Couldn't agree more with both Marco and Rob below. A well-structured semantic model already has SO MUCH of what would allow an LLM to succeed, because it's the same inputs that your humans need to succeed in running your business. Business context? ✅ Synonyms specific to your business? ✅ Human-vetted translations? ✅ Structured relationships connecting multiple tables of data? ✅ Human-vetted measures to ensure that your core KPIs/OKRs are being calculated correctly? ✅ LLMs can do a lot of good for your company, but they shouldn't be net-new. Look beyond the hype train and build on top of what already works inside of your organization if you truly want to get the most out of their potential.

This morning, I was 𝐬𝐡𝐨𝐜𝐤𝐞𝐝 😲 (not kidding) to hear that there are consulting firms suggesting that data be moved out of semantic models to achieve a better consumption experience with LLMs. Really? What a lie. Don't listen to my words. I have a vested interest in Power BI semantic models. Hear what a lie this is, straight from Rob Collie in his podcast (linked below - if you don't have much time, go straight to 26:30). 𝑇ℎ𝑒𝑟𝑒 𝑖𝑠 𝑛𝑜 𝑏𝑒𝑡𝑡𝑒𝑟 𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒𝑑 𝑑𝑎𝑡𝑎 𝑠𝑜𝑢𝑟𝑐𝑒 𝑜𝑛 𝑡ℎ𝑒 𝑝𝑙𝑎𝑛𝑒𝑡 𝑛𝑜𝑤 𝑜𝑟 𝑖𝑛 𝑡ℎ𝑒 𝑓𝑜𝑟𝑒𝑠𝑒𝑒𝑎𝑏𝑙𝑒 𝑓𝑢𝑡𝑢𝑟𝑒 𝑓𝑜𝑟 𝑎𝑛 𝐿𝐿𝑀 𝑡𝑜 𝑡𝑎𝑙𝑘 𝑡𝑜 𝑡ℎ𝑎𝑛 𝑎 𝑃𝑜𝑤𝑒𝑟 𝐵𝐼 𝑠𝑒𝑚𝑎𝑛𝑡𝑖𝑐 𝑚𝑜𝑑𝑒𝑙. If you heard stories like "consulting firms suggesting to get all data out of Power BI semantic models", please write your story in the comments. I understand competition, but I cannot justify lying to customers. Rob Collie has already expressed my thoughts, so I don't need to add anything else; I agree with him. p3adaptive.com/rawdatapodcas…

With all of the #MVPBuzz posts going around right now, I wanted to send a shoutout to the Power BI and Fabric community at large. To all of the MVPs: Thank you for taking us on your journey with you, whether you're cataloging your own learning along the way, providing throught leadership and new ideas, or helping digest complex topics into something more digestible for the masses. To all of the community members who don't create their own content, but contribute through asking questions, providing advice, and pushing the conversations forward, y'all are my personal favorites (and I've done a lot more of this than content creation). This is what I really see as the lifeblood of the community and the reason why I continue to absolutely love working in this field. To all of the folks at #MicrosoftFabric on the CAT, Product, Engineering, and Community teams, y'all are really doing something special. Even if it feels natural at this point to have open and honest conversations with your stakeholders, customers, and community, it's important to remember how much of what you do is NOT the norm across other large tech companies out there, and it has always been what makes this platform special. I've been very lucky to not only form some great professional relationships from this community, but also some very fun friendships with interesting people along the way, and have been able to leverage community connections to land me jobs in the past, as well as my new role that will be announced later this month. Whether it's through live streams, user groups, feedback sessions, or just partaking in all of the great social media conversations — get plugged in. It might have a way bigger impact than you imagine.

This is the best single article I've read so far that describes what you can and cannot do with AI-generated code. Suitable for developers and non-developers to understand what is going on beyond the hype. diwank.space/field-notes-fro…

Last week on SQLBI we published an article about tools in Power BI. Here's an overview of some (not all!) external tools that can help you, and at what stage in. Check below for a downloadable PDF. #powerbi #microsoftfabric

It was awesome to see the massive #MicrosoftFabric community at #FabCon. The keynote is now available with tons of announcements and demos. Check it out. t.ly/Owy2W #PowerBI

So it looks like my #FabCon session ended up ranking #1 in attendance for non-Microsoft presentations, and #3 overall for entire conference What a surreal experience to have over 1000 people want to hear me speak, according to the app I love the @MicrosoftFabric community

Exporting your model metadata and DAX calculations isn't "documenting your model". Why would I look at the same information that's in the actual model, but structured in a less useful and functional way? Good documentation should focus on purpose, processes, and people.

Thrilled to be able to say that I'll be speaking at the upcoming #MicrosoftFabric FabCon in Vegas! Hoping to see some familiar faces in the room, check out my session below azuredataconf.com/#!/session…

After sending out my same "starter kit" to a few handfuls of people over the years, I finally decided to just type it up into a blog post! This is meant to take you beyond the MS Learn certification-prep courses, and into "building it right" dataafterdark.tips/2024/12/1…

📝 Myths, Magic, and Copilot for Power BI 🔗 data-goblins.com/power-bi/co… ❓ Explaining and evaluating #Copilot in #PowerBI and #MicrosoftFabric ❓ How it works, what it costs ❓ Walking through three different scenarios and asking "why would I use this"? A 🧵 with a summary:

Shoutout to the #PowerQuery UI/UX teams for adding in this amazing error message when pointing to Excel docs in the web!! Known issue for a long time, but now a CLEAR and EASY TO UNDERSTAND error message! and shoutout to Matthew Roche for this old gem: youtube.com/watch?v=kpRrK89t…

If you have already or going to install May 2024 Power BI Report Server update - you need to create environment and system variables if you are using direct queries to connect to your servers. I have written a blog post with the steps here dbanuggets.com/2024/07/31/ma…

Working in #PowerBI is always such an adventure of finding new features, whether they're actually "new" or just "new to me" Today's highlight: refresh all datasets in a workspace via deployment pipelines. Has this always been a thing?

New Repo! This time is DAX queries. Imagine you want a list of all your field parameters? github.com/bernatagulloesbri…

𝗖𝗼𝗻𝘁𝗼𝘀𝗼 𝗗𝗮𝘁𝗮 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗼𝗿 𝘃𝟮 is available. Stop using ancient crappy data for your next demo. 𝗥𝗲𝗮𝗱𝘆-𝘁𝗼-𝘂𝘀𝗲 𝗳𝗶𝗹𝗲𝘀: CSV, Parquet, Delta Table, PBIX, SQL Server; from 10k to 100M orders; fresh data (2014-2024). 𝗖𝗼𝗺𝗺𝗮𝗻𝗱-𝗹𝗶𝗻𝗲 𝘁𝗼𝗼𝗹 for Windows, macOS, and Linux to generate customized data distribution and/or different data volume. 𝗢𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 𝗠𝗜𝗧 𝗹𝗶𝗰𝗲𝗻𝘀𝗲 for C# code that generates files. Read the announcement for links and details: sqlbi.com/blog/marco/2024/07…