Power BI's 'agentic era' arrived this month. Here is the part an SME can actually use

If a small business runs any reporting at all, it tends to live in one of two places: a spreadsheet someone updates by hand on a Friday, or a single Power BI report a member of staff built once and now nobody quite wants to touch. So when Microsoft spends a keynote describing artificial intelligence that builds reports for you, it is reasonable to wonder whether the whole job is about to get easier, or whether this is another announcement that turns out to be for somebody much larger. This month gave a clear answer, and it is a useful one to be able to read for yourself.

What Microsoft announced

Two dated, first-party events sit behind the headlines. At Build 2026, on 2 June, Microsoft framed what it calls the agentic era of analytics: artificial-intelligence agents that can build the data models and reports a business relies on, described in its own Power BI Updates post 4. Then on 9 June the monthly release landed, the routine What’s new in Power BI update that ships features on a fixed cadence, several of them reaching what Microsoft calls General Availability 1.

The two are easy to blur together, and the gap between them is the whole point. The keynote set out where the product is heading. The monthly release is what actually became dependable. Reading a release well means telling those two things apart.

The headline: agents that build the report for you

The genuinely eye-catching announcement is the agent. In Microsoft’s words, you describe what you need to an artificial-intelligence agent, even from a screenshot, and it can create models, generate reports, and iterate on the visuals 4. There is a companion idea, Fabric Apps, custom data applications built on top of a business’s data for jobs like planning or stock control. It is a real direction of travel, and worth knowing about.

The status is the part the headline tends to drop. These agent features are in Preview 4. Preview means available to try, not finished and not yet something to build a business process on. It is the difference between a feature that has been announced and a feature that is safe to depend on, and for a small team with no spare capacity to babysit a half-finished tool, that difference matters.

There is a deeper dependency underneath. Microsoft’s own framing puts it plainly: artificial intelligence without that grounding is, in its words, a confident guesser 4. The agent reasons over a governed data model, a properly structured and trusted layer that defines what the numbers mean. That model is precisely what most small firms running a spreadsheet or a single report do not yet have. The clever agent assumes the unglamorous groundwork is already done.

What actually reached General Availability this month

Underneath the keynote, the 9 June release shipped a batch of smaller, finished improvements, and one of them is quietly worth a small team’s attention: DAX user-defined functions reached General Availability 1. DAX is the formula language Power BI uses to calculate things, and a user-defined function lets whoever builds a report write a calculation once, give it a name, and reuse it everywhere, rather than copying and pasting the same formula across a dozen places 2.

That sounds technical, and it is, but the benefit is the opposite of technical. When one person maintains all the reporting, every duplicated formula is a place a rule can quietly drift or break. Microsoft’s own guidance frames the gain as maintainability: you update the logic in one place rather than hunting for every copy 2. For a business where the reporting lives in one head and one set of files, fewer places for a number to go wrong is a real, if modest, improvement. The honest caveat: it is a feature for whoever builds the model, not the person reading the dashboard, and it needs the current modelling engine to use it 3.

The rest of the General Availability list is reporting polish rather than headline material: clearer axis behaviour on bar and column charts, tables that expand their rows and columns more sensibly, plainer tooltip options 1. None of it changes what is possible. All of it makes an existing report a little easier to read and maintain, which is usually what a small team actually wants from a monthly update.

How to read a Microsoft release: GA, Preview, and Frontier

The single most useful habit when an analytics release lands is to check the label on each feature, because Microsoft ships at three different levels of readiness, and only one of them is something to plan around.

• General Availability (GA) is finished and supported. This is what a business can sensibly adopt and depend on. The reusable DAX functions and the chart improvements above are GA 1.
• Preview is available to try but not finished. Good for experimenting, not for a process the business runs on. The agent features sit here 4.
• Frontier is narrower still: the earliest, most limited access, often the flashiest demonstrations. This month it covers some of the conversational, ask-your-data features grounded in a governed model 1.

The pattern worth noticing is that the most impressive items tend to sit at the least-available end. The reusable formula reached General Availability quietly; the agent that builds your reports is a Preview; the conversational data answering is Frontier. Reading a release by what reached GA, rather than by what was demonstrated on stage, is a habit that saves a small business from planning a quarter around something that is still months from being dependable.

Does this matter for an SME yet? The honest take

The agentic-analytics direction is genuine and worth watching, but most of it is still in Preview and built for organisations that already run a governed data model, which is the layer most small firms have not put in place. So for a small business this release is a signal of where reporting is heading, not a tool to adopt this quarter. The change a small team can genuinely use today is the quieter set of generally-available improvements, reusable calculations and clearer chart defaults, that make an existing report easier to trust and maintain.

That is the unflashy reading, and it is the right one. The robots are not building your dashboards yet. What is worth doing now is the groundwork the clever features will one day assume: getting the numbers a business relies on into one trustworthy place, defined once and reported consistently, so that when the agent era does arrive it has something solid to reason over. That is the part that pays off whether or not the keynote features ever land.

Worth a closer look

See how hirevolution helps small teams get reporting they can trust