Intelligent Search: Semantic, Intent-Based Content Discovery for Learners

Description

Intelligent Search uses semantic and vector-based technologies to produce more relevant search results for learners browsing content in Cornerstone. Rather than matching the literal words in a query against title, keyword, and description fields, it interprets the meaning, intent, and context behind what a learner types, surfacing relevant content even when the exact wording is not used. A search for dog training, for example, can return semantically related results such as pet obedience guidance that a pure keyword search would miss.

The feature is available in Learning Search and Global Search Training for Learning Management customers, and in Global Search Content for single architecture customers. It is built on top of the existing Enhanced Search technology and uses a form of Hybrid Search, which combines semantic embeddings generated from titles, descriptions, and keywords with traditional keyword matching to preserve the unique identifiers customers have built their catalogs around. Intelligent Search uses vector and semantic technology with natural language processing; it does not use LLMs, and no customer data is used to train any models.

Business Need

Learners often do not know the exact words to search for to find the training they need. Traditional keyword search matches terms only within title, keyword, and description fields, which can return unrelated content while omitting genuinely relevant material — leading to dead ends and slower discovery.

By moving beyond keyword matching to understand intent and context, Intelligent Search reduces these dead ends and speeds up the moment a learner finds the right content. This creates a more intuitive and efficient learning experience and is particularly valuable for organizations that encourage self-directed learning and content discovery over a purely compliance-driven approach.

Key Features

• Intelligent Search applies semantic, vector-based technology to interpret user intent and context, surfacing relevant content across Learning Search, Global Search Training, and Galaxy Global Search Content even when descriptions vary.

• Administrators govern the feature through three preferences — AI Activation Preferences, Cornerstone Search Preferences, and Galaxy Global Search Preferences — and both the activation and search preferences must be enabled for learners to use it.

• End users can switch between Intelligent Search and Keyword Search directly on the page, via a menu icon in Learning Search or a checkbox in the Refined Search area of Global Search Training.

• Hybrid Search combines the benefits of semantic embeddings with traditional keyword search, balancing intent interpretation against the unique identifiers customers have used to build their catalogs.

• Intelligent Search supports 30 languages, covering 99.7% of all searches in the application, and automatically falls back to standard keyword search for any unsupported language.

• The maximum number of results increases from 10,000 with Keyword Search to 100,000 with Intelligent Search, reflecting a focus on a limited set of high-quality, relevant results rather than high volume.

Deployment & Considerations

Intelligent Search reaches general availability with the July 2021 release and is available to customers using Cornerstone and Xe Single Architecture. The feature is available with the start of UAT and with the production release. In stage portals, AI Activation Preferences are on by default and Search Preferences are off by default; in production, everything is off by default. No new permissions are introduced.

To enable the feature, administrators navigate to Admin > Tools > Core Functions > Core Preferences > AI Activation Preferences and enable Intelligent Search, then enable it in Search Preferences or Galaxy Search Preferences for the relevant OU; the search preference is not available until the AI Activation Preference is enabled. Both preferences are available to Compliance Managers and Learning Administrators. On first enablement, a data vectorization delay applies — roughly 3–5 days for small portals and 5–10 days for large portals — occurring in both stage and production; this runs once per portal but may recur if the feature is disabled across all OUs and later re-enabled. Intelligent Search is not a keyword-matching search, so exact title matches are not guaranteed to appear at the top, and its results will not match predictive search results. Existing search customizations and custom scoring logic are retained but not applied while Intelligent Search is active, and custom URLs with custom queries (beyond filters) may return different results. The feature is not currently planned for administrator use cases such as Content Manager. Filters, availability, result display, and localized language support continue to work as expected, and there are no plans to deprecate keyword search.

FAQs:

What is Intelligent Search?

Intelligent Search uses semantic and vector-based technology to understand the meaning and context behind a user's search query, providing more relevant search results than traditional keyword matching. It helps learners find the right content even when they do not use the exact words contained in course titles, descriptions, or keywords.

Where is Intelligent Search available?

Intelligent Search is available in Learning Search and Global Search Training for Learning Management customers. For Single Architecture customers, it is available in Global Search Content.

Does Intelligent Search replace Keyword Search?

No. Cornerstone has no plans to deprecate Keyword Search. Both Intelligent Search and Keyword Search will continue to be available, allowing organizations to choose the approach that best fits their needs.

Does Intelligent Search use AI?

Yes. Intelligent Search uses semantic and vector-based technology along with natural language processing (NLP) to understand search intent. However, it does not use Large Language Models (LLMs).

Is customer data used to train AI models?

No. Customer data is not used to train any models. Intelligent Search only uses content metadata such as titles, keywords, and descriptions to generate search results.

Why doesn't an exact title match always appear first?

Intelligent Search prioritizes relevance and intent rather than exact keyword matching. As a result, an exact title match may not always appear at the top of the results list if other content is determined to be more relevant to the user's query.

Why did I receive search results that do not contain my exact keywords?

Intelligent Search evaluates meaning and context rather than simply matching keywords. Results may contain semantically related content that addresses the same topic, even if the exact search terms are not present.

Will filters, availability settings, and localization still work?

Yes. Existing features such as filters, availability restrictions, search result displays, and support for localized languages continue to work as expected with Intelligent Search.

Do Intelligent Search results match Predictive Search suggestions?

No. Predictive Search is designed to suggest exact matches while a user types. Intelligent Search focuses on returning the most relevant content based on meaning and intent. Therefore, results may differ between the two features.

Can Intelligent Search be used for administrator searches?

Currently, Cornerstone does not plan to extend Intelligent Search to administrator use cases such as Content Manager. Administrative users often search for specific IDs, exact titles, or unique identifiers where keyword matching is more appropriate.

What permissions are required to enable Intelligent Search?

Users who need to enable or disable Intelligent Search must have access to AI Activation Preferences. If access is not available, users should contact their system administrator to obtain the required permissions.

What data is used to generate search results?

Intelligent Search uses content metadata including titles, keywords, and descriptions. User data is not used to generate search results.

Can Intelligent Search be tested before enabling it for all users?

Yes. Organizations can test Intelligent Search in a lower environment such as Stage or Pilot before enabling it in Production. It can also be enabled for specific Organizational Units (OUs), allowing a controlled rollout to a limited audience before broader adoption.

How long does it take for Intelligent Search to become available after activation?

When enabled for the first time, Intelligent Search requires time to build and vectorize portal data. Small portals typically require 3–5 days, while larger portals may require 5–10 days before the feature becomes available.

How many languages are supported?

Intelligent Search currently supports 30 languages, representing approximately 99.7% of searches performed in the application. If an unsupported language is used, the search automatically falls back to traditional keyword search behavior.

What is the maximum number of search results?

Keyword Search can return up to 10,000 results. Intelligent Search can return up to 100,000 results, focusing on delivering the most relevant results rather than simply the highest volume.

What is Hybrid Search?

Hybrid Search combines semantic vector-based search with traditional keyword search techniques. This approach allows Intelligent Search to understand user intent while still benefiting from existing catalog metadata and unique identifiers.

Which organizations benefit most from Intelligent Search?

Organizations that promote self-directed learning, content discovery, and exploration are likely to benefit most from Intelligent Search. Organizations that rely heavily on exact keyword matching, custom search logic, or highly structured search guidance may need to evaluate the impact before adoption.