Where should we Use AI With BI? – Ep. 251

AI is showing up everywhere in the analytics stack, but “use AI” isn’t a strategy. The real question is where it reduces friction without adding risk.

In Ep. 251, Mike, Tommy, and Seth talk through practical places to apply AI in BI—especially in the messy middle of projects (requirements, definitions, documentation, and iteration). The theme: use AI to speed up thinking and drafting, but keep humans accountable for truth, security, and business definitions.

News & Announcements

• Power BI September 2023 Feature Summary — A roundup of the month’s Power BI improvements and release notes worth scanning before you plan your next rollout.
• Fabric capacities: everything you need to know about what’s new and what’s coming — A clear explanation of capacity options and what Microsoft is signaling for Fabric’s near-term direction.
• Tips+ Theme Generator — Standardize report look-and-feel (colors, typography, visuals) so your org stops shipping a new “design system” every dashboard.
• Subscribe to the podcast — New episodes, show links, and the full archive in one place.
• Suggest a podcast topic — Send a mailbag question or a scenario you want the crew to break down.

Main Discussion

A useful way to think about AI in BI is: AI is great at drafts and options; terrible at accountability. It can help you brainstorm measures, rephrase requirements, generate documentation scaffolding, and explain patterns—but you still need governed definitions, controlled data access, and validation.

Takeaways you can apply:

• Use AI to clarify the question, not to invent the answer: start with “What decision are we trying to make?” and “What would change our behavior?” before you generate any visuals.
• Treat AI-generated DAX as a first draft: ask for multiple approaches, then validate with test cases and known totals—especially around filter context.
• Automate documentation and naming consistency: AI is strong at turning existing model metadata into readable descriptions, measure explanations, and field lists that stakeholders can actually use.
• Keep governance and security non-negotiable: if the tool can see sensitive data, assume it will leak unless you’ve explicitly designed for isolation, auditing, and least privilege.
• Use AI to improve communication: have it rewrite technical findings into executive language, or turn a messy meeting note into a crisp set of requirements and acceptance criteria.
• Build a “verification loop” into your workflow: every AI suggestion should end in something checkable—numbers, rules, sources, or reproducible steps.
• Don’t skip semantic modeling: AI can speed up building, but it can’t rescue a model with unclear definitions, duplicated logic, and unowned metrics.

Looking Forward

Pick one safe, high-leverage use case (documentation, requirements drafting, or DAX prototyping) and standardize it with guardrails—then expand only after you can prove it improves quality, not just speed.

Episode Transcript

0:30 good morning and welcome back to the explicit measures podcast with Tommy hello good morning happy Tuesday gentlemen happy Tuesday it is a great Tuesday dude we are now in the the full Crux of like school right for your kids so I I’m still relatively new to the school game with kids we’re getting really into the thick of it how do you guys separate the office area or where there’s where their desks are because

1:01 there’s where their desks are because right now I have a first grader and a Pre-K and there’s more and more need of not having a thousand papers on tables trash can speak cool okay so no I have you guys are much more organized than I am in the physical world my wife is so yeah right actually I I am don’t tell her that wear it out to the world good thing she doesn’t listen to the podcast it’s a good thing you can guarantee our wives are probably

1:33 you can guarantee our wives are probably not listening to us talk more they’ve had enough of it at home so I’m sorry go ahead organization of papers is a problem because especially towards the end of the year stuff comes back in like droves we’re killing full forests at school these days and bringing everything home and I’m not seeing I’m not saying like the Thousand E’s that she’s drawing is important

2:06 how do you set up like do you have the their desk in another room do you do it in their bedroom their desk area like yeah their workstation their own work station they they both have a desk set up in their room in their room okay see I didn’t want that okay so we have desks right no we don’t we didn’t put desks in our room though we put desks in another room of the house so there’s all three of them have their desks together and you’re right like it every couple days it’s like it’s like a snowstorm right it’s just piling up piling up piling up piling up and then someone

2:37 piling up piling up and then someone comes in the room and says clean this stuff up and then everyone cleans things up and it figures itself out but our kids are a little bit older than your kids so yeah I’m prepping because my girls are so good and I’m raising and my son he’s I’m apparently I’m raising a gorilla like I don’t like he’s a good-hearted Soul he’s a good kid but man like just maybe it’s going from two girls straight to a boy and he’s different all they’re very different yeah and there’s just things I’m like okay that’s gonna break a window that’s gonna break a window so I’m trying to plan for that

3:08 window so I’m trying to plan for that yeah yeah I don’t know so it’s interesting that you didn’t have a follow-up question right because you it’s two things where do the kids do their work and where does everything pile up and how do I change myself my kids my kids have desks in their room they don’t use them that’s true everything is oh everything is done on the island kitchen island and that is just where things accumulate until until someone comes through and swipes it across into the corner of the kitchen

3:39 across into the corner of the kitchen where that piles up probably for the week at least and then things are discarded so what you’re telling me there’s really no hope they’re pot man it it just I think it depends on where your piles are as parents and then periodically you clean them up that’s a great topic there’s another you were you’re also commenting on like the kids practicing their letters over and over again I found I found or something for my daughters they like the idea of having like a little notebook notebook so something that was not just loose

4:09 so something that was not just loose paper everywhere so we got them like a smaller notebook thing this was in paper with some lines in it and then my youngest daughter just sits there and she was at night when she before she goes to bed or she’s not tired yet she’ll just sit there and write or practice things there so she was actually really diligent around practicing her handwriting that seems to work because it’s kind that seems to work because it’s like a fun little extra thing and of like a fun little extra thing and that way you’re not practicing just random pieces of paper but you’re practicing in a place where it’s collected at least so that would be one one trick or

4:39 so that would be one one trick or tick tip or trick that I would give there there I’m trying to think about like like where do we put things in general so with their desks came like a pull out drawer and with that came like a little shelf organizer so we bought some Ikea stuff stuff it’s Ikea and so that helped out a little bit as well just because then we have like okay the papers go we had to like help them label things a little bit like okay the Small Things the rulers the pencils they go here right the papers and things go here right it’s still a mess but at least

