Learn about the Unstructured extraction and Unstructured search workflows in Unstructured augmentation.

Unstructured augmentation helps you turn unstructured documents into usable Reltio data or searchable knowledge bases through the following workflows:

• One time upload: Upload a single file, extract its data, and map it to your tenant in one pass.

• Automated bulk extraction: Create a reusable extraction template, connect a document source, and set up a pipeline to process multiple files manually or on a schedule.

• Unstructured search: Test retrieval against a sample document, create a searchable knowledge base from a configured source, and make it available to custom agents.

For One time upload and Automated bulk extraction, the extracted data becomes available in Reltio for review, matching, and downstream use.

For Unstructured search, documents from the configured source are processed into a search index. The resulting knowledge base becomes available for natural-language questions through custom agents created in Custom agents.

Unstructured extraction workflows

Unstructured extraction supports two workflows: One-time upload, which processes a single document, and Automated bulk extraction, which processes multiple documents from a configured source manually or on a schedule.

The following diagram describes the workflow for One time upload, which processes a single document.

The following table explains each stage of the One time upload workflow:

Table 1. One time upload workflow

Stage	Task	Description
1	Document upload	Upload a single document containing the information that you want to extract.
2	Data extraction	Unstructured augmentation processes the document and extracts structured information from its content.
3	Data mapping	Unstructured augmentation maps the extracted information to the corresponding Reltio entities and attributes in the selected tenant..
4	Output review	Review the extracted and mapped data before it becomes available for matching and downstream use in Reltio.

The following diagram describes the workflow for Automated Bulk Extraction, which processes multiple documents from a configured source manually or on a schedule.

The following table explains each stage of the workflow:

Table 2. Automated Bulk Extraction workflow

Stage	Task	Description
1	Template creation	The workflow begins with template creation. A template defines what structured data Unstructured augmentation must extract and how the extracted output maps to Reltio entities and attributes in your data model. You can create templates for specific document types and reuse them across processing runs. For example, create a template for contracts and use it to process a set of contracts that share a similar structure, which supports more consistent and efficient extraction.
2	Extraction and validation	As part of template creation, Unstructured augmentation processes a sample document and generates extracted output for review. You then review the document, examine the extracted JSON, and verify how the extracted fields map to Reltio entities and attributes. This validation stage ensures that the extraction output is complete, accurate, and suitable for repeated use.
3	Publish template	After reviewing the extraction results, publish the template. Publishing makes the template available for use in pipelines, enabling consistent and reliable processing across future document runs of similar type and structure.
4	Source configuration	Sources define where Unstructured augmentation retrieves files for processing. You can configure sources for supported cloud storage services, such as AWS S3 and Google Cloud Storage. Unstructured augmentation document pipelines use these source definitions to access files for processing.
5	Pipeline configuration	A pipeline is configured to use a selected source and a published template to process documents. It allows you to define the following settings: Crosswalk source, which defaults to the cloud source Whether to create a data change request before publishing extracted data to Reltio, if needed The schedule for pipeline runs
6	Pipeline execution	When a pipeline is executed, Unstructured augmentation begins processing documents from the selected source. The system creates a new execution record per run. Each execution captures run-level details, documents processed, including start time, end time, status, and processing outcomes.
7	Execution review	After a pipeline runs, you can review execution results at both the pipeline level and the document level. Execution views show overall run status, success and failure counts, and the extracted entities associated with processed documents. This helps you confirm outcomes and identify runs that need follow-up.
8	Publishing structured output to Reltio	As part of this workflow, Unstructured augmentation processes extracted and approved data for use in Reltio. Depending on the configured mappings and processing outcome, the results will contribute to entity creation, linkage, and downstream master data workflows.

Unstructured search workflow

Unstructured search processes documents from a configured source and makes their content searchable through a knowledge base. The following diagram describes the Unstructured search workflow.

The following table explains each stage of the Unstructured search workflow:

Table 3. Unstructured search workflow

Stage	Task	Description
1	Source configuration	Configure AWS S3, Google Cloud Storage, or Google Drive as the source containing the documents that you want to make searchable.
2	Sample document upload	Upload a representative document from the source content. Unstructured search uses the sample to prepare the test playground and evaluate the retrieval configuration.
3	Retrieval configuration	Use the recommended configuration or customize the search method and number of document chunks returned for each question.
4	Retrieval testing	Ask representative questions in the test playground. Review the generated answers and retrieved chunks to confirm that the configuration returns relevant content.
5	Knowledge base configuration	Apply the tested configuration to the complete document source. Select the source, enter the knowledge base name and description, and choose a Manual, Daily, or Weekly schedule.
6	Index creation	Create the knowledge base index. Unstructured search processes the documents in the selected source and makes their content searchable.
7	Index monitoring	Review the document count, last run, execution frequency, indexing dates, and indexing statuses to confirm that the documents are ready for retrieval.
8	Custom agents integration	In Custom agents, enable the Knowledge bases capability and select one or more knowledge bases that the custom agent can use.