How to Use Copilot Capacity – Ep. 393
Mike and Tommy break down what Copilot Capacity is and how to think about sizing and governance so your organization can safely roll out Copilot features in Power BI. They share practical guidance for planning adoption, controlling cost, and setting expectations as teams move from experimentation to production.
News & Announcements
-
Introducing Power BI Embedded with Direct Lake Mode — Microsoft announced Power BI Embedded support for Direct Lake mode, enabling embedded analytics scenarios that query Delta tables in OneLake without importing or duplicating data. The post highlights performance and real-time benefits, automation scenarios for ISVs, and important limitations (like token requirements and identity/auth considerations) to understand before going to production.
-
Add Copilot Summaries to Power BI email subscriptions (Preview) — Power BI email subscriptions can now include Copilot-generated summaries (in preview) for standard subscriptions on Copilot-eligible capacities. The feature adds a lightweight way to push “what changed” insights to stakeholders, with a few caveats around supported subscription types, cloud availability, and report complexity.
Main Discussion: How to Use Copilot Capacity
Copilot features in Power BI are exciting, but they also introduce a new type of planning problem: you’re no longer only sizing for report rendering and refresh—you’re sizing for AI interactions. Mike and Tommy walk through what “Copilot Capacity” really means in practice, and how to approach it with the same discipline you’d use for any shared service: clear use-cases, controlled rollout, and measurable outcomes.
Start with the “Why” (and the actual workload)
Before you provision anything, they recommend getting specific about what you’re trying to enable:
- Are you trying to help report consumers quickly interpret a page?
- Are you enabling authors to generate measures, summaries, or narrative?
- Is the goal productivity for a small creator group, or broad consumption across the org?
Those choices determine whether usage looks like sporadic spikes (a few authors experimenting) or sustained demand (large groups using Copilot daily).
Capacity planning: treat AI like a shared resource
A key theme is expectation-setting: Copilot capacity isn’t “free magic”—it’s a finite pool of resources that can bottleneck if you roll it out broadly without controls. Mike and Tommy discuss practical levers you can use to keep the experience consistent:
- Roll out to a pilot group first (then expand intentionally)
- Monitor usage patterns and performance as you add teams
- Align enablement with training so prompts are effective (and not wasteful)
Governance and guardrails matter more than ever
With Copilot, governance isn’t just about who can see which dataset—it’s also about how users interact with and interpret AI-generated outputs. They stress having a plan for:
- Who is allowed to use Copilot features (and where)
- How you validate AI-generated insights before they reach stakeholders
- How you prevent “prompt chaos” by giving people examples of good prompts and good review habits
Sizing for adoption: the hidden cost is success
If Copilot works, people will use it—and that’s the point. But success changes your capacity needs quickly. Mike and Tommy encourage thinking ahead:
- Build a path from “experiment” to “production”
- Decide what the organization will standardize (and what stays optional)
- Treat early wins as signals to revisit capacity and operational support
Looking Forward
Copilot in Power BI is moving fast—new experiences like Copilot summaries in email subscriptions are a preview of how AI will show up in more places (not just authoring). Mike and Tommy’s advice is to embrace the momentum, but do it with intention: start small, measure impact, and scale with guardrails so Copilot stays helpful instead of becoming the next unmanaged bottleneck.
Episode Transcript
0:31 welcome back to the explicit measures podcast with Tommy and Mike good morning everyone and welcome back to the podcast good morning Mike how you doing oh we’re doing well it’s just the week is clipping along I can’t believe January is almost over it’s been moving by incredibly fast so that’s good I think it’s a good sign things are moving along quickly and I can’t complain but there’s a lot of nice things we’ve hit January typically January is a very quiet month there’s not a lot of noise happening in the Microsoft fabric space or power space
1:01 Microsoft fabric space or power space for that matter but it’s been nice to have a couple announcements come out we’ve had some actual news topics come out here recently but before we get into the news I’ll just do a quick topic today our main topic today will be co-pilot will be insert your word here copilot will be blank to your organization we’re going to just unpack this new co-pilot capacity copilot capacity is a new feature that has just been recently announced from the Microsoft on the blog and we’re going to unpack that idea and just try
1:32 going to unpack that idea and just try and go through the article a little bit just to understand where co-pilot is fitting Tommy I know you’re all up on the AI space have you heard about this rumbling in the in the market here oh the Deep seek okay I knew you would have heard about it so deep seek has come out and so my understanding is deep seek is this chinese-based AI platform that has just recently been released in the last four or five days and already it’s it’s turning people on their heads apparently shockingly
2:03 their heads apparently shockingly there’s been a chip shortage or or a lack of high-end Nvidia chips going over to China but they were able to produce a model that is on par with open ai’s expensive chips for a fraction of the cost to one not only fraction of the cost because they’re using older Hardware but fraction the cost to even train the model it was much much cheaper and much faster to train so it’s very interesting to see that this is coming out so quickly they’re saying this is going to shake up the industry and the AI bubble may be popping so anyways
