I Quit! AI Will Take Over The Power BI Tutorials

I'm considering to quit this power bi YouTube channel I'm pretty sure that most of you have seen examples of the capabilities of different AI tools popping up here on YouTube or on your LinkedIn or Twitter feed and we're basically at a time when you can ask these tools anything and they will give you an answer so when you need a certain Dax formula well just ask an AI to like chat GPT and it gives you the whole formula with a step-by-step explanation or maybe you need a little bit of inspiration for your next power bi dashboard well ask it tool like mid journey and it pops up something like this which looks pretty good so that made me wonder are my power bi tutorials still relevant and do they add any value especially going forward or is it time to quit or maybe it's still a bit early well that question we're going to answer in this video [Applause] welcome to how to power bi my name is boss and if this is the very first time for you visiting this channel then make sure to hit that subscribe button to stay up to date on all of my videos in which I share everything I know about power bi now let's see if I should quit this YouTube channel now there are different AI tools that popped up recently and they got a lot of attention in the social media because not many people were aware of these tools tools especially chat GPT was standing out there and what we are going to do in this video is check them out see if they are any good and if they basically replace my job here okay now the first thing that I wanted to try out is to see if these AI tools can come up with a design idea for my next power bi dashboard now one AI tool that is able to do this is called Journey so let's go to the website of mid journey and try it out now I went here to make journey.com and you get to this website over here now you shouldn't stare too long at it so let's quickly click here on getting started okay now here you have all kind of how to get started and how to create your first AI generated image with mid-journey okay so just read through it but basically you just have to join a Discord group and there you can ask your question okay so let's do this quickly so here we have the link to that Discord group that's follow it and here we just have to put in our username and check the box that we are agreeing to the terms and conditions and click on continue and of course we have to say that we are human all right now let's see if we can pass this captcha so over here we have rough Seas here there there and click on verify okay we are in now here we have to go to one of the rooms for the newbies which we are so let's choose one of them okay and then we can try it out now the whole explanation of how this works you find in the quick start guide but it's relatively simple so let's go to the Discord group and here we can type in forward slash and then you see there imagine okay so that's what we need and then the prompt so now here we can ask what it should visualize so let's ask it to create a dashboard design so I want to have maybe a minimalistic dashboard design for a car business okay now let's see what it comes up with and that took about a minute or so and here we have our AI generated dashboard design for car business and we can have a closer look by clicking on it and it looks pretty interesting so you see nice color schema with the gold and the black but maybe you want to have different variations and you want to have a higher resolution no problem you see over here these buttons below it and here you can ask for more variations of one of the four and if you want to upscale it and have it in better quality you use those there at the top okay but what if we want to have maybe a lighter color background instead of the black silver gold well let's see if we can ask for a lighter design so I'm going to repeat this step so over here imagine then we have the prompt and here I'm I'm just going to copy over the same question as before so minimalistic dashboard design for car business and with light and bright colors and you see we have some interesting designs again now this one is still kind of dark however we also have some light and brighter designs now the one here in the bottom right the corner is very bright maybe not the prettiest ones but it can inspire you with some new ideas for your next dashboard so you see a lot also depends on the question that you asked this AI tool of course here it's also the question can you replicate it in power bi but even that we could also put in the question now let's type in again imagine and then the question which is the same as before has a minimalistic dashboard design for car business but now I want to have a power bi dashboard and maybe we go for not bright columns but just with light colors okay and there you go we have some new designs popping up over here so let's click on it to zoom in and indeed it is a little bit more in power bi style however with the lighter colors did not really work out so well so we probably would have to readjust the question a bit further now this is quite impressive but the question of course is how helpful is it really because well the visualization types that it chooses might not be the most optimal one and that's the same for where they are placed on the dashboard however other Design Elements like the color schema the imagery that is being used how it separates the visualizations from one another well from that you might get some new ideas for the next dashboard and the Beautiful Thing is also you can choose any topic or any color schema right so here we did this for a car dashboard but maybe you work in the shipping industry and then you get kind of a customized design idea for that industry so that is really unique and could be super helpful now of course this is just the beginning stages right so just imagine in power bi we already have the option to automatically generate a report now if you combine this with this type of functionality that is really impressive because then you can just click on create a report make maybe here and there's some small adjustments and then play around with different designs that are AI