Our most recent community show featured an in-depth exploration of the Reltio AgentFlow™ platform, specifically focusing on how the new Profiler agent is revolutionizing the data onboarding process. The session, led by Saurabh Agarwal, Principal Product Manager, and Kriti Agarwal, Product Manager at Reltio, covered the technical architecture of AgentFlow, the diverse library of prebuilt agents, and a live demonstration of "shifting left" on data quality.
The AgentFlow™ Platform: A New Era of Agentic Intelligence
Saurabh introduced Reltio AgentFlow™ as an agentic intelligence platform built natively on the Reltio Data Cloud®. At its core is the Model Context Protocol (MCP) server, which hosts all Reltio APIs as "tools" that agents can call. By combining these tools with sophisticated system prompts and LLMs (like Claude 3.5/4.5 Sonnet), AgentFlow creates a conversational, interaction-driven experience for managing complex data tasks.
Meet the Reltio AgentFlow™ Prebuilt Agents
The session provided a comprehensive breakdown of the specialized agents currently available or in development:
-
Profiler Agent: Profiles data directly at the source (S3, Azure, GCP, SFTP) before ingestion to assess match readiness and schema alignment.
-
Match Resolver Agent: Researches potential matches and provides merge/no-merge recommendations with full auditability.
-
Validator Agent: Uses natural language prompts to automatically generate and post complex data validation functions (e.g., regex for emails).
-
Unmerger Agent: Analyzes merge history to identify and reverse incorrect merges.
-
Work Assigner Agent: Monitors and redistributes Data Change Request (DCR) tasks to balance steward workloads.
-
Data Explorer Agent: A read-only agent that allows users to query their tenants using natural language instead of complex APIs.
-
Address Enricher Agent: Automatically corrects ambiguous addresses by performing multi-source web research.
-
Product Recommender Agent: Generates personalized product suggestions based on customer interaction and buying history.
-
Workflow Configurator (In-Progress): Automates the creation of BPMN diagrams and workflows through simple prompts.
Solving the "Data Timing" Problem
Kriti highlighted a critical blind spot in modern enterprises: discovering data quality issues only after the data has been mastered. This "data timing problem" costs companies millions in rework and failed downstream processes.
The Profiler agent addresses this by acting as an intelligence layer that evaluates data before it touches the tenant. It doesn't just provide a static report; it maps source attributes to your specific tenant schema, identifies corrupted values, and suggests correction strategies.
Key Features of Reltio AgentFlow™ Profiler
-
Pre-Ingestion Validation: Scans datasets in cloud storage (AWS, Azure, GCP, SFTP) to find nulls, invalid formats, and inconsistencies before they enter the system.
-
Intelligent Mapping: Automatically aligns source columns to Reltio entity attributes (e.g., Individual, Organization) with confidence scores.
-
Conversational Cleanup: Users can probe the agent for specific corrupted values (like invalid gender codes or email patterns) and receive fix recommendations.
-
Automated Data Loading: Once the mapping and quality checks are approved, the agent executes the data load job directly from the interface.
-
Scalability: Built to handle enterprise-grade volumes, with testing successful up to 10 million records.
Conclusion
The latest community show demonstrated that the Reltio AgentFlow™ Profiler is more than a tool—it’s a control point for data trust. By moving quality checks upstream, organizations can drastically reduce onboarding time, eliminate trial-and-error ingestion, and ensure their Reltio Intelligent Data Graph is fueled only by high-quality, reliable data.
#CommunityWebinar
#Featured
#AIML