SQL Bits 2026

2 minute read

Background

I had heard a lot about SQL Bits from colleagues over the years, so it had a lot to live up to. I decided to brave the three-and-a-half-hour drive to Newport for the free Saturday event, and I am glad I did.

There was a real sense of energy around the event, and it was clear from the start why so many people speak so highly of it. The day was packed with interesting sessions, practical insights, and plenty of ideas to take away.

Sessions

There was far too much to choose from, so I decided to focus on areas where I am less familiar and could learn the most.

AI

I attended a two-part session with Scott Bell and Mihaly Kavasl called “Building context, not vibes: practical AI augmented Data Engineering”. It was a deep dive into current capabilities, trends, and ways of thinking more practically about how AI can support data engineering work.

The second AI-focused session I attended was Kevin Arnold’s “Next-Level BI Development: AI-Driven Semantic Models in Power BI”. I found this especially interesting because I had already used some of the AI capabilities in Fabric, so it was useful to see how similar ideas are being applied in other parts of the Microsoft data ecosystem.

Engineering

The UDF-specific session by Oskari Heikkinen, “User Data Functions: Development, Testing and Integration”, led to some of the biggest lightbulb moments of the day for me. The session focused on Transact-SQL User-Defined Functions and how they can be developed, tested, and integrated more effectively.

It was one of those talks that gives you immediate ideas for how you could improve your own approach, which always makes a session memorable.

Security

The other two-part session I attended was “OneLake Security - Centralized Data Security for Microsoft Fabric” with Gabi Münster, Thierry Houy, and Edgar Cotte from Microsoft.

As the speakers said, integrated security is only valuable if it is also understandable and manageable. I found this session particularly useful because security can often feel like something you only think about once a problem appears, when really it needs to be designed in from the beginning.

Update: I left too much time between starting and finishing this post, and by the time I published it, OneLake Security had reached general availability.

Soft Skills

The final session I attended was by my colleague Robert French, who, in his kilt, brought his usual level of passion to a fantastic session on lessons learned and practical approaches to working well with people.

It was a strong reminder that technical delivery is only part of the job. Communication, trust, and how we work with others still make a huge difference to whether projects succeed.

Themes and Takeaways

AI Is Everywhere

Every session I attended either referenced AI or included it as a core part of the demonstrations. That is probably not surprising for the AI-specific sessions, but after speaking to others who had attended different talks, it seemed clear that AI was a recurring theme across the event.

The analogy I keep coming back to is email in the workplace. Its impact was not immediate, but over time it fundamentally changed how work got done. I think AI is following a similar path. Adoption may be uneven at first, but it increasingly feels like it will become part of almost every workflow.

Testing is not optional

During Robert’s talk, I wrote down the phrase “Popodies Nerfect”, and it stuck with me. Pipelines will fail eventually, and mistakes will happen. What matters is how well prepared we are for that reality.

I do not mind apologising when something goes wrong, but I want to be able to show that the right effort was made to prevent the issue in the first place. For me, that was the biggest takeaway of the day: what can we do to make sure we are building and testing things properly before they fail in production?

Conclusion

Thank you to everyone involved in SQL Bits. I really enjoyed the day, learned a lot, and came away with plenty to think about. It was well worth the drive👍

🤖 For transparency Github Copilot has been used to assist the writing of this post