2:34 AI bubble may be popping so anyways we’ll have to talk about that a little bit more too later on in there as well especially now that we’re GNA discuss a little bit more about co-pilot we’ll us see where this Market is going to shift and where it’s going to go moving forward I’ll tell you what I wish I was Gordon Moore because if I founded the idea of Moore’s Law that’s going to keep proving true which is insane if you never heard of Mo’s law it’s simply this observation that the number of transistors how like how things compute doubles every two years and I think this
3:04 doubles every two years and I think this was something in the 60s like 65 and he’s like the way we’re going with technology is going to keep doubling it’s going to keep getting cheaper to do things it hasn’t been false like and it’s insane for someone to make that claim I feel like I feel like that Trend was good until we hit this parallel Computing and like AI space I feel like Mo’s law was doing was clipping along I believe Mo’s law was applying to and I’m not sure you can’t quote me on this one but yeah I thought it was applying to the the CPU the
3:35 it was applying to the the CPU the computer processing chip itself so every was it every year it’s doubling two years every two years every two years the amount of transistors they could put on in actual computer chip was doubling and so they kept dropping the size of these resistors 30 nmet architectures 10 nmet 2 N like I don’t know what they’re at right now but they’re at some like absurd like five seven nanometer transistors on these chips to make them one they’re
4:05 these chips to make them one they’re much more efficient when they’re so small but then they can stick a lot more of than on a chip so I think I think we’ve hit the limit the physical limit of Moore’s law on a single chip but I think Tommy maybe what you’re referring to now is people are now building the architecture has changed on computers it’s no longer just one processing unit and how fast can that one unit go it’s like how many can I how many units can I put on a machine right this AI trading stuff is like what thousands tens of thousands of computer
4:37 thousands tens of thousands of computer processing units that are all stuck together all working together this is incredible anyways it’s pretty neat stuff I know you’re all into this stuff as well all right with that let’s jump into some news Tommy we have a one that I’m passionate here about is I do a lot of powerbi embedded projects and there is a new announcement so we have import mode and we have direct query mode in our semantic models we have now with the Advent of The Lakehouse we have the ability for us to add a third option which is direct
5:09 add a third option which is direct lake so this is the ability of having a lake housee support your power be reports I will admit I have been doing some internal testing on models that were on import mode versus direct Lake mode that first load of that model if it’s been evicted takes a little bit longer for that report to show up so it definitely spins a bit more while the computer figures out how to get the model data into the memory of the computer but once it’s in memory it
5:41 computer but once it’s in memory it actually hums along pretty well it doesn’t it doesn’t yeah has very little lag or lead time so that’s nice to have that working smoothly there the article is here I’ll put the article in the chat for people Tommy do you use parb embedded for anything at this time there’s been projects not nearly I I think to to the capacity that you you’ve done it and products as some of the things youve built but definitely have done a few things where it’s like this is pretty neat like the potential solutions that you can
6:13 potential solutions that you can develop around this are incredible it’s an impressive impressive thing that Microsoft’s built yeah I like I like this whole direct L thing the my only comment here on this one so there is some distinct distinct limitations on this one that I do want to point out here you must use if you’re going to use the direct Lake models in embedded which is great because now I can just focus more my attention on doing the DAT Engineering in the lake house and not have to worry about the import and dude I just love direct Lake I think direct lake is a really neat feature fact that
6:43 lake is a really neat feature fact that I don’t have to wait for data to load into a semantic model is well I don’t have to load it in an import method right it whatever is in the lake is what will be represented in the model which is very nice this is only supported with the embed token version two so there’s are two versions of embeded tokens so you need to make sure you’re using the most modern version of the embedded token and then there’s only fixed identities so you need a fixed identity to be able to and it’s described in the article but what that means there no there is no single
7:14 means there no there is no single sign on so this means basically the app owns the data mentality you need a single service principle that will be accessing the data if you’re already using embedded with app owns data then this will work this so this is really more for like a feature for an isv solution solution independent solution provider that what that is yes yeah okay old yeah old old life sounds like it should have been a p but or vendor vendor maybe it’s vendor it’s vendor it’s vendor vendor and then the final
7:44 it’s vendor vendor and then the final thing here is they’re allowing access to these direct Lake models using the service service principle Microsoft’s best practice is to use service principles and a service principal profile mhm with embedded Solutions the downside of this is if you’re using the direct Lake models the direct Lake models cannot use a service principle with profile they must only use a service principle which is a