5:10 right it’s still a mess but at least there’s a door on it so you can at least shut the mess out a little bit too so that might also help our schools your school like ours have and maybe maybe part of this has reduced the amount of paperwork like they they are really diligent around the binder you guys have this or no we have our school paper we just use RBI for everything I built a power app for my child what are you talking about we only use powerapps the binder is a big thing in our in our school yeah a lot of the folders a lot of the loose stuff gets

5:42 folders a lot of the loose stuff gets consumed into that thing best notes I’ve taken in 10 episodes by far maybe I need a three ring binder I do too yeah to print off all my P all my yeah reports I’ll print them all out yeah I’ll put them in my binder so I can email them to myself and print them on my printer PBI binder yeah better yet binder Jupiter notebooks that that’s a huge yeah there we go right there yeah nope nope yeah on your Jupiter notebooks works

6:12 yeah on your Jupiter notebooks works until you have an error and then you get like 60 pages of error because the the error message is just so stinking long stop being so efficient slow down and print that’s what our topic is about today right right excellent It’s a couple a couple internet updates across the web here so one of the ones I’d like to point out here is an article that came out from the Microsoft fabric blog and this is talking more around the upcoming not changes in fabric but there’s we’re

6:43 changes in fabric but there’s we’re getting closer to getting to General availability of fabric as it’s being rolled out and with that comes some additional charges some other features they’re doing a lot more explanation as preparing you for what will be coming inside sounds like they’re trying to release it in October based on this article message I’m not really sure if that’s the actual go live date with a lot of fabric things but we’re getting close at that point with the Microsoft fabric blog so I’ll put that in the chat here any reactions to this one did you guys get a chance to read this one I ripped

7:14 get a chance to read this one I ripped through parts of it and like the the capacity metric like how how and where you’re using things are coming through here that is pretty exciting because I know that’s been a big Boon of like I agree where is this where is this happening how much how much is it going to cost and mine that’s not part of this right to your point though also I’m also looking at the the one Lake storage reporting in capacity metrics that’s going to be very interesting and my

7:44 going to be very interesting and my initial thought came with this was cool I’m able to make Delta tables or Delta tables from Power query I’m able to make Delta tables in my Lake I can I can load it in there no problem what happens when I want to clean up that data and one thing I’m not very clear about right now is inside a notebook for or inside a dataflow’s Gen 2 right does the data flow Gen 2 automatically clean up old RK files when they’re not being used my suspicion is

8:14 they’re not being used my suspicion is no because I have not seen any documentation on this so the idea being like let’s say for example you load a series of files every single day all those files are generating new parquet files because Delta is is an immutable file format parquet files are immutable you don’t create new one you don’t update old ones you just create new ones and so they update these files so as you edit every time you edit or optimize or do something to a table if you’re making transactional changes on a table you’re

8:46 transactional changes on a table you’re just creating more files which is fine but in data bricks what I’m familiar with is there’s this stuff called optimize and vacuum so optimize takes the files that you have consolidates them down to a single amount and then the vacuum step actually goes through and removes the files so as you’re making new data the old data or things or the old files that you’re no longer using anymore they slowly disappear disappear yeah but that’s not what vacuum is doing vacuum is just cleaning up like old versions of things like you need a different step if you’re like I have a data threshold I no longer need to keep

9:17 data threshold I no longer need to keep yeah yes correct stable no but yes but It’s it’s a lot of those vacuuming mean it’s it’s a lot of those vacuuming of the old files that are no longer being referenced by that there’s no yes but think about like no fabric isn’t doing that right out of the gate like I don’t know any other more of a mystery to me you have to you have to manage well okay so that’s your question yes are they going to manage your storage so it doesn’t explode or they’re just gonna let it go crazy can I get a button on

9:40 let it go crazy can I get a button on that here’s my threshold right just keep things clean for me and and I can do that a little bit inside I don’t know what I don’t know if if notebooks have a vacuum equivalent step yet but I do want to vacuum that in my my notebooks but data flows what about data flows what about that how does that work because I can log data in through a data flow there’s no my from my understanding there’s no vacuum stuff today I’m personally intrigued on the visual side when it comes to the new capacity

10:10 when it comes to the new capacity storage and how they’re showing it because right now you can explore that fabric capacity’s usage but it’s per I don’t want to say it’s cryptic but it does not I think from an Administration point of view both from if you’re looking at cost and just saying what’s what in the world is happening in my tenant it’s pretty difficult to have a pretty good idea like walk in jump in and get what you need so some of the things I’m looking at from the explore graph to the drill through like I hope that from the visual side out of the box they give us what we need

10:42 the box they give us what we need from both where are artifacts and then what’s going to use like if there’s a data flow that’s one particular data flow now all data flows that are taking up consumption and when they are and who’s who’s in charge of that very similar to the premium capacity app that I am looking forward to I’m I’m I’m cautiously optimistic around what you say there Tommy because I’ve already seen how well they’ve done reporting on the existing power bi tenant yes and then now we’re adding

11:12 tenant yes and then now we’re adding a whole bunch of other fabric capability pieces here and this is this is been my one hesitation with fabric since day one is great we’re getting a lot of new features and capabilities but do I really have the ability to administer and govern it and tune it the way I need to if we have unlimited money sure we’ll just turn everything on let everyone run wild no problem but I’m not going to do that and organizations are going to continue to cost optimize their environment they want to make sure that it’s running the right things at the right times so I’m not sure if we’re going to be there yet

11:43 not sure if we’re going to be there yet day one day one yeah there’s go ahead there was a key piece in here that that there is a blurb that we should lean into which is no you’re right September 30th is the deadline for in the preview period for basically for free capacities done yes whoa so starting October 1st workload usage will count against capacity limits and capacity limits will be enforced yes

