Knowledge Intelligence Settings

The Knowledge Intelligence Settings page is the central configuration area for KI. It contains three tabs: Overview, Insights, and Saved Scopes. A persistent Create Job button appears in the top-right corner of all tabs.

Knowledge Intelligence honors existing Knowledge Base read and edit permissions and does not introduce new roles or permissions. The Switch to Settings button is visible only to Partition Administrators and Knowledge Managers, who can create jobs, manage content scopes, and add or edit knowledge insights. Authors do not see this button and cannot access the Settings page. All users can view the dashboard, but article visibility within findings is governed by existing Knowledge Base read permissions — authors see only the articles they have access to.

Overview Tab

The Overview tab provides a summary of Knowledge Intelligence activity across your knowledge base and quick access to jobs and run history.

Summary Metrics

Three metric cards appear at the top of the Overview tab:

Articles Analyzed: Total number of articles that have been processed by at least one KI job.
Completed Runs: Total number of job runs that have finished successfully.
Findings: Cumulative number of issues identified across all completed runs.

Jobs

The Jobs section shows a preview of your most recent jobs. Each job card displays:

A badge indicating the job type (Knowledge Insights or Duplicates and Conflicts).
The job name.
The number of articles and insights configured for the job.
A Run Job button, or a Job Running indicator if the job is currently active.
A toggle to enable or disable the repeat schedule, if the job is configured to repeat.
Select View All to open the full Jobs list, which shows all saved jobs and their actions.

Job Actions

On the full Jobs list, select the Options menu on any job to access the following actions:

Edit — modify the job configuration.
Duplicate — create a copy of the job with the same settings.
Delete — permanently remove the job.

Run History

The Run History section shows a preview of the most recent job runs. Select View All to open the full Run History table.

The Run History table includes the following columns:

Job Name: The name of the job that produced this run.
Job Type: Knowledge Insights or Duplicates and Conflicts.
Run Date: the date and time the run was started.
Status: Running (yellow indicator), Completed (green), or Failed.
Articles: Total number of articles analyzed in this run.
Findings: Number of issues identified.

You can filter the Run History by job, type, and status using the dropdown filters at the top of the table. Select a completed run to view its detailed results.

Run Detail Panel

Selecting a completed run opens a detail panel that shows:

Job type and completion status with timestamp.
Articles Analyzed, Total Findings, and Insights Run.
A View Job Details section that can be expanded to show scope, insights used, schedule, and analyzed scope.
A Findings Distribution bar showing the number of findings per insight type.
Individual findings listed by article title and article ID, with a snippet of the AI-generated response for each.

Insights Tab

The Insights tab lists all insights available for use in Knowledge Insights jobs. Insights are grouped by type and displayed as cards. Predefined insights provided by eGain cannot be edited.

Insight Types

Insights are organized into the following types:

Compliance — checks related to regulatory obligations, legal language, and procedural requirements.
Correctness — checks for factual accuracy, grammar, and content correctness.
Consistency — checks for content style adherence and brand voice alignment.
Accessibility — checks for readability, sentence complexity, and skimmability.

Insight Card Details

Each insight card displays:

A description of what the insight checks.
The insight type.
The Check Prompt — the prompt used to analyze articles for issues.
The Fix Prompt — the prompt used to suggest or apply a fix (if configured).
A Show on Dashboard toggle — when enabled, findings for this insight appear on the Knowledge Intelligence Dashboard. When disabled, findings appear only in job run reports.

Creating a Custom Insight

You can create your own insights using prompts from the Prompt Library.

To create a custom insight:

Click the + Custom Insight button.
In the Add Custom Insight dialog, complete the following fields:
- Insight Name (required): Enter a descriptive name for the insight.
- Insight Type (required): Select an existing type from the dropdown, or select + Create new type to define a new category.
- Check Prompt (required): Select the prompt that will analyze articles for issues. Selecting this field opens the Select Check Prompt dialog, which lists all prompts in the Prompt Library that are tagged for KI use. Please note that these prompts are created in the AI Console by the Prompt Manager and are tagged for Knowledge Intelligence Use either as a Check or Fix prompt.
- Fix Prompt (optional): Select a prompt to use for generating fix suggestions. When a fix prompt is configured, a Fix button appears in the AI Insights section of the article Properties panel, allowing authors to apply suggested fixes directly.
- Show on Dashboard: Toggle On to display findings from this insight on the Knowledge Intelligence Dashboard. Toggle off to limit findings to run reports only.
- Click Save to create the insight. The insight is then available for selection when configuring Knowledge Insights jobs.

Saved Scopes Tab

Saved Scopes define which articles are included in a Knowledge Intelligence job. A scope specifies a set of articles — based on folder selection or search criteria — that a job will analyze. Saved scopes can be reused across multiple jobs.

