Discovery Agent discovery-agent

The Discovery Agent delivers AEM content on demand through natural, conversational prompts for a streamlined, click-free discovery experience. It intelligently searches across Assets, Content Fragments, and Adaptive Forms to deliver relevant materials such as images, videos, PDF documents, articles, and form templates. Using natural language, you can search for content without building complex queries or applying filters in the AEM Assets interface. Based on your prompt, the agent returns curated results along with asset metadata and delivery URLs, ready to be embedded in other applications.

Some of the key benefits of Discovery Agent include:

Skills skills-discovery-agent

The Discovery Agent provides the following skills:

Personas personas-discovery-agent

Campaign Managers campaign-managers

Discovery Agent enables Campaign Managers to quickly identify and reuse trusted, high-performing content for ideation.

Channel Marketers channel-marketers

Discovery Agent allows Channel Marketers to efficiently find relevant assets to create cohesive, multi-channel experiences.

DAM Librarians dam-librarians

The DAM Librarians can flag assets that are missing the metadata standards set by the organization, supporting consistent governance and ensuring assets remain complete and ready for use across channels.

Agencies and Partners agencies-partners

Agencies and Partners can easily find brand-approved assets within Content Hub and reuse them to accelerate creative work while staying aligned with brand standards.

How to access Discovery Agent? access-discovery-agent

You can access AEM Business Agents via AI Assistant. Log on to experience.adobe.com and you can start interacting with AI Assistant by specifying your prompt in natural language using the search box:

For information on the MCP endpoint to access Discovery Agent, contact Adobe Support.

Common use cases and sample prompts use-cases-prompts

Assets discovery-agent-use-cases-assets

Tag-based asset discovery

The Discovery Agent uses natural language prompts to find assets associated with specific tags in the AEM repository, helping users quickly access content organized according to their organization’s taxonomy.

Sample prompt:

Show images tagged office in folder WKND.

Folder-based content discovery:
The Discovery Agent can identify assets by interpreting natural language prompts that reference folder names in AEM. Users can simply mention the folder in their prompt, without manually navigating through the repository, significantly reducing the number of clicks needed to locate the right content.

Sample prompts:

Resolution and format-based asset discovery

The Discovery Agent can identify assets that meet specific quality requirements, such as file format or minimum resolution, allowing users to quickly locate product visuals that are ready for high-quality delivery and reuse across channels.

Sample prompt:

Find product packaging PNG images at least 2000 px wide.

Orientation-based content discovery

The Discovery Agent can filter assets by recognizing visual attribute, such as the presence of people and the orientation of an image. This allows users to quickly narrow down content to the most relevant visuals without manually applying multiple filters in AEM.

Sample prompt:

Show assets with person in landscape orientation.

Content Fragments discovery-agent-use-cases-content-fragments

The Discovery Agent helps users quickly locate the right Content Fragments by interpreting natural language references to campaign names, product brands, publication status, and recent creation activity. It allows teams to surface campaign-ready fragments and view brand-specific content, all without manually browsing through folders or applying multiple filters in AEM.

Sample prompts:

Forms discovery-agent-use-cases-forms

The Discovery Agent helps you quickly find adaptive forms using natural language prompts. It searches through form content and metadata to find matches based on keywords from your prompts. This means you can successfully discover relevant forms even if your search terms are not in the form’s title or description.

Sample prompts:

Note: Form discovery currently supports Edge Delivery Services forms only and tag-based search is not available for forms at this time.

Search Results discovery-agent-search-results

Assets discovery-agent-search-results-assets

The Discovery Agent returns the top results for each query, sorted by relevance to ensure that the exact matches appear first. The Agent combines metadata-driven queries with semantic search to assemble a focused set of likely matches, then uses an LLM to rank them based on user intent. This blended approach delivers accurate, context-aware results without depending entirely on a direct keyword match.

Each result includes asset name along with key asset metadata such as the asset path, creator, creation date, title, description, format, last modifier, last modified date, file size, dimensions, Dynamic Media URL, and associated tags. If an asset is in approved state, the results also include Dynamic Media with OpenAPI URL.

You can click the asset path to seamlessly navigate to the asset location within AEM.

You can use these asset details to quickly evaluate if an asset meets the requirements without navigating to each asset to view these details.

Content Fragments discovery-agent-search-results-content-fragments

The Discovery Agent provides full-text search capabilities for Content Fragments, returning the top results that best match the specified prompt. Each result includes Content Fragment name along with key metadata fields such as Content Fragment path, creator, creation date, variations, last modifier, and last modified date fields.

You can click the Content Fragment path to seamlessly navigate to the Content Fragment location within AEM.

Prompting best practices prompting-best-practices-discovery-agent

Specify concise details in your natural language prompts so that the agent can return accurate and relevant results. The more clearly you describe what you are looking for, the better the agent can refine and narrow the output. For example, you can: