AI-driven suggestions

AI-driven suggestions are the first type of notification in the Notification Center. The platform quietly watches your data quality trends and speaks up only when something meaningful can be improved — so instead of hunting for gaps yourself, you get high-quality recommendations brought to you.

This first version focuses on suggested new rules: data quality rules the platform recommends based on your data, your existing quality patterns, and gaps your current rules don't cover.

At a glance

• Recommendations to improve data quality, gathered in the Notification Center

• Generated per ruleset, for rulesets you subscribe to and monitor

• High-confidence only and limited per week, so the panel stays low-noise

Turning suggestions on

Two settings control whether you get suggestions for a ruleset:

• AI monitoring — whether the platform analyzes a ruleset at all (a per-ruleset toggle, off by default)

• Subscription — whether you get notified about that ruleset

To start:

1. Open a ruleset you have edit rights for. The subscription bell and the AI-monitoring toggle live in the ruleset header and are only shown to users with edit rights.

2. Subscribe to the ruleset (bell icon) and switch AI monitoring on. Monitoring runs the suggestion logic (which adds some extra computation); subscription routes the resulting suggestions to you.

You may already be subscribed automatically. Auto-subscription applies to rulesets you created, the Default ruleset of a table you connected, and workflows you created. When you connect a new connector, the platform also offers to switch monitoring on for the newly connected tables right away.

You can turn monitoring on later from the ruleset header, and unsubscribe at any time.

How suggestions are generated

Once monitoring is on, the platform reviews the ruleset about once a week — looking at your data, profiling information, previous ruleset checks, detected issues, and your existing rules — and generates suggestions for the subscribers of that ruleset.

Because analysis is weekly, it can take a few days after you enable monitoring for the first suggestions to appear. The platform only surfaces suggestions it's confident about, and never shows internal confidence scores or technical noise.

Reviewing and acting

Open the Notification Center from the top-right corner. For each suggested rule you can:

• Review it — including a plain-language explanation of why it was recommended — and Accept it to add the rule, or

• Dismiss it.

Once a suggestion is accepted or dismissed, it's removed from the panel automatically. You can also filter for unread suggestions or deleted ones you no longer need.

Working as a team

Suggestions are generated per ruleset, not per person:

• When any subscribed teammate accepts or dismisses a suggestion, it's resolved for all subscribers.

• If you try to act on one a teammate already handled, you'll see a message such as "This suggestion has already been accepted by another team member."

• Dismissed suggestions are not re-suggested.

It learns from feedback

Over time the platform learns from how suggestions are handled — favoring the kinds that get accepted and easing off patterns that get dismissed. This feedback is tracked per ruleset, reflecting what's valuable for that data asset rather than individual preferences.

The DQ-AI Assistant knows your suggestions

The DQ-AI Assistant is aware of the active suggestions for the ruleset or workflow you're working in. You can ask it to explain a suggestion in more detail or walk you through accepting it, and it avoids repeating suggestions you already have.

What's coming next

Future versions will add more suggestion types beyond new rules — for example rule-update suggestions (when changing data warrants adjusting an existing rule), quality investigations (flagging notable quality-score changes worth looking into), and improvement suggestions that recommend workflow automations to resolve recurring issues.