AI Managed Checks Introduction

Overview

An AI Managed check is a quality rule generated from observed data behavior. Most are created during a Profile operation. For example, Profile may find that a customer ID is consistently present or that a price remains positive, then generate and validate a matching check. AI Managed Volumetric and Freshness checks can instead use Observability measurement history. These observations create useful starting coverage, but they do not supply the business context that determines whether a rule is appropriate for every use case.

Future Profile runs maintain Profile-generated AI Managed checks as observed data changes. AI Effort controls the breadth of pattern analysis, while Oversight determines whether new checks become active automatically or remain drafts for review. If profiling no longer supports a rule, Qualytics may remove that Profile-generated check. Observability automation maintains Volumetric and Freshness checks as new measurements are recorded.

When you change a property that controls how an AI Managed check evaluates data, Qualytics converts it to an Authored check. From that point, your team owns the rule and future Profile runs leave it unchanged. This preserves the business decision behind your edit.

Why this matters

Without the conversion, future profiling could replace a deliberate adjustment. For example, a data steward may lower coverage from 100% to 99.5% to account for a documented exception. Converting the check to Authored preserves that decision.

It also makes ownership visible. The Authored label tells teammates that a person has reviewed and changed how the rule evaluates data.

Edits that turn an AI Managed check into an Authored check

The conversion happens automatically when you change any of the following on an AI Managed check:

• The rule properties: the values that define what the rule checks. Examples include the threshold on a Min Value check, the regular expression on a Matches Pattern check, or the list of accepted values on an Expected Values check.
• The coverage: the percentage of records the rule must satisfy.
• The filter clause: the SQL WHERE expression that limits the rows the check evaluates.
• The fields the check applies to: adding, removing, or replacing a field.

You don't need to confirm the change in a dialog: the conversion happens silently as soon as you save the edit.

Edits that keep the check AI Managed

Some edits are organizational and don't change how the check evaluates data. They are saved without converting the check:

• The description.
• The tags.
• The additional metadata (key-value pairs).
• The anomaly message field.
• The owner.
• The anomaly assignee.
• The check status (for example, switching between Active and Draft).

This means you can document, categorize, and triage AI Managed checks without taking them out of automatic management.

What changes after the conversion

Once an AI Managed check becomes Authored, three things happen:

1. The check's badge in the Check Summary switches from AI Managed to Authored, so anyone looking at the check can tell at a glance that it is user-managed.
2. Profile runs no longer touch the check. Qualytics will not change its rule, coverage, filter, fields, or status during future Profile operations, and it will not delete the check even if profiling no longer detects the same pattern in your data.
3. If the AI Managed check was originally owned by Qualytics, ownership is transferred to you, the user who edited it. The check then shows up in lists like "checks I own", and notifications about the check follow you instead of Qualytics.

How to go back to an AI Managed check

There is no one-click "revert to AI Managed" action. For a Profile-generated check, delete the Authored version and run Profile again. Qualytics AI will recreate an AI Managed version if it still detects the same pattern. For Volumetric or Freshness, later Observability measurements can recreate an AI Managed check when the matching automation setting is enabled and enough history is available.

Next Steps