12:14 and capacity limits will be enforced yes so starting October 1st you’re gonna see just how mad actually dive in right now like if you’re using stuff right like you’ve got a you’ve got to understand how much how much you’re spending in these capacity metrics right before it turns on and that becomes part of your bill so to reiterate Seth on October 1st anything you’re doing in fabric you’re actually gonna have to pay for for that is what I’m reading so this is this is also my question as well before wow

12:45 is also my question as well before wow well and and so this is I just read that and thought okay when I go into fabric right now as I look at my fabric today it still says 59 days left on my I had mine too so does that mean 59 days goes to zero days and 10 days or does 59 days go to like 58 days on October 1st I’m not sure exactly what that means yet so there’s a little bit on not Clarity there yet or do I need to spend three hours deleting a lot of workspaces or

13:16 hours deleting a lot of workspaces or that yeah that yeah [Laughter] anyway lock into the article goodread yeah very good read it out I think one I think another couple notes they make here is about is one security that’s not even out yet so I don’t even know what that looks like I don’t know what that means means it’s coming with fabric it’s noted in their little image at the top they have a single security model which I think is the reference to One Security so I’m interested also to see how that’s going to roll out as well because I don’t think we actually have a full picture of how that’s going to be interacted with

13:46 how that’s going to be interacted with inside fabric so that’s yeah I I have to mention also part of the article is probably like what they would want us to talk about right in the beginning is bursting and smoothing so they’re there yes some detail about how the the fabric capacity is going to work or smarter than that which is indicative or or just we saw right through your shenanigans there scroll to the bottom I will say if you lead with bursting and smoothing there’s a reason why you’re leading with

14:17 there’s a reason why you’re leading with bursting and smoothing I’m just going to say that we only leave that one there anyways anyways the only other noteworthy note here for openers would be Power bi desktop has now updated for September there’s a couple release notes around that I think most of the improvements around power bi right now have been focused a lot on fabric they’re getting that outdoor out the door and also with a now October 1st due date for things to go real real there’s probably a lot of feverish

14:48 there’s probably a lot of feverish pace of building of things and code at the Microsoft level so we didn’t really see a ton inside the power bi desktop there’s a couple modeling changes you now have the mobile layout interactive canvas which is something I never use anyway so hope you do thank you for building it but I don’t really need it so anyways there is a upgrade for Microsoft desktop rbi. net 7. 4. 72 and newer I was really looking for that feature I was very very pleased and happy I was wondering when we

15:18 and happy I was wondering when we were going to hit at least I personally won my power bi Bingo for the year with that so anyways not too much going on all right let’s jump into our main topic today so this is actually a carryover topic from I believe it was last week on Thursday I think if you start talking about this got a little bit into bit into where should AI be applied and I’ve been actually actually LinkedIn knows me right LinkedIn knows I click on all the articles that have ai related things in there and if there’s a

15:48 related things in there and if there’s a chart and AI bolted together I’m really interested so so LinkedIn knows I like charts and AI all in the same stuff and I scroll was scrolling down through Linkedin and I saw an article that said companies are now doing studies around around the efficiency of employees and whether or not they use or do not use AI in their workloads or either developers right so how efficient is someone if they have or do not have ai or supplementing AI to support their workload

16:19 workload and as I was reading the article it was it had two bell curves and they had done a a run against many different employees and we’re seeing how long it took them to achieve certain tasks and all the employees that were either self-taught or had been trained in AI or have been learned how to be prompted using prompting to get the interest they need they were statistically more efficient than the employees that did not use that so it feels to me like there is this rowing

16:51 it feels to me like there is this rowing knowledge base around yeah you may not have ai built into your programs but you should start trying to use AI to do different things for you and I’m finding personally I’m getting better at understanding what prompts I need and I think of anyone in this in this call or this podcast Tommy is probably the most AI person I know right now Tommy’s got like libraries and libraries of things and he’s really studying it and trying to figure it out so I’m I need to learn more about where I can

17:22 I need to learn more about where I can best apply the AI inside what we’re building whether it be reports whether it be data modeling whether it be measures for like I think everyone thinks oh yeah AI can write tax measures I think it does a lot more than that now and we should be think we should open our minds around the space around where can AI be applied inside the bi realm you’re seeing a lot of services now too if you watch any football game now there’s like 18 commercials on how’s generative AI can help your business

17:53 generative AI can help your business Watson X and all these things and there’s a whole lot of fluff with not a lot of work like content yeah so we’re seeing there’s two areas Mike that you highlighted there’s the independent the the individual user the micro sure impact of this and I think that organizational impact I think we can focus from the bi point of view where there is a major significant difference of utilizing AI in both aspects aspects yeah I think there I think there’s a you yeah I think there I think there’s a how do we I think there’s actually know how do we I think there’s actually a strategy around both of these things

18:24 a strategy around both of these things right so we talk about me as a user user how do I interact with an AI thing where do I best use it and then we can step back and say okay let’s look at what the corporation is doing with AI what how are they using the corporate aips here and I think this is where we got really excited last time we talked about this was we need AI to help us design the requirements of report what if it what if AI could capture the requirements or do some of these more mundane things inside this and I I’m actually very Pro

18:56 and I I’m actually very Pro why can’t the AI document my data model right I saw a I think this was a while ago a tabular editor script someone was using chat GPT and they were sending a sending a the Dax formula so it was in tablet editor someone had written a call I think this was daring gospel maybe who did this one but he basically had it okay for each measure in your model send the measure to chat GPT and explain the

19:22 the measure to chat GPT and explain the measure and what would come back would be a paragraph or a little a two or three sentences that explains the business logic of what this Dax statement was doing and so what was coming back was a human readable explanation of what was going on in the model which I thought was genius because what is the one thing you don’t do when you’re building a model you don’t spend a lot of time thinking about how does how would someone consume this measure what does it mean and how would I give an explanation as to how it works yeah I