generated so although maybe now we are just at the beginning just imagine where we are in five to ten years now ai can do much more than providing you with some dashboard design ideas it can also help you write text formulas or write your custom and function now let's pick two of my previous videos and see if we kind of could get to a similar solution using AI I'm first going to use chat GPT to give me a custom and function that calculates the not working days so the number of working days between two different days excluding the weekends let's see if it's capable of doing that now here I already went to the chat GPT website so if you go to open up ai.com and then just forward slash block forward slash chat GPT then you get here and you can just click on try and it will open it up and that brings you here now I like dark mode so let's switch to the dark mode now let's write our instructions to chat GPT I want it to write an amp function in power query that calculates the network days so Network days between two dates okay now let's press enter and there it goes to calculate the number of weekdays also known as Network days between two days in power query you can use the following amp function boom and it starts generating the M code all right it's done and that looks quite impressive and so here we have this whole code that I can just copy over with comments in between of what's going on and well let's just follow the instructions and see if it works okay so over here I'm just copying it and then I'm gonna go here to power bi now here I'm in power creamy I have here a simple data set with start and ending date and I want to have an extra column that calculates the networking days so not just simply had a difference between that and the starting date but at the difference between the end and starting date and then also exclude the weekends and that is what a custom and function is supposed to do now to create a new function go here to new source plan query done to view Advanced editing and this is where we then paste in that code that just GPT generate for us now you see no syntax errors have been detected so that's already quite impressive okay so no errors just yet now let's first see if it works and then we can go over the codes line by line okay so let's click on done and I'm going to give this function a name so this is going to be my net work days function and then we can choose over your date so over here today is 21st I'm going to select Monday over here until Friday next week and then invoke the function you see it doesn't work okay an error occurred in the query expression error two arguments were passed to function which expects only one okay so let me see chat GPT unfortunately failed but maybe we can fix it in a relatively easy way so let's see if this is possible let's go here to our function Advanced editor and then here it starts with the input parameter so start and ending data State that's good then calculate the number of days between two dates so we have the duration and days and date my start date plus one if you want to include the day so that's okay then over here the number of weekends it's going to calculate now the Latin end statement let us break down a query into separate steps so we have here the list of dates with the list dates function that's good which starts with the starting date then over here how many pmas you want to add so number of days comes from there and then here the duration is in day so that also looks good now the next step that is probably where it goes wrong because here we have the list that count function which expects one argument but it actually gave it two arguments and that's not possible so we have to take this part basically out over here and then we have to first adjust the dates that we have here we have the date list that only at the weekend right so weekends how can we do that well over here we need to use a function like list dot select and then we have a previous list so they list from here and we want to have as a selection function well there we could use what we what it just created not let me just copy that over so control movie that is our function there and we need a comma there at the end and what we then want to count then we can count that is this list there okay now you see no syntax errors anymore and over here we can click on done now the function here returns 11 which is actually not correct because it should be done not 11. so we have to make an adjustment to the function now let's go back again view Advanced editor and if you have a closer look at that function date Point day of week you see it returns a number from zero to six and so if we have Saturday and Sunday well this should be not six or seven but five or six okay and because it's null based no it's cling on okay and a function now returns 10 which is the correct answer let's also see if it works here if we apply it for creating an actual column also add column invoke the custom function and then here we have the project Lang in a network days okay now we can choose a function from here then as inputs we're going to have columns start and ending date Okay click OK and there you go now seems to be correct now so we have a function now here if we go back to the function in the advanced editor now here you see that the function that we are using was mostly generated by chat GPT that's open AI tool which is really impressive however it was not working straight away 100 we still had to tweak it a little bit and to be able to make that adjustment you probably need a little bit of um knowledge or how you need to Google a bit trial and error look at the documentation to get there okay but that can be time consuming but the main structure well that was given by chat GPT so I would say pretty helpful now the next question is can we also use chat GPT to write our Dax formulas well let's have a look at an example now here I have some HR data sets and the visualization that I created is a simple Matrix that shows the total amplitude higher Over time however these are the employees that are hired in a certain period And what I want to have is the total number of active employees at any given point in