AgentFlow™ Quality uses natural-language prompts to guide you through data quality assessment and publishing validated data to Reltio. The following samples show how to start an assessment, adjust rules, publish validated data to Reltio, and revise or repeat steps in the current workflow.

Assessing data quality

✅ Prompt: "Start data quality on my Databricks source."

Why it works: TThe phrase "start data quality" matches the assessment workflow and opens a new conversation.

Adjusting quality rules

Use prompts that describe the rule you want to add, or the change you want to make to the suggested rules.

• ✅ Prompt: "Add a rule that flags rows with duplicate customer IDs."

  Why it works: Describing the rule logic in natural language lets the agent add it alongside the suggested rules.

• ✅ Prompt: "Add a few more rules for this table."

  Why it works: The agent adds to the existing rule set without replacing rules you already selected.

• ✅ Prompt: "I only need 2 rules for this table."

  Why it works: Specifying a number reduces the suggested rules to that count.

• ✅ Prompt: "Why did the duplicate customer ID rule fail, and how do I fix it?"

  Why it works: Naming the specific rule and asking for both an explanation and next steps returns a targeted diagnosis.

Publishing validated data to Reltio

Use prompts that confirm you want to continue from an assessment into a Reltio Data Loader job.

• ✅ Prompt: "Map and load this table into Reltio."

  Why it works: The prompt triggers the agent to start a Data Loader workflow to map the assessed table to your Reltio tenant schema, reusing the Databricks connection from the current conversation.

• ✅ Prompt: "Yes, publish to Reltio."

  Why it works: After a data quality run, the agent offers to publish. Accepting keeps the same connection and table without re-authentication.

Revising, repeating or canceling a step

Use prompts that describe the step you want to revise or the workflow you want to stop.

• ✅ Prompt: "Go back."

  Why it works: The agent returns to the previous step so you can change your selection.

• ✅ Prompt: "Regenerate the rules."

  Why it works: The current rules are discarded and a new set is generated based on the same table.

• ✅ Prompt: "Redo table selection with a different table."

  Why it works: The agent replays table selection without restarting the connection.

• ✅ Prompt: "Cancel."

  Why it works: The current workflow stops. You can start a new conversation at any time.