19:52 an explanation as to how it works yeah I could just figure it out I would be interested to know for you guys personally has AI seeped into any of your workflows I would say yes my non-technical workflows especially right so what by that is yeah being being in the position I manage a team I have a lot of strategy cameras et cetera so like interviewing people with job descriptions roles and responsibilities like it it in

20:23 responsibilities like it it in especially in use cases where it’s like I I need a new role written like job description of like Hey we’re gonna like upgrade you or we’re gonna promote you hear your new responsibilities and then it’s just like the bullet points here are the new roles and things and it’s like create create a what I would like you to do is write this you would like you to do is write this format in the form of a business know format in the form of a business letter but it just cuts down on the a lot of the administrative type activities that you you any workflow it’s not necessarily business intelligence right

20:54 intelligence right but a lot of the administrative stuff at least just gives you a framework that you can edit quickly right because it’s not going to be perfect you’re not going to like copy pasta something into your business but at the same time like it absolutely is a something that speeds up that workflow and reduces your time spent in trying to word things appropriately etc etc the other the other areas I think you already addressed right related to business engagement

21:24 related to business engagement right trying to figure out like what what a form looks like for requirements Gathering Gathering like what set setting up processes like give me ideas of like what is what should my business intelligence workflow do or if I here’s my input parameters here’s what I’m trying to do like just give you cycle things that you may not spend a lot of time on that could exponentially help your your workflow another way to do that is just listen to our podcast a long time like we’ve been through all these things of like how do you create your workflow but at the same time like it it’s a to me it’s an idea

21:57 time like it it’s a to me it’s an idea generator right like I I know I want to do this thing I have three of the steps but are there other things I need to be aware of and and that’s where I found a lot of usage so far that’s a really good point I didn’t I don’t think I’ve used those as far as that industry workflows I will say this I built an entire website using AI oh that that became School you did that’s right yeah the website fabricators. tips that’s the one that we we it’s just a fun little it

22:28 we we it’s just a fun little it started out as a fun little experiment I was like I wonder if I can build a website and so I went to I bought a domain name fabricators. tips I said I went over to chat TPT and said okay I just literally fat fingered build me a website around being a fabricator someone who works in Ai and bi and does all these things and like literally it spit out all the texts I look great and I it the the website started as just a letting AI generate most of it now I did use an editor so it was like a oh oh what you see is what you get

22:59 see is what you get editor and I use that to like take the code or the the words that was out of the one website and dropped it into the website but then what happened was I got excited about being a fabricator after I had a website for it and then I made like a sticker for it and then a sweater for it and then I like okay I went back and I added other things to the website so now it looks more like a perfect and Tommy and I did learn Fabric and we have a YouTube series video on on the playlist so I added the playlist right there in the web so it’s it’s been slowly enchancing itself to do some different things so it was more like an

23:30 different things so it was more like an experiment but I was like oh this is this wasn’t that bad and it sounds kind this wasn’t that bad and it sounds interesting and I and I did words of interesting and I and I did words with some of the words so it made more sense to how I felt I wanted the website to look and feel but to your point set that was a great for like the initial idea the generation a lot of the initial code and then I just like you code and then I just like clean it up a little bit and you know clean it up a little bit and you know clean it up a little bit and fabric Fabricators are made know fabric Fabricators are made possible Life by Microsoft Fabric and the letter F that’s that’s the ending of the website so that’s amazing let me give you one I think I don’t want

24:01 let me give you one I think I don’t want to put the expectations high but this may blow your mind I do a lot of Discovery calls on understanding departments what data sources that they use what are they how do they utilize them what’s their process who do they share it to where do their what are their problems like where in terms of their current workflow how much time do they spend and I have a pretty like solid agenda at this point in my life doing doing so many the problem is you have to follow up and then write that Discovery recap especially when it’s happening a lot well since I

24:31 when it’s happening a lot well since I knew I followed this agenda always kind knew I followed this agenda always to the T in terms of it’s the same of to the T in terms of it’s the same understanding for a department or organization I they actually created a pretty long prompt saying I want you to take this transcript this vtt and I’m going I want you to spit back out for me this format I said write me the data sources like what do they use thing and in this format The Prompt is long but for anything I upload comes back perfectly you would never believe how how much content AI gets out but also the quality

25:03 content AI gets out but also the quality and right to the teeth here’s the data sources they use SharePoint they have problems with SharePoint here’s the process that they use when they use SharePoint and it breaks down each data source based on just the conversation and we talked about but this is good though this is what I want this is awesome this is this is where I think AI should be being used is like if we can and I think to your point Tommy you change your process a little bit by having here’s my prompts here’s like you you have a format that you can use and

25:33 you have a format that you can use and so the idea is like you’re using the AI to summarize something that you’re talking with verbally so we can have the normal natural conversation and out of that pops an automatic piece of documentation that this is exactly what I think the use case is and I think the more of and this made me think more about our conversation last week we were like how can we create like Discovery or a request reporting requests and the more I think about it like if we as from The Human Side because I think there’s a big part here where AI doesn’t do

26:03 big part here where AI doesn’t do everything and just establishing that right up front all these situations are cool but it’s still very much especially in our space takes our ability to one prompt but also to understand what that team or process is trying to do where I know this agenda that I’m following I know what I need the recap to look like I like that a lot actually and I will share to you privately yes exactly right

26:34 yes exactly right so yeah the question I had though is where do you upload your VT so VVT or vtt I don’t know what it’s called it’s a vtt vtt okay that is a transcript that is generated from a video mm-hmm so is that where does that get uploaded to that way your check GPT can use it is it you’re using like vs code to this module all local baby also local okay that makes sense then so because I was wondering because I was looking like I’m familiar with using chat gbt in the

27:04 I’m familiar with using chat gbt in the web browser and that’s like my main interaction with the AI or that that particular AI that’s coming out excellent okay the the recognize with chat GPT that that like what was it Samsung started putting proprietary codes yes but so like it is not it’s not proprietary it’s not your local domain it’s not no going to the channel yeah what it is so it’s it’s good that you’re doing that and it completely separate way but I think