time now let's see if GPT can give me at the right formula so I'm gonna go back and let's start a new chat over here and I already prepared a question for it so the instruction is write a text message that calculates the active number of employees at any point in time where active is defined as being between the hiring date and the contract ending date so you see I'm quite specific the more specific you are the better the answer will be that you get from this AI tool okay now let's run it and let's see what it comes up with and there it goes again here's a sample Dax measure all right and it starts generating some code so let's see now that formula looks already quite promising and you see it contains also with the whole explanation and so this measure uses the calculate function to modify the filter context of the current pivot table that's a little bit interesting and it failed as the employees stayable to only include rows where the hiring date is less than or equal to the current date and the diamond termination date is either blank or greater than the current date okay and then continuous and of course the code that gets generated here also and makes the Assumption on other table names that we probably need to adjust okay but let's try it out and let's see if we can make it work so I'm gonna go to my power bi file new measure and this is going to be my active employees measure okay now here it shows just errors but that is just naming issues right so over here and we can do account on the employee ID or I could also say I'm count rows and then employee table so let's see if there if I have a similar column so employee and then we have amplitude name amplitude number okay so let's go for employee number okay so that's good then we continue over here now every time we have employees that doesn't work so let's select all of them so Ctrl shift L and then my data model in my data model it's called amplitude the stable not employees okay so that fixes that part then it checks over here the hiring date now I don't have a hiring date I have a higher date okay and then here the termination date is the contract ending date so let me also adjust that okay and my calendar is called then date okay so I just make all of these adjustments first and to see if there's any obvious error okay but that seems to be all right now let's have a look at the logic so a function starts with a calculate function to modify the filter context which is necessary is good then here the expression itself that's the count on amplitude numbers there's only one employee row and one row for each employee so we can do that or we could also change it to count rows on the employee table but this is fine and then we have here the filter function that lets us modify the filter that gets applied to our expression over there and this is probably where it goes wrong so over here we have the first argument employee now we still need to adjust that in a second but let's see what it tries to do then it wants to iterate of the amplitude table and only keep the rows where the hiring date is equal to are lower than the maximum date that's the maximum date within the field of context so for example here for q1 that would be the 31st of March 2019 and so the hiring date needs to be before so that is actually really good and for it to be an active employee the contract ending day needs to be blank okay or the contract ending date needs to be after the last date within the field of context well that's actually impressive that it takes both into consideration especially that this one is blank amplitude contract and um of course I tried this example before and generated a little bit of a different code every time it's a bit different and then it did not take this into consideration so I'm quite impressed um and that's it okay now if we make just a little adjustment here at the first argument because here for this to work we still need to wrap the amplitude table inside of all or all selected to remove any filter that gets applied to the employee table first okay and once we have this let's see if it works I'm going to add a two all right Matrix Video Only C starts to accumulate yeah perfect that works so we just needed to do one little adjustment which I think is very impressive and so to be able to get so close just like with the m code before just a little bit of adjustment was necessary but also here kind of the same story to be able to make that adjustment you do need tax knowledge and it is not so easy to spot where it went wrong if you're just starting off with power bi okay now this also goes in a similar direction as quick measures right now we also have the ability there I had to write a question or our instruction and generate different measure Alternatives and so that's kind of going in the similar Direction now the advantage that quick matches of course has is that it knows your data and so on and this openai tool chat GPT doesn't so time for a conclusion am I gonna quit this power behind YouTube channel maybe just not yet I think these AI tools are super impressive and every time they got us to almost the correct answer almost but to tweak it you still needed a little bit of Dax knowledge a little bit of arm knowledge and I I still think I can provide a bit better value there and that's the same for the design part also there it wasn't perfect but I do think that if this is where we are now and over the next five to ten years this gets even more integrated to have power bi with all of these different AI tools well then our role might change a bit now I'm curious to hear your thoughts about this topic put it in the comment section below and if you like this video then make sure to hit that like button subscribe and check out one of my other videos over here now I want to thank you for watching and I hope to see you in the next video

Channel: How to Power BI

Views: 151,343

Keywords: Power BI, howtopowerbi, powerbi, bas dohmen, datatraining, data training, datatrainingio, ai, powerbi ai, chat gpt, midjourney, power bi chat gpt, power bi ux, m function ai, dax ai, auto dax, auto m

Id: 6m2ZGeX-EyM

Length: 21min 40sec (1300 seconds)

Published: Thu Dec 22 2022