Saved Scopes tab with type filter open

Scopes List

The Saved Scopes list displays all scopes available in your department. The table includes:

Scope Name — the name assigned to the scope.
Description — an optional description of what the scope covers.
Language — the language of the articles in the scope.
Articles — the number of articles matched by the scope definition.
Type — Folder-based or Search-based.
Used By — indicates how many jobs are currently using this scope, or shows Not in use.
Created — the date the scope was created.
You can filter the list by scope type using the dropdown at the top of the list. Use the search field to find a scope by name.

Creating a New Scope

Select + New Scope to open the Create New Scope dialog. You can also create a new scope directly from the Select Scope dialog when configuring a job. The dialog provides two tabs for defining the scope:

Folder-Based Scope

A folder-based scope includes all articles in selected folders from your knowledge base hierarchy. To define a folder-based scope:

Select the Folder-Based tab.
Browse the folder tree and select the checkboxes next to the folders you want to include. The article count for each folder appears to the right.
If no folders are selected, the scope includes all folders.
Select a KB Language in which the scope is created. The dropdown displays all KB primary languages configured for the partition.
Select Save this scope for future use if you want to save it for reuse.
Enter a Scope Name (required).
Enter a Description (optional).
Click Save.

Search-Based Scope

A search-based scope includes articles that match filter conditions you define. To define a search-based scope:

Select the Search-Based tab.
Define one or more filter conditions. Each condition row consists of an attribute, an operator, and a value. Supported attributes include: Article Type, Article Name, Article Content, Create Date, Last Modified Date, and Article ID.
Use AND to require all conditions to be met, or OR to include articles matching any condition.
Select + Add Condition to add more rows.
Select a KB Language in which the scope is created. The dropdown displays all KB primary languages configured for the partition.
Select Save this scope for future use if you want to save it.
Enter a Scope Name and optional Description.
Select Save.

Similarity Scoring Tab

The Similarity Scoring tab lets you define how Knowledge Intelligence measures similarity between articles for Duplicates and Conflicts jobs. A similarity formula weights four components — title, client, topics, and question-answer (Q&A) pairs — and sets the thresholds used to flag two articles as duplicates or identical. You can use the built-in system default formula as-is, or create your own custom formulas and select which one applies when you run a job.

Similarity Components

Each formula scores article pairs using the following components. The weights assigned to each component must always add up to 100%.

Title — Compares the titles of the two articles. Dates or version numbers that are incidental to the subject (for example, "Maintenance Guide (April 2018)") are ignored during comparison, while dates or versions that are part of the subject itself (for example, "Annual Report 2024") are kept and compared.
Client — Compares the specific named client (company, firm, or organization) each article is written for or about. Generic or categorical terms, such as "SMB companies" or "eCommerce," are not treated as clients. If either article does not name a specific client, the client similarity for that pair is 0.
Topics — Compares structured details extracted from each article, such as product, module, region, policy, intent, training, and audience. The more of these details two articles share, the higher their topic similarity.
Question-Answer (Q&A) pairs — Compares the questions and answers extracted from each article. Articles with the same or similar questions and matching answers are treated as duplicate content. Articles with the same or similar questions but different answers are flagged as a conflict.

Scoring Fields

In addition to the four component weights, a formula includes the following fields:

Duplicate threshold — The similarity score, as a percentage, above which two articles are considered duplicates of each other.
Identical score — The similarity score assigned when two articles' content matches exactly.

System Default Formula

Every partition includes a System default formula, identified by the System and Default tags. It applies the following weights: Title 20%, Client 10%, Topics 10%, and Q&A 60%, with no minimum duplicate threshold. The system default formula cannot be edited or deleted, but you can create custom formulas alongside it.

Creating a Custom Formula

To create a custom similarity formula:

Select + Add Formula.
Enter a Formula name.
Enter a weight for each component — Title, Client, Topics, and Q&A — so that the four values total 100%. The total shown below the fields updates as you type and confirms when the weights are balanced. Select Distribute evenly to split the 100% equally across all four components.
Enter a Duplicate threshold — pairs scoring at or above this value are flagged as duplicates — and an Identical score — the value assigned to an exact match.
Optionally, enter an Additional extraction prompt to give the AI extra context for how this formula should score articles. This field has a 2,000-character limit.
Turn on Use as the default scoring formula if you want this formula to be used automatically for future Duplicates and Conflicts jobs. Only one formula can be set as the default at a time; turning this on for a formula turns it off for any other formula that was previously set as the default.
Select Save.

Access to the Similarity Scoring tab is subject to your assigned Knowledge Base permissions.

Setting the Default Formula

A Duplicates and Conflicts job always uses whichever formula is currently set as the default — there is no option to choose a formula at the time you run the job. If no custom formula is set as the default, the System default formula is used automatically.

To run a job with a different formula, come back to the Similarity Scoring tab before starting the job, open the formula you want to use, and turn on Use as the default scoring formula. This replaces the previously selected default. Formulas used by a completed job run are preserved with that run as a historical record, even if the default is later changed.