27:36 completely separate way but I think naturally people are gravitate towards chechi PT is like oh the solution to and it’s no there are solutions that you can use like for large know you can use like for large language models that are like locked into your domain so look into those as opposed to carrying out your proprietary yeah because there’s Enterprise Bing now there’s actually a GPT for all I saw that the local yeah the ping yeah security around that yeah yeah it’s all

28:06 security around that yeah yeah it’s all big now changing the messaging like oh this chat this this chat session is only for you it’s only for your organization oh so I’m running a company here I understand I’ve gone through the legal ramifications with all this so we’ve yeah checked them no and yeah I’m not questioning you Tommy certainly not doing that but it is it is it is worth noting like in this world I think actually Microsoft’s got to step up on competitors regardless because the level of security

28:36 of security that they have always instituted in their products and are very very cognizant of I think set themselves set them in a realm of of a trusted position when people are like you’re locking stuff into Azure ID like all of this you want to use this okay great it is part of your company it’s not going to be aired out to wherever and that is one of these things that I think with a lot of

29:02 these things that I think with a lot of these large language models or all these services that are out there you’ve got to read the fight and print before you start utilizing them for proprietary things yeah right like like I was saying like I think there’s a lot of usage a lot of things you can use it for that are generic that are you get a lot of free wins around like administrative work like I was outlining but like once you get into the code aspect of things and proprietary information like yeah definitely definitely know what what

29:32 definitely definitely know what what you’re using for one of the questions that I want to follow up with regarding this like how how have you guys seen this seep into organization it’s because it is it’s a major topic I I would think across many especially mine right like how do we how do you use AI or large language models to to impact I think the production of the business or deliver a better product or whatever the case may be is is that is that seeping into

30:02 is is that is that seeping into conversations or is like like are you guys engaging with customers that are talking around it on the periphery of their reporting Solutions or stuff they want to do with data well I think we’re in the beginning stages of that where that that’s well yeah well yeah but like really early stages I personally have not seen any organizations heavily utilizing or even correct I would say it’s like the initial seed right now in terms of where it could be utilize I know some organizations are I haven’t worked with

30:33 organizations are I haven’t worked with any that are utilizing that from not again not just an independent point of view but our example of hey we’re going to do our recap our report request through AI where it’s the users interacting with a service at the organization it’s more we’ll provide this to you for your independent work I’ll say where I see Chachi PT or where I’m where I’m seeing AI being utilized at least initially right now for organizations I think it’s a lot around the developer so I think if you look at

31:04 the developer so I think if you look at the the spectrum of people that are willing to start leveraging AI unless it’s if it’s not built directly into a product like I’m in PowerPoint and it’s AI generating slides for me or AI generating information yeah it’s it’s typically not designed like it’s not nor business users are not immediately picking it up or I would say I’m finding most people using it right now is it’s in the developer community so the developers

31:34 developer community so the developers are the ones that understand how to write code and I heard someone say you write code and I heard someone say likening the prompting experience know likening the prompting experience is like writing code or writing programming language but using just straight regular language so the developers I think are already understand what’s going on there and there’s a lot of these little things like like stuff that just falls out of your head like hey I need to write a regex that parses this thing I need the second element of this regex element right so it’s the little tastic that

32:05 right so it’s the little tastic that like yeah I could figure it out but it’ll probably be easier if I have it do this or hey can you write a function that does this Tommy and I were doing learn fabric yeah we were we’re using learn and this would have been a great place where AI should be put in place but Tom and I are using fabric we needed to write a pandas function that would create a new column and so Tommy basically said using this and this and this create a new column by multiplying this column times that column and it came back with a very close to the right

32:36 back with a very close to the right answer and we were able to interpret the answer and then rewrite the language using a Lambda function we’re able to build out the Lambda function that we needed in the python notebook so it was very close and the other thing that is interesting here with that as well is okay maybe you don’t understand what a Lambda function is once it Returns the answer you can say explain this answer to me and then it will it will go through and it will read that code that it wrote you and trying to explain it in a different way so I think again that’s

33:07 a different way so I think again that’s my experience so far where am I seeing Ai and things being actually used inside organizations I just paid for the Microsoft teams premium so you can buy teams premium and with teams premium every meeting you record now comes with a full transcript and chat stuff and there’s a whole bunch of extra things that go along with it and what you’re seeing now and I think this is where it’s going to get really interesting is to your point Tommy right take the V BVT file and say summarize right it’s already it’s now

33:37 right it’s already it’s now automatically doing that for every meeting so you go into a meeting invite you now have who talked when as a bar as a line chart and then you have the actual transcript and then you have meeting minutes or summary of that meeting as well right inside teams so so I want to lean into that one a little bit though like there’s there’s there’s ethereum thinking of like right away where and I forget what it was it’s not a

34:07 and I forget what it was it’s not a theory it’s it’s a practice where essentially when you’re taking notes when you’re writing something down you’re more prone to remember it than not agree and to me I do the same things when note taking and I’ve developed I think very proficient note-taking skills over the years as I engage in any meeting because I need I need to make sure that I understand in common right like three weeks later what we talked about about yep do you find though that

34:38 yep do you find though that the like using AI in these cases is just CR is creating value or is it cheapening that experience or or disengaging people further to to the point where they’re not gonna retain or not understand like what were the key points to them instead of an AI engine going oh these were the key points of the meeting that’s your methodology I would wholeheartedly describe that it cheaps out disagree I disagree that it cheaps

35:08 out disagree I disagree that it cheaps out the what you would retain without it and I think it’s part of the understanding of what AI is doing if I’m just relying on that going oh AI will take care of it yes cheaps out I still write notes during it I know though I don’t have to rely on every single thing that’s set more importantly what the AI does is provide a little it provides great structure both from Mike’s Point with the development and the notes that come out where it’s like oh yeah I remember that well let me go back to the recording or let me go back and yeah and

