RDM Autopilot

Learn how to configure RDM Autopilot for automated and assisted reference data mapping.

RDM Autopilot (Intelligent Mapping) is an optional automation capability in Reltio Reference Data Management (RDM) that helps organizations automatically resolve unmapped source values during transcode operations. When a source value does not have a mapping to a canonical lookup value, Autopilot analyzes the value using semantic similarity, fuzzy matching, AI reasoning, and frequency thresholds. Based on this evaluation, Autopilot automatically maps values, generates suggestions for steward review, detect misspellings, or recommends creating new lookup entries.

Note: This capability is not enabled by default. Because Autopilot uses AI-based evaluation, it works only when you explicitly enable the Autopilot feature in the configuration settings.

You maintain full control over whether the feature is used. If the feature is not turned on, the system continues to behave as usual, and unmapped values must be handled manually..

Note: Autopilot does not work for MDM tenants configured with the "All in memory" RDM cache. In such tenants, Autopilot will not act on transcode errors originating from MDM. However, Autopilot can still be used independently on the RDM tenant through the Transcode API.

Once enabled, Autopilot monitors transcode failures and evaluates unmapped values in the background. The processing is asynchronous, which means it does not impact transcode API performance or latency. This allows organizations to introduce automation safely while maintaining governance through configurable thresholds, workflow approvals, and monitoring tools.

Based on its configuration, Autopilot supports the following capabilities:


Capability	Description
Automatic Mapping	Automatically creates mappings when confidence thresholds are met
Suggested Mapping	Generates a Data Change Request (DCR) for human review
Misspelling Detection	Identifies likely misspellings and caches the resolution
Lookup Creation	Automatically creates new lookup entries
Suggested lookup	Recommends creating a new lookup entry and generates a Data Change Request (DCR) for steward review when the value belongs to the lookup type but does not match an existing entry.

Why would I use RDM Autopilot

Organizations often receive reference data from multiple source systems. These systems may send values with typos, abbreviations, naming variations, or new entries that do not yet exist in the lookup table. When this happens, the transcode operation fails and data stewards must manually review and map the value.

As data volume increases, these unmapped values can accumulate quickly and create delays in data processing. Manual resolution also increases the risk of inconsistent mappings.

RDM Autopilot helps reduce this effort by automatically analyzing unmapped values, identifying likely matches, and either creating mappings or generating suggestions for steward review.


Common Issue	Example
Typos	Untied States → United States
Naming Variations	U.S.A → United States
Different Conventions	UK → United Kingdom
New Values from Source Systems	New region, currency, or product code

Autopilot Capabilities

Autopilot combines multiple technologies to intelligently resolve mappings.


Capability	Description
Semantic Matching	Uses semantic score to understand meaning and find similar canonical values
Fuzzy Matching	Calculates textual similarity between source values and lookup values
AI Decision Support	Uses AI models to evaluate ambiguous matches and provide reasoning
Frequency Validation	Requires a value to meet the configured frequency threshold before automation occurs, including cases where the threshold is set to 1
Workflow Integration	Generates DCR tasks for steward review when needed

Autopilot Modes

RDM Autopilot supports two operating modes that control how the system handles unmapped values. These modes allow you to evaluate the feature safely before enabling full automation. Youcan choose the appropriate mode based on their governance requirements and confidence in the configured thresholds.

Preview Mode

In Preview mode, Autopilot analyzes unmapped values and identifies potential matches but does not automatically create mappings. The system records similarity scores and tracks how frequently values appear. This mode allows you to observe how Autopilot performs and refine configuration settings before enabling automation.

Active Mode

In Active mode, Autopilot automatically creates mappings when the configured similarity score and frequency thresholds are met. If a value does not meet the automation threshold but still qualifies as a potential match, Autopilot generates a suggestion for steward review through a Data Change Request (DCR) workflow. This mode allows you to automate routine mapping decisions while maintaining governance for cases that require manual validation.

Note:

Switching between Preview and Active modes does not delete Autopilot's learning or existing stats. Use mode switching to pause or resume Autopilot processing without losing learned behavior or tracked data.
Before you switch from Preview to Active mode, review the Insights (statistics) tab. If the AI reasoning looks incorrect, adjust the thresholds, lookup type descriptions, or blocked values first. Active mode starts acting on data immediately.

How Autopilot Evaluates and Resolves Unmapped Values

When a transcode operation fails because a source value does not have a mapping, RDM Autopilot evaluates the value to determine the appropriate action. The system analyzes the value using similarity scoring, and frequency tracking. Based on this evaluation, Autopilot decides whether to automatically create a mapping, suggest a mapping for review, detect a misspelling, or recommend creating a new lookup entry.

This evaluation process ensures that automation occurs only when sufficient confidence and usage frequency are observed.

Note: Autopilot tracks unmapped values separately for each source. The same value from different sources is evaluated independently. Autopilot also calculates the required frequency per source within a rolling time window.

Evaluation Factors

Autopilot evaluates unmapped values using several factors.


Evaluation Factor	Description
Similarity Score	Measures how closely the source value matches existing lookup values using semantic and fuzzy similarity
Frequency Threshold	Ensures that value appears within a defined time window before automation occurs
AI Evaluation	Uses AI models to validate ambiguous matches, detect misspellings, or recommend creating new lookup values

Possible Outcomes

After evaluating a value, Autopilot assigns one of the following outcomes.


Outcome	Description
Automatic Mapping	Creates a mapping between the source value and the canonical lookup value when confidence and frequency thresholds are met
Suggested Mapping	Generates a suggestion and creates a Data Change Request (DCR) for steward review
Misspelling Detection	Identifies a spelling error and caches the corrected mapping for future occurrences
Lookup Creation	Creates a new lookup entry automatically when the value belongs to the lookup type and the configured threshold is met.
Suggested lookup creation	Recommends creating a new lookup entry and creates a Data Change Request (DCR) for steward review when the value belongs to the lookup type but does not match an existing entry.
No Match	No suitable candidate is found, so the value remains unresolved

Note: DCR_PENDING rows are not terminal but remain paused. Autopilot does not process them until the DCR is completed in the workflow. If a DCR is stuck or no longer needed, close or reject it, then delete the stat to allow a new evaluation.

Autopilot Dashboard

The RDM Autopilot Dashboard allows users to monitor Autopilot activity.


Metrics	Description
Automapped	Number of values automatically mapped by Autopilot based on thresholds.
Mappings suggested	Number of mappings generated as suggestions pending steward review.
Lookup Created	Number of new lookup values created automatically, which appear under the Lookup Suggested tab.
Lookup suggested	Number of suggestions to create new lookup values when no match is found.
Misspelled	Number of values identified as potential misspellings.
Ignored	Number of values excluded from processing due to configuration or rules.

Note: Ensure that the tenant is registered in the Workflow Service and has the required permissions to access and use workflow features for Autopilot in the UI.

Unify and manage your data