The first version of our Smart Narrative solution proved the concept. A lightweight layer sitting inside Power BI, pulling KPIs from the data model, passing them through an API to an AI engine, and returning a polished narrative directly inside the report. The output was good. The business case was clear.
We wanted to take it further.
The next step
Certain deployment environments - embedded reports, users accessing reports without a Power BI licence presented a constraint worth solving. We wanted the narrative to work everywhere the report works, without exception.
So we rebuilt the packaging.
How it works
The second version is a fully custom Power BI visual. From the outside it looks like a text box - clean, unobtrusive, sitting naturally inside the report canvas. What sits behind it is the complete narrative engine, contained entirely within the visual's own code.
When the report loads, the visual reads the KPIs and measures already defined in the data model - values produced by complex calculations executed at the data layer, the numbers the business already trusts. Those values are bundled with a precise brief: what to write, for whom, in what tone, with what focus. That brief travels through an API to the AI engine. The response comes back as a fully formed narrative, rendered inside the visual.
No separate scripts to manage. Everything self-contained.
Change a filter - a different region, a different period, a different segment - and the visual responds. The narrative updates to reflect the new context. Same report, same visual, new story in seconds.
Why this matters
It works on Power BI Desktop. It works in Power BI Service. It can be embedded anywhere - a client portal, a third-party platform, a customer-facing dashboard. Wherever the report lives, the narrative travels with it, and it works for users regardless of whether they hold a Power BI licence.
That last point is significant. Most organisations have far more report consumers than licence holders. Dashboards get embedded in portals, shared via links, surfaced inside other platforms. The people reading those reports are often the ones who most need a clear written explanation of what the data means. Now they get one.
What stayed the same
The underlying principle has not changed. The AI does not invent anything. It reads what is already in the model - the measures, the comparisons, the logic that has been carefully built at the data layer. The quality of the narrative is a direct reflection of the quality of the data foundation underneath it.
A well-built data model already contains the story. This is just a more robust, more deployable way of letting it speak.
Where this leaves us
The Smart Narrative is no longer a proof of concept. It is a production-ready solution - custom-built, fully contained, and deployable across the environments where real business reporting actually happens.
The last mile problem is not fully solved. But it is considerably shorter than it was.
If your organisation is sitting on a well-built data foundation and the gap between the numbers and the people who need to act on them is still being bridged manually - we would be glad to talk. Get in touch with us at Plumgate.