35:38 recording or let me go back and yeah and I can review that with my notes we’re saying here are some of the action items that were talked about here’s some of the Summers I guess I remember that I go back to my notes compared to me when without any recording you’re going oh my gosh I need to remember that let me remember that in a certain way I I get a lot more spit out information it’s the same exact thing in my head when it comes to to when I’m writing it for like the Lambda function it didn’t give us the right answer but it sure as heck gave us the right structure of an answer it was like oh I see how it’s Lambda rho column and I can then begin to format

36:10 column and I can then begin to format that’s how not just that one line of code needs to get output but anytime I’m adding a new column it’s the structure okay now I understand that it’s the same thing right now with the notes conversations and and again from the development too where it’s providing that structure I still have to do work if I’m just relying on it it’s not going to help but just allowing knowing that’s almost like a supplemental side has been is that has been the Catalyst

36:40 has been is that has been the Catalyst from a value point of view yeah there’s a couple chat comments around this as well and I think I think that’s a good it’s a it’s a good different perspective too right because I I I started the comment with I’ve I’ve it’s taken years to perfect how I want to do that but in in most cases I think that is actually a challenge for folks to consolidate a meeting with the appropriate bullet points and if it’s doing that for them because it that’s a difficulty then that makes it like there’s a ton of value

37:11 makes it like there’s a ton of value there so there’s there’s also actions that come out of there too yes okay which I do like right so go ahead sorry no no I’m sorry like the the outcomes of a meeting yeah right because that is one of the things that I would hope is in every meeting I’m attending is there’s action items for somebody to do something because that’s why we’re getting together and I have seen even when when the folks on my team utilizing that is like a bullet list the the challenge or not

37:43 a bullet list the the challenge or not challenge like the thing I would look forward to is like an integration at some point where if there are actions I want to point I want to push those into the system of where I keep track of actions right or build them out like jira or something like that oh this is okay this is awesome that you’re saying these things so in the Microsoft teams AI stuff right so so let me say this way I’m not going to stop taking notes right when I’m in the call and I’m listening to what’s happening I’m I’m actively listening by focusing on what you’re

38:14 listening by focusing on what you’re saying and I’m taking the notes that still happens like normal however there were things that I miss and so as as one who’s letting the AI also listen to it it’s like having a second set of information that’s also appending so to your point Seth like is it going to weaken my note taking abilities in a week in this no I don’t think so for me personally no because I’ve already formed a good pattern around this now new users of or or young employees of companies May lean on the AI too heavily

38:43 companies May lean on the AI too heavily potentially they’re not actually forming good habits around listening to what’s happening in the meeting however what I would say is not everyone always can attend all meetings and sometimes you’re catching someone up to a meeting so the other part of AI that I think really is very relevant here is you use the team’s premium it has this little area called notes and AI notes so right Microsoft is integrating this thing called Loop now into inside their meetings an inside loop you can add follow-up tasks so I can actually add

39:13 follow-up tasks so I can actually add tasks there it goes to your to do and it automatically synchronizes them across everything so now what I can do is I can go back to the meeting I can look at the AI notes and the AI notes is doing something interesting it’s almost doing like topical summaries when I so as I take notes someone says a statement or a sentence and I write down okay I need to do XYZ thing I need to build this in the model right mine are very like mechanical and very Tactical for what we’re seeing there at the very pointed element

39:43 pointed element the AI is is taking like an example of their AI notes it’s taking like four bullet points Mike said something customer said something Mike said something Mike said something it’s taking like multiple statements and saying okay I’m going to take these four key points and summarizing all of them to say we talked about model optimizations certain visuals are running out of capacity here’s what we’re gonna have to do to fix this and they talk about solutions for fixing that problem so it’s more of like an you

40:13 that problem so it’s more of like an you that problem so it’s more of like an a thought as opposed to just single know a thought as opposed to just single things right so it’s it’s those items that are coming up so yeah I will say this for me I think teams premium from for where I’m spending it right now is probably worth it because there’s things that are moving so fast and particularly in the Consulting space it goes quick for every organization to add every user with a team’s premium I’m not sure it’s going to be there and where I’m worried about some of this from an AI perspective is how many different AI things do I need

40:45 how many different AI things do I need to subscribe to to pay a couple bucks per month to get value out of them right so this to me there’s another part of this too we’re saying like am I paying ten dollars to chat GPT am I paying ten dollars to Microsoft teams am I paying five dollars over here so look at the end of the day you’re paying 100 a month for all these random AI generation tools to do things for you to help you automate is that you have to think internally how well will that summarize what you’re trying to do well but at the same time like so I was

41:16 well but at the same time like so I was looking and I tried to find it but I couldn’t I saw a LinkedIn post yesterday where I think it was a consulting firm that had done some analysis across a team of users I don’t know much more detail around like what it was they plugged AI into but what they saw was there was a they they basically did to maybe testing here’s a group of people here’s the output that they had based on the the work items that they do here here’s a group where we taught them nothing about it and their quality and

41:49 nothing about it and their quality and like slightly less quantity but the quality of work or was it both I think let’s just say better quality better quantity quantity and then the people that they did teach like about how to utilize AI were even better than that right so if that’s indicative where I see this going is things like the team’s integration yeah of course all this is going to cost more money right but if there are if there are proven methods to show that this 10 bucks a month is

42:22 to show that this 10 bucks a month is saving you like a hundred dollars a month because now all these people are much more efficient than they were before before there’s a lot of value there right because people would agree I I could say hey there’s a great use case here what I would like to do is I’m a heavy Microsoft user I use jira if I can take a meeting dump out these notes and Consolidated summary of This plus actions and I know this is going to turn into a work item I want to like let me select the things and then it just goes and creates a work item which gives me

42:52 and creates a work item which gives me the framework to go add the specific requirements at some point because it’s the what I call like just the I’m trying to find a nice PC word here like it’s the wasted time of like I have to go create this virtual space to create it I have to create the description I have to do all these things is it like man if you simplified that that’s 15 minutes of my time boom I I need every time I have one of these meetings right so that’s a huge Time Saver and I I think that if there are

