QAiry 3.1: A Major Step Forward for Audience Building in SFMC

Building an audience in Salesforce Marketing Cloud is still, in 2026, an exercise in patience: a SQL query to write, Data Extensions to join from memory, an Automation to schedule, and a quiet hope that the refresh chain holds together. QAiry 3.1 goes after all three frictions at once, with an AI that actually understands your data model, a visual mode that has grown into a genuine builder, and waterfall automation that reaches past what Automation Studio can express. Here is what the release changes, feature by feature, and what each one solves inside a working CRM team.

What 3.1 changes, in one sentence

QAiry stops being a tool where you ask an AI for an audience and becomes a workspace where you build one, by whichever route suits you. Three entry points now live in the same screen: Chat, Manual Edit and SQL. Start in one, continue in another, come back to the first, and nothing is lost along the way.

That is a meaningful shift in posture. SFMC teams are not homogeneous. The data analyst wants to write SQL. The CRM project manager wants to assemble filters. The marketer wants to describe the requirement in plain language. Tools have historically privileged exactly one of those people. 3.1 stops choosing on your behalf.

A chat that finally understands your data model

The first pillar of the release is answer quality. QAiry 3.1 reads your request in the context of your Salesforce Marketing Cloud environment rather than in the abstract. The difference shows up most on complex estates: the tool has a much better grasp of your Data Extensions, their fields, and above all the relationships between them.

That last point is what tips the experience. A realistic SFMC model is not three tables. It is Clients, Orders, Products, Invoices, Payments, Optin and Events, wired together with 1:1, 1:N and M:N cardinalities that will happily manufacture duplicates the moment a join is chosen carelessly. An AI that is blind to cardinality produces SQL that is syntactically perfect and functionally wrong. That is precisely the failure mode 3.1 targets.

On top of accuracy sits sharper guidance: QAiry works out what you are trying to achieve and points you toward the right next step while you refine the audience, rather than answering only the question you happened to type.

Connecting a Data Extension without leaving the conversation

The most quietly useful change is elsewhere. When QAiry needs a Data Extension that is not connected yet, you can search for it, select it and connect it straight from the conversation. No trip to a configuration screen, no return to chat, no rewriting the prompt, no losing your train of thought.

Across one real audience-building session that is three or four interruptions avoided. It does not look dramatic in a demo. It is very noticeable in production.

Four kinds of Data Extension, one workspace

QAiry 3.1 widens what you can pull into audience logic. Four families of Data Extension are now available:

Type	What it gives you
Data Extensions	The DEs connected to your Business Unit, the usual foundation of your audiences.
Shared Data Extensions	Shared data enters your audience logic, so you can finally work across Business Units.
Synchronized Data Extensions	CRM data synchronised from Salesforce, usable directly in the audience-building workflow.
Salesforce Data Extensions	Salesforce-generated DEs, bringing even more CRM data into your targeting.

For multi-BU organisations this is the most structural item on the list. Cross-BU segmentation has historically been solved by copying data around, with all the freshness drift that implies. Handling it at the audience-logic layer removes a duplication step and a standing source of error.

Manual Edit is now a builder

In earlier versions, Manual Edit was where you went to fix what the AI had produced. In 3.1 it is a legitimate starting point: you open Manual Edit and design the logic step by step without writing a single prompt.

The flow is five steps:

Select the connected Data Extensions you want to use.
Choose the join logic between them and define how the data should be connected.
Add the filters your contacts must satisfy to be included.
Apply exclusions, defining who should be kept out.
Review the output and organise the fields in the final audience.

The exclusion step deserves a pause. It is the one that SQL generators handle worst, and it is also the one protecting your contact pressure and your deliverability. Treating it as a first-class step rather than a clause to bury inside a WHERE measurably improves the audiences that come out the other end.

The real story of 3.1 is not that the AI got smarter. It is that you are no longer obliged to go through it. Build first, ask the AI later, or combine the two — the entry point is yours to choose.

Prefer SQL? Start there

The third entry point is the SQL tab. Write or paste your query, then carry on refining the audience inside QAiry. Movement between modes is unrestricted: SQL to Manual Edit to Chat, or the reverse.

In practice, a deduplication query like this one — most recent order per opted-in client — becomes a starting point you then enrich visually:

SELECT t.Clients_Subscriberkey,
       t.Clients_Email,
       t.Orders_Date
FROM (
    SELECT c.Subscriberkey AS Clients_Subscriberkey,
           c.Email         AS Clients_Email,
           o.[Date]        AS Orders_Date,
           ROW_NUMBER() OVER (
               PARTITION BY c.Subscriberkey
               ORDER BY o.[Date] DESC, o.Id DESC
           ) AS RowNum
    FROM Clients c
    INNER JOIN Optin  op ON op.Subscriberkey = c.Subscriberkey
    INNER JOIN Orders o  ON o.Client_id      = c.Subscriberkey
    WHERE op.Optin_email = 'true'
) t
WHERE t.RowNum = 1

This is the kind of query nobody wants to rewrite for every campaign, and nobody wants to hand blindly to an AI either. Being able to bring it in as-is, then add exclusions with a mouse, is an honest compromise.

Waterfall automation: refreshing in the right order

The final pillar is the most technical, and probably the most valuable to teams already running QAiry in production. Waterfall automation chains dependent Data Extension refreshes sequentially, so each step starts only once the previous one has finished.

The canonical shape looks like this: base audience, then enrichment with profile and behavioural data, then exclusions and suppression rules, then the final published audience. Every link waits explicitly for the one before it.

Anyone who has tried to reproduce that logic in Automation Studio knows the usual method: estimate how long each Query Activity takes, add a safety margin, schedule the automations at fixed intervals, and hope. The day volumes grow, query one overruns into query two and the final audience ships with partially stale data — with no error raised anywhere.

Declaring a dependency instead of a clock time eliminates that class of silent failure and takes timeout pressure off long refresh chains. It is paired with an API-triggered refresh, so those chains can hang off your own business events rather than a calendar.

Where to start

Map your Data Extension relationships in QAiry first. The quality of every chat answer depends directly on it.
Identify the multi-BU audiences you currently build by copying data. Those are your first candidates for Shared Data Extensions.
Reopen an existing audience in Manual Edit and isolate its exclusions. You will usually find suppression rules buried inside a WHERE.
List every automation chain you schedule at a fixed time. Each one is a waterfall candidate.
Have your technical people try SQL and your business people try Chat on the same audience, then compare notes.

Key takeaways

Three entry points, one workspace. Chat, Manual Edit and SQL share the same screen, so every profile on your team can work the way they already work, on the same audience.

The chat understands relationships, not just tables. That is the difference between SQL that parses and SQL that is correct on a model full of 1:N and M:N joins.

Cross-BU becomes native. Shared, Synchronized and Salesforce Data Extensions enter your audience logic and remove a data-copying step.

Exclusions are promoted to a first-class step. It is the single best lever you have on contact pressure and deliverability.

Waterfall automation replaces the scheduling gamble. A declared dependency beats an estimated margin, particularly on the day your volumes double.

Want to evaluate QAiry 3.1 against your own Salesforce Marketing Cloud data model, or audit the refresh chains you run today? Talk to our SFMC consultants.