The AI-based Wiki serves as a central knowledge base for various content such as FAQs, work instructions, documents and news.Content in the AI-Wiki is not solely for documentation. It can also contribute significantly to the quality of suggested responses, phone calls, and chat contexts.
What is a Knowledge Source?
A knowledge source is a single exploitable knowledge entry in the AI-Wiki.
It can originate in two ways:
- as manually created articles
- as content from an externally linked source, for example, via a website connector
Manually created articles are suitable for content that should be consciously formulated and technically reviewed. These include work instructions, frequent questions, process descriptions, internal requirements, or mandatory formulation.
Externally linked sources are suitable for regularly using the existing content. This could include help centers, documentation pages, or public information pages.
Why Structure is Important
Knowledge in the AI-Wiki is not just stored, but also categorically arranged. For this, groups and classifications are used.
Groups form the thematic structure. They help make content findable and bundle technically related information.
Classifications describe the kind of content. This distinction is essential because different content is used differently:
- An FAQ answers a specific recurring question.
- A work instruction describes an operational process.
- A document contains background information or mandatory regulations.
- News inform about current changes and appear on the homepage.
If these types of content are cleanly separated, the knowledge stock remains more manageable for review. Simultaneously, AI functions can categorize the content better.
Website Connector
The Website Connector links an external website with the AI-Wiki.
Path in Enneo: Work aids → AI-Wiki → Sources
During crawling, Enneo reads out the connected website and takes over suitable content as knowledge entries. These then appear in the knowledge structure.
A connector is controlled through multiple settings:
- Source URL
- Include paths
- Exclude paths
- Maximum page number
- Frequency for renewed crawling
The source URL defines which website is connected.
Include paths establish which areas of the website should be considered. Exclude paths exclude areas that are not relevant. These include, for example, login pages, privacy pages, imprint, blog areas, or technical overview pages.
The maximum page number limits the scope of the transferred content. The crawling frequency determines how regularly Enneo reads the source again.
The connector makes content technically available. However, it does not evaluate whether these contents are technically suitable. This review remains part of editorial responsibility.
Crawling and Currency
During crawling, content from the external source is read out and made available in Enneo. The connector then shows a status, for example, ongoing, completed, or faulty.
A renewed crawling updates the transferred content based on the current connector configuration. As a result, changes to the external website can be included in Enneo.
Simultaneously, the quality of the result depends on the external source. If URL structures, navigation, or website content changes, so can the knowledge structure in Enneo.
For stable results, connectors should be configured as precisely as possible. A narrowly defined source is easier to review and usually provides better results than a very broadly connected website.
Files as a Knowledge Source
In addition to manually maintained articles and linked websites, files can also be used as knowledge sources in the AI-Wiki.
Path in Enneo: Work aids → AI-Wiki → Sources
The Files Connector is used to centrally provide technically relevant documents and make them useful for AI-supported functions. Uploaded files are processed, converted into knowledge entries, and can then be considered in suggested responses, chat contexts, or AI agents, provided they are released for this purpose.
Files should not be used as unchecked storage. The decisive factor is that the content, file name, and storage location are technically precise. A clear folder structure helps manage larger knowledge stock comprehensibly and review it purposefully later on.
If the content of a file changes, the knowledge source must be processed again so that the updated content becomes effective in the system. Re-indexing ensures that not only the file itself but also the search and AI contexts derived from it are up to date.
As with website connectors, the connector technically provides content. The technical responsibility for accuracy, currency, release, and structure remains with the responsible users.
Files and Media in Articles
Articles can be supplemented by uploaded media. These include, for example, images or videos.
Media is useful when it makes a technical statement more comprehensible. This applies, for example, to form examples, process representations, or screenshots of certain states.
It’s important: Media should supplement the text, not replace it.
With the KB Files Connector, files can be used directly as a knowledge source in Enneo. Documents can be uploaded via drag-and-drop, structured in folders, managed, and if necessary, previewed for checking. After uploading, the content is automatically processed and made findable for the AI, so that it can be considered in suggested responses or chat, for example. Supported files are validated, saved, and converted into a format suitable for knowledge search. Individual files can also be re-indexed if content has changed or processing needs to be updated.
For AI-supported functions, the written content remains particularly relevant. Text can be searched uniquely, processed, and taken over into response contexts. Therefore, critical information should always be present in the article text as well, and not only contained within a screenshot or video.
Visibility for AI Functions
The setting Make knowledge source publicly accessible controls whether a knowledge source can be used for AI-supported functions.
If a knowledge source is released, it can be considered, for example, in suggested responses, chat contexts, or AI agents.
If it is not released, it primarily remains a part of the internal documentation in the AI-Wiki.
This setting is technically relevant. A released knowledge source can influence the results of AI functions. Therefore, only content that is reviewed, current, and formulated clearly should be released.
Content that is confidential, outdated, incomplete, or technically unclear should not be released.
Impact on AI Agents
AI agents can use knowledge sources as context if they are accessible and technically relevant.
The quality of agent results strongly depends on the quality of the knowledge stock. Ambiguous formulations, contradictory articles, or unreviewed connector content can lead to inaccurate results.
The AI-Wiki, therefore, is more than just a storage for information. It controls which knowledge is available to the AI and how this knowledge is technically classified.A well-maintained AI wiki improves the reproducibility and stability of AI-supported processing. On the contrary, an unclear knowledge base can lead to inconsistent responses and misinterpretations.
Maintenance and Responsibility
A good knowledge base is not necessarily large, but reliable.
The following points should be checked particularly for manual articles:
- Is the content technically correct?
- Is the article unambiguously formulated?
- Is the classification appropriate?
- Is the group selected sensibly?
- May the content be used for AI functions?
For connectors, additional questions are added:
- Is the external source technically reliable?
- Are the include and exclude paths set sensibly?
- Is the maximum number of pages chosen appropriately?
- Is the crawling frequency appropriate?
Changes to the AI wiki do not only affect the documentation. They can also influence the quality of suggested responses, chat contexts, and AI agents.
Therefore, articles, connectors, and visibility settings should be maintained as technical system components.