43:24 Saver and I I think that if there are use cases like this where it’s it’s just this verbose activity that we have to do every single time and that that’s definitely worth the money and if you like multiply that by team multiply that across the company and all the dev teams and all the managers that are product people that are creating these work items like yes of course building the framework and allowing me to automate a lot of the construction of oh this is coming from Seth he needs the these five properties set up on all these things

43:54 properties set up on all these things and this is the project that it needs to go to and like that is not hard to figure out if we’re consolidating entire conversations right the the consolidation of like what’s the configuration to go link into systems that are the the workhorses and engines for review in processes that we already have set up right like and I don’t have to do all that work man that frees up so much time for me a few few quick shout outs for my Take My Money Services

44:28 read AI is like that teams premium on steroids with the action items I’m also I use notion for note taking for milestones for articles it’s like a second brain but also for all stuff I do with work it has an AI feature you click space it’s basically chat gbt and grain one thing I’m shocked about today I I really want to get to is we’re talking about Ai and bi and one thing we have not if you were to say in the beginning of this where would my money go in terms of where this conversation would lean I would say it’s going to be our coding or development data data data data modeling Dax we

45:00 data data data data modeling Dax we haven’t touched on that and I think it’s very indicative I don’t what I’m trying to figure out is that indicative of the technology where it’s at now or where our needs really are what we see on a daily basis yeah I think Tim to me my comments are based around like what where is AI helping me right now okay right like I don’t once once we start to get implementations of easy use code editors in for Dax and things like that dude we’ll be

45:31 Dax and things like that dude we’ll be talking about that for many many podcasts but at the same time like how do you perfect this how do you make it better like am I excited about it yeah but I I think right now nothing’s integrated much in the same way that we’re talking about teams integration like yes they’ve implemented AI into this thing there’s nothing that that is in the wild that we’re currently utilizing that’s in fabric that’s in you utilizing that’s in fabric that’s in power bi that’s in this that just know power bi that’s in this that just lets me go hey okay let me validate my Dax and is it good is it bad do I need to prompt it do I need to like give more

46:02 to prompt it do I need to like give more details so like that’s where I took this conversation today is like what are the things I’m using right now without a hey I’m copy and pasting and moving this over here and then it doesn’t understand the model and yeah it’s there but not is that is that part of a conversation I guess but it’s not something I’m really excited about at this point it’s not a Time Saver for me yet okay let me let me give you one a real good Perpetual example on this one so you’re right Seth and I think what we’re going to start seeing is we’re going to start seeing AI come closer and closer to where we’re doing real workloads this is this is the trend I

46:33 workloads this is this is the trend I think is happening right teams is one example of this another good example is I actually use data bricks and we use databricks premium inside databricks premium you get a databricks assistant which is dang stinking good it’s literally like a new fly out panel that as you’re writing code in the notebook in the different cells it’s prompting you or you can ask questions about things on the left hand side so you can ask questions around help me find this python function help me write this thing help me help me and what it’s doing is the assistant

47:04 and what it’s doing is the assistant is actually one of the biggest things you do especially when you’re using data bricks at least for me you can help co-pilot yeah it’s it’s like it’s like a co-pilot but it’s more it’s more focused around assisting you writing those notebooks and code pieces right so you’re writing if you’re trying to you have a data frame you’re trying to run a data frame to back to the lake somewhere right the AI assistant realizes is understanding a bit more of your code and it understands here’s some articles that you may want to read that are helping you optimize this thing hey you’re using Delta live tables hey you

47:35 you’re using Delta live tables hey you may want to read this article around here’s a section here’s some summary of that information that’s coming back to you as you’re writing it so you can just quickly glance over or see things you obviously don’t know what you’re doing so here you go use some contacts here’s some context but it’s it’s that stuff it’s it’s the things that and I’ve heard it said to me a number of times right if someone’s new to something you don’t understand the keywords that you need to Google or search for to get the answers you want right so part of this now is the context

48:06 right so part of this now is the context of the code you’re writing the context of what you’re doing can automatically start being presented in this and I would highly recommend if you use databricks today go look up or research what this databricks assistant is doing inside this because I couldn’t also Imagine like you’re writing like one thing that

48:24 you’re writing like one thing that Microsoft is totally missing is there’s a whole bunch of documentation on their learned at microsoft. com there’s a whole bunch of documentation on microsoft. com around doing things in power bi what we need is we need desktop to have an AI assistant in the same way like don’t try to write the Dax code for me try to point me to where things can happen or things can be and the neat part about this would be really helpful would be as a user of power bi or other tools let me Source where the

48:54 tools let me Source where the information comes from that’s a good example of this one Tommy you you pointed an AI at Dax Dax dot guide guide right you point to the AI and said here’s here is the source of Truth and information it’s a website that creates content point to that location and then use it to help me build things or if I ask a question about add columns or examples around different it’s already right there the the website has the links and the code for everything so use that to help me produce what I want to

49:24 that to help me produce what I want to produce let me do you one better with that not only can you have that little Dash script but I can have a page or a CSV on a table saying here’s my table format now point to this file saying this is what my stable structure is point to Dax guide and even other things we utilize externally not just the power wheel PBI tools the command line it’s great but no intellisense in the terminal you got to really look at the description I told it I said create a new project for me create a python or a Powershell module with intellisense around all the parameters around this

49:55 around all the parameters around this PBI tools command line pointed to the article I have all intellisense now for every command in PPI tools so we’re getting to a really crazy point in terms of what’s already available in data bricks what power bi is developed where this is going to be part of the micro workflow of that the personal developer we’re Mike we didn’t have to know the python Lambda function or we in terms of or we didn’t have to we learned it like here we go now we know it here’s what here’s what need to be written in

