Some years in this field are incremental — a few new features, a steady march, nothing that changes the shape of the work. 2023 was not one of those years. Two things happened, more or less at once, that genuinely moved the ground under the profession, and a third quietly reminded us that some things don't move no matter how loud the year gets. As the calendar runs out, it's worth marking what actually shifted, because distinguishing the real change from the noise is the whole job of a year-end reflection — and there was an unusual amount of both to sort through.

The obvious one: generative AI ate the conversation

There's no pretending otherwise: the story of 2023 in data was generative AI going from a curiosity to the thing every executive on earth wanted a strategy for. A tool the whole world could use put the capability in front of everyone, the vendors responded with a wall of announcements, and by mid-year "what's our AI plan?" was echoing through every organisation with data to its name. Microsoft threaded AI assistants — Copilots — through its entire product line, promising that natural language would sit on top of everything.

What actually changed for those of us who build the foundations is subtler than the hype suggested, and I spent a good part of the year saying so. The genAI wave didn't replace the need for good data architecture; it raised the stakes on it, because a confident AI answering questions on messy, ungoverned data is worse than no AI at all. The organisations that spent 2023 quietly getting their data house in order — cataloguing, quality, governance, access — are the ones who'll actually be able to use this stuff. The ones who rushed a fragile demo to impress the board mostly learned an expensive lesson about foundations. The capability is real and important. The panic around it produced more bad decisions than good ones. Both of those were true all year.

The Microsoft one: everything folded into Fabric

The other genuine shift, at least in the Microsoft corner where I spend most of my time, was Fabric. Announced in preview in the spring and reaching general availability in November, it represents Microsoft folding its sprawling data portfolio — Power BI, Synapse, Data Factory, and more — into a single platform over one storage layer. It's the most significant restructuring of the Microsoft data stack in years, and it's a real bet on consolidation: the idea that most organisations would rather have one coherent platform than a best-of-breed sprawl they have to integrate themselves.

Now that it's generally available and people are actually asking whether to move, the honest guidance is measured. GA means it's real and worth serious engagement, but a brand-new platform is still a young one, and "available" is not the same as "battle-tested in your context." The organisations that spent the preview period preparing — getting data into clean open formats, sorting governance, understanding their workloads — are positioned to get real value now. The ones waiting for a signal to start: this was it, but start with your eyes open.

The reassuring one: the fundamentals didn't budge

Here's the part I find genuinely steadying, and it's the thread through both big stories. For all the ground that moved in 2023, the fundamentals of good data work stood exactly where they always have. Every one of the year's big shifts increased the value of the boring foundations rather than replacing them. Generative AI made clean, governed, well-described data more valuable, not less. Fabric concentrates whatever you bring to it, rewarding good practice and punishing bad — so the unglamorous groundwork mattered more, not less. The profession spent the year being told everything was changing, and the deepest truth turned out to be that the things worth doing were the same things, now with higher stakes and better reasons.

The one I'm still uneasy about

If I have a worry carrying into next year, it's the speed at which "we need AI" outran "we understand what we're deploying." I watched organisations this year commit to AI initiatives with genuine enthusiasm and almost no clarity about the data risk underneath — who can see what, where the answers come from, whether the thing can be trusted or explained. That gap between ambition and readiness didn't close over 2023; if anything it widened, because the ambition accelerated faster than the groundwork. It's the quiet story under the loud one, and I suspect it's where a lot of 2024's expensive lessons are going to be learned. I'd rather we learned them cheaply, in preparation, than publicly, in production — but the year's momentum wasn't pointing that way, and hoping it corrects itself isn't a plan.

Into 2024

I don't think the pace lets up. Generative AI will move from "everyone wants a strategy" to "everyone wants results," and that shift will be brutal for the pilots built on sand and kind to the ones built on foundations. Fabric will mature, and the interesting questions will move from "should we adopt it" to "how do we run it well." And governance — the least fashionable word in the field — will keep quietly rising in importance, because every one of these shifts makes it matter more. If I had to bet on a single theme for next year, it's that the gap widens between the organisations that did the boring, foundational work and the ones that chased the demos. 2023 opened that gap. 2024 will show it.

For now, the ground has shifted, and it's been a genuinely remarkable year to be doing this work — equal parts exhilarating and exhausting, as the interesting years tend to be. But the compass still points the same way it always did: get the foundations right, translate honestly, distrust the hype without dismissing the substance. The tools changed a great deal in 2023. The job, underneath, changed hardly at all. See you in the new year.