manager for administrators to review and moderate. Enabling this setting reveals two additional fields: Prompt and Guardrails.

Prompt

The instruction text sent to the AI to define how comments are assessed. The AI uses this prompt to decide whether a comment is acceptable (true) or violates netiquette (false). A default prompt is pre-configured. The placeholder {{ADDITIONAL_GUARDRAILS}} in the prompt is automatically replaced with the content of the Guardrails field at runtime.

Guardrails

Additional terms or phrases the AI should consider during moderation, for example platform-specific rules or forbidden words. Multiple entries can be provided as a comma-separated list and may include terms in different languages.

Blocked Words List

A list of words or phrases that are flagged automatically, without AI involvement.

New Flag

This setting determines the duration for which the "New" flag remains visible on content. The flag is removed earlier if a user interacts with the content before the specified time expires. If the content is not accessed within the defined period, the flag disappears automatically once the time elapses. The time can also be configured so the flag remains indefinitely unless the user takes action.

Show Publication Date on Channel Content Cards

When this setting is enabled, the publication date of a content item is shown on its card within the channel.

If the content has been updated in the respective channel after its original publication, the date of the most recent update is shown instead. This allows users to immediately see whether they are looking at new or recently revised content.

Enable AI Moderation

When this setting is enabled, every comment posted in a channel is automatically assessed by an AI model before it becomes visible to moderators. The AI checks whether the comment violates netiquette rules — for example, whether it contains profanity, insults, harassment, or discriminatory language.

Comments flagged by the AI are highlighted in the Review Manager, where administrators can either hide the comment or confirm it as acceptable using the Keep comment action. Comments that pass the check are not flagged.

The AI does not block or delete comments automatically. It only assists administrators in identifying potentially problematic content more efficiently.

Prerequisite: This feature requires a specific deployment configuration. If it has not been set up for the environment, enabling this setting and saving will result in the following error message:

Input error! Enable AI Review Moderation: AI Review Moderation is not available in this environment. The "Enable AI Review Moderation" option must be disabled before saving. To request the required deployment, contact Scheer IMC via the Service Desk.

Data privacy: Comment moderation is processed by a self-hosted AI model running within own infrastructure. No comment content is transmitted to external services or third parties.


Prompt

The prompt defines the instructions sent to the AI each time a comment is submitted for moderation. It determines how the AI evaluates comments and what it considers a violation.

The AI is expected to return only true (comment is acceptable) or false (comment violates netiquette) — no explanations or additional text.

A default prompt is pre-configured and works out of the box for most use cases. It instructs the AI to flag content containing profanity, insults, mockery, harassment, hate speech, or threats, regardless of the language the comment is written in.

The prompt contains the placeholder {{ADDITIONAL_GUARDRAILS}}, which is automatically replaced at runtime with the content of the Guardrails field. This allows to extend the moderation rules without modifying the base prompt.

The prompt supports up to 4,000 characters.

Warning: The prompt must instruct the AI to return only true or false as its response. If the prompt is modified in a way that causes the AI to return a different format or additional text, the moderation feature will stop working. In addition, the {{ADDITIONAL_GUARDRAILS}} placeholder must not be altered in any way — any change to its exact spelling or syntax will prevent the Guardrails content from being applied. It is recommended to use the Guardrails field for customisations rather than modifying the base prompt directly.

Guardrails

The Guardrails field allows to define context-based rules that the AI should take into account when assessing comments — on top of the instructions already defined in the Prompt.

Unlike the Blocked words, which flags comments automatically based on exact word matches, Guardrails are evaluated by the AI with context. This makes them suitable for cases where a simple word match is not sufficient — for example, when a term is generally acceptable but problematic in certain contexts, or when platform-specific policies apply that require interpretation rather than a binary check.

Typical use cases include rules such as:

  • Competitor names or product references should not appear in comments

  • Comments should not contain advertising or external links

  • Certain topics are off-limits for discussion on this platform

Multiple entries can be provided as a comma-separated list. Terms in different languages are supported.

The content of this field is automatically inserted into the Prompt at the position of the {{ADDITIONAL_GUARDRAILS}} placeholder.

The field supports up to 4,000 characters. It can be left empty if no additional rules are needed.v

Blocked Words List

The Blocked Words field allows to define words or phrases that are flagged automatically when found in a comment — without the comment being sent to the AI for assessment. A comment containing any of the listed terms is flagged immediately, regardless of context, and the AI assessment is skipped entirely for that comment. If no match is found, the comment proceeds to AI assessment as normal.

This makes the Blocked Words field the appropriate tool for terms that are always unacceptable, such as profanity, slurs, or other explicit language. Because the check happens before AI processing, it acts as a fast and reliable safety net for clear-cut cases.

Entries are provided as a comma-separated list, for example: word1, word2, word3. Terms in different languages are supported.

Blocked words list vs. Guardrails


Blocked words list

Guardrails

How it works

Exact match — no context

AI assessment — with context

Best for

Terms that are always unacceptable (e.g. profanity, slurs)

Platform-specific policies, context-dependent rules

AI involvement

None — flagged before AI check

Yes — evaluated by the AI