Practical thinking on data, BI and AI

Deneb, HTML and SVG let you show data the way it needs to be shown, not just make it look good. Here’s how far you can take them, and when to stay native.

Fabric Apps, now in preview, gives you a database, an API, sign-in and hosting from one command. We spent time with it, and we are not yet sure who it is for.

We build agents that do in a few minutes what we estimate would take an analyst hours. So can they replace analysts? Partly. With the right context behind them, more business users can become their own analysts for everyday questions, freeing analysts for the new and complex work. Whether that happens depends more on the business than on the technology.

We have used Claude every day on Power BI for a few months. Here is what it has changed, where it is unmatched, and what it took to get dashboards most of the way there.

Databricks is building the connectors we used to hand-build into the platform. Our early experience was not great, but the effort since is clear, so we are revisiting them and will report back.

AI can now build a working first version of a Power BI model and dashboard in a few hours. That changes how we deliver Power BI, and what clients should expect from us. It doesn’t remove the need for experienced developers. It moves where their time goes.

Building one AI agent is no longer difficult. Turning many of them into a single system the whole business can rely on is where projects stall. These are the six decisions we made at the start that took our own platform from prototype to everyday use in weeks.

One began as a single simple flow, the other moves around 15,000 records a day across a hundred stores. Different contexts, different scales, and both built on the same integration accelerators. Here is what stays the same whatever the scale, and what changes.

Five questions an email marketing team should be able to answer from its data: from the true impact of a campaign to leaking discount codes and lapsed customers worth winning back.

Most renewal risk models treat usage as one number, when it isn’t. What people click and what actually flows through the product are two different signals. Where they disagree is where risk and opportunity can be identified first.

Here’s what’s changed, what AI is really capable of, and how to use BI and AI together without losing trust in the numbers.

Every business owner has sat in a meeting where two reports show two different numbers, and nobody can say with confidence which one is right. If that sounds familiar, the problem is usually your systems, and the fact that they were never actually connected.

Here’s why it happens, what it’s costing your business, and how the right business reporting solution fixes it.

Two parallel workstreams, eight data sources, one Claude interface. An inside look at the architectural choices that make AI on enterprise data actually work, and what they tell you about how to do this properly.

The problem usually isn’t the data. It’s the absence of a shared, trusted layer that makes that data mean the same thing across the whole business.

Plenty of organisations have BI. Far fewer have seen what BI can really do. The gap between functional and great is bigger than most people realise, and it changes how people actually engage with their data.

Started with a single proof-of-concept. Now powering every part of the business. Around fifteen source systems, a full suite of Power BI Apps used across the organisation, an integration layer carrying thousands of transactions a day, and a marketing analytics programme building on top of all of it. Here is what a properly built retail data platform actually looks like.

For a luxury brand, every customer touchpoint has to feel considered. So why should the dashboards that run the business be any different? Inside the build of a Power BI environment designed to match the brand it serves.

The most practical first step is not a big AI strategy. It is putting a data agent on top of the data and BI you already have.

Most organisations invest in storing data. Fewer invest in making sure that data means the same thing everywhere. That’s the gap semantic modelling closes, and it’s the gap that determines whether your data platform actually delivers.