50:25 what here’s what need to be written in this format for data Wrangler like okay cool like yeah this is the thing data Wrangler needs the same stuff like this is yes and it and it should be like literally everywhere like that this is what I think is going to start happening and so we’re talking about a lot of code based things but I think that makes the most sense to people because these are things that you have to go figure out and learn how they work but I think there’s a lot of need here as well to what we were talking about was was we need to use AI to help us gather requirements around reports we need AI to present hey what we what we

50:57 to present hey what we what we what I’d like to have is hey here’s my report here’s the documentation here’s the model. bim file right hey Mr Mr user of this report right go to my SharePoint site ask a couple questions of that report and what should be happening is there should be a system where I can say that that user request form that comes to me and says hey I need a report that does this this and this that input that question should be asked into an AI the AI should already have

51:27 into an AI the AI should already have all the definition of the existing data models and all the existing reports that are out there be able to scan and read all that the user should ask the question and the AI should be able to say hey did we had this report called sales report it’s it’s located here and it answers these kinds of questions and so the AI should be pointing people to things we’ve already done and work that has already been created created that that way we don’t have to build a new report and if it doesn’t exist the AI should be able to say

51:57 AI should be able to say we have something similar does this meet your needs the user can say no I need some more help I’m looking specifically for these types of calculations and the AI should be able to read that and be able to say oh let me help gather some more requirements for you and then it then prompts there should be a series of prompting that says okay what is this report for okay it’s going to do this what is the audience of this report okay here’s the audience how long would you expect to be in the report okay do you think you need an overview page that does like there should be there should be all this like natural language communication style that says

52:29 language communication style that says here’s my prompt as a bi developer here’s the things I care about talk to the AI and have all that done up front and and spit back out to me and then actually get a summary of that thus the the Crux the Crux oh I was gonna say the Crux of my argument here is all the coding is great cool but I my argument is the reason why we focused on the soft skills in that area I can make the argument and I’m convinced myself from all the people I

52:59 convinced myself from all the people I know in bi all the tools available the largest gap between bi is not a sister is not the coding right now it is that Gathering requirements it is that communication with the business where I see the largest gap and the largest need for AI or these types of services the coding great cool awesome I can still Google it if I didn’t have it but this ability and a lot of organizations they struggle or have a wider gap between their ability to gather requirements

53:30 their ability to gather requirements effectively to gather insights effectively and to communicate back to the organizations effectively more than they do with their Dax or their visualization skills so I see AI having a much larger impact and much larger need than right now the development side I’d agree and it’s just going to be like how how do you build that like so to me right now the question is how do you build that tool like us as developers or us as report creators how can we feed like some the question in my mind right

54:01 like some the question in my mind right now is how can the developers of power bi and fabric level things how can we feed the AI what it needs in order to be more accustomed to what we’re building inside our organization how does that work like to me that’s the unknown at this point like I know it’s needed I think it would be great to have these additional enhancements in this way I think this would be very effective for what I need for a business use case but how can I how could I get a

54:31 what is the what is the virtual assistant right how can I get the virtual assistant program to utilize something that would say okay put my virtual assistant on a page that talks about Pirate Bay reports I’m going to feed the virtual assistant all the information about my model and it will then be able to direct people users questions to what what is happening right I would love to feed an AI all my process documents right I have how to request a workspace how to how to ask for for a new change to my certified data set right these are all processes that we’ve

55:01 right these are all processes that we’ve built I want to communicate them to the AI so the AI knows how to answer those questions so I’m not spending the time to do that so the first the first tier one support should be somewhat generated by an AI system and then only the best stuff of that comes out to me and summarizes things more for me and that’s where that’s where my head goes I’m not sure if we’re if we’re all there yet there yet no I think all the things for me well I my dream my dream is a dynamic form

55:32 I my dream my dream is a dynamic form where they say what’s your request you start filling it out sounds like you’re you’re looking at sales well what type of sales and having that conversation sounds like you’re looking for a data dump please find another company yeah yeah exactly it sounds like you’re looking for passing reports we recommend you read these five articles from Microsoft to figure out how to build your own reports please find some targeted kpis and get back to us thanks click yes close window yeah anyways we’ve done it again we’ve wasted a perfectly good hour of time I think this has been a great conversation I still am thinking there’s going to be

56:02 I still am thinking there’s going to be more conversations around AI where it fits how it integrates with different products or tools if you have any examples of AI areas or AI that you’ve integrated with so I’m actually very interested myself is if you have found anything that you’ve worked on in your company or other companies put in the chat window we’d love we’d love to talk to or see examples of where this AI is best utilized Insight organizations I have a very firm feeling that this is going to continually changing the

56:32 going to continually changing the landscape of how we do business and it’s going to become an expectation that if your tool does not have some AI generative feature to it it won’t help it you you won’t use it you won’t you won’t want to pay for it right if I’m going to pay for some service it needs to have some generative AI piece a part of it that helps me build better things anyways that being said we really appreciate all your listenership everyone has been great the chat has been very vibrant today lots of good comments there so thank you very much

57:02 comments there so thank you very much chat for jumping in and adding additional context and comments with that our only ask if you like this content if you found something of value here if you learn something new please share it with somebody else we’d love for you to communicate with somebody else about the value that you were able to see or or bring from this conversation and this podcast Tommy where else can you find the pack the podcast you can find us an apple and Spotify or wherever you get your podcast make sure to subscribe and leave a rating it helps us out a ton have a

57:33 rating it helps us out a ton have a question an idea or topic that you want us to talk about in a future episode head over to powerbi. tips slash podcast finally join us live every Tuesday and Thursday 7 30 a. m and join the conversation on all of power bi tips social media channels we should have ai generator outro can we do that can we just take we should do that in the music for our AI episodes exactly I guess as a bare minimum it’s a fight yeah interesting I have to think about that one anyways thank you all very much

58:04 one anyways thank you all very much we’ll see you next time

Thank You

Thanks for listening to the Explicit Measures Podcast. If you enjoyed this episode, share it with a coworker and drop a question or topic suggestion for a future show.