8:15 use a service principle which is a little bit different from their actual guidance around if you’re doing a multi-tenant embedded solution they want you to use profiles service principal profiles with powerbi to go get the data so I find that to be a little bit it’s in public preview so this is a preview area I’m hoping that they figure out the profiles thing fairly soon here I think it’s a big Miss if they don’t go to GA with actually having the the profiles attached because when you’re building embedded and you’re trying securely right that’s a security
8:45 trying securely right that’s a security thing right that’s Microsoft says this is what you should do for security well then why can’t I use direct leg with security or this type of security seems to make a gap in in understanding of what the feature should be doing so it is impressive though because with the direct Lake honestly my most impressive feature and I think a lot of Saving Face with especially if you’re doing embedden Solutions I’m honestly still amazed when I change a measure in the service and automatically that change is reflected without a refresh in that model it’s
9:15 without a refresh in that model it’s that live connection almost look but it is so seamless it’s almost scary especially when you come from nine years of if you started with obbc and you’ve now gotten to direct Lake wow yes agree with that one let’s go the features Tommy you you picked up a feature from the blog here let’s transition to other news item yeah well before I do I got I got a mik so go ahead and my back is a little little sore I had my first old main twang on
9:45 sore I had my first old main twang on my back this previous weekend yeah this has never happened but I I need to preface it because I want to bring a little humor to to the podcast not a little more humor I should say so we just got a Black Stone steel griddle Grill basically Y which is pretty cool but we won it the reason we won it was we went to the schools Gala the auction thing the really nice thing that you do at your school to raise money for teachers it’s an awesome thing I don’t
10:15 teachers it’s an awesome thing I don’t do well at these events because Mike you may you may have picked this up over the last four five years we’ve done the podcast I’m a clumsy clumsy person I’m
10:26 podcast I’m a clumsy clumsy person I’m not a slob but I’m not classy and these events fence it’s a tie it’s you got to put everything down when you sit when you sit up you got to un you sit when you sit up you got to un buckle you want Buckle I don’t even know buckle you want Buckle I don’t even know the right word button when you sit down if if I sit at a table for dinner and there’s more than two cups more than two cups for you I’m not going to do well it’s just not g to turn out so we’re at this event you’re talking to other dads at J like so what are you doing it’s always still hard to explain what I what I do thing
10:56 explain what I what I do thing and like and I you’re like ah yeah my own business I do powerbi they look at you like okay I don’t know what that is and I my go-to bit is baseball stats for grown-ups that’s a good one that’s a good analogy immediately we sit down I like reach of course and I’m none of this I’m about to say I’m make I’m exaggerating this is just what happened in my This Is My Life The Bread came to the table I’m like great so I reached for the bread immediately did not know
11:26 for the bread immediately did not know my wife’s wine glass was right next to her the whole wine glass on her dress so this is how we start thing oh boy oh this feels like a movie starting out yeah and I have to give a lot of props to my wife very classy how she handled it she just got the black towel she just slapped you and left no she didn’t she’s classy she’s classy she she she levels me up and immediately kind Mar and just cleaned
11:57 Mar and just cleaned it well I’m arly embarrassed people looking at me was a table we didn’t know who who they were they’re like who’s this guy I’m like only got one piece of bread I want to take the whole basket because no one’s touching it you basket because no one’s touching it so it’s like what are we doing here know so it’s like what are we doing here but then they do these auctions and the one that they did that you they do is called paddle timer so the principal comes up undisclosed time on this timer yeah the last person to raise their paddle for like 25 $50 wins this griddle so for $50 an impeccable time in
12:29 griddle so for $50 an impeccable time in I got the last one no way now what I should have done was maybe one of these like little Prince Waves King waves like thank you thank you thing Tommy did not do that his emot his Italian emotions I can see this already taking over this the excitement of the win was a fist pump yeah yeah thing and then I used and then you just look and everyone’s like you very nice thing and I’m like oh oh I just I and then there’s the or derves Mike like the ORD derves are
13:00 derves Mike like the ORD derves are incredible they have the coconut shrimp the hot dog with the bacon around it I want to just take the plate go to the other room and watch a game but you can’t so anyways go back to the while the back’s a problem we had a pick up the gittle yesterday we’re putting in the van and it the first time I ever felt the in your back which if you are young and go like that’ll never happen to me it will it will because it’s just what life is so I’ve had a couple I I’m
13:30 what life is so I’ve had a couple I I’m a little bit older than you Tommy if you remember I have a little more gray in my a little more wisdom we’ll call it wisdom whatever stupidity we’re doing based on the beard a lot more wisdom yeah well I have done a couple of those and I have found it’s the ex like if I work hard a day or two it’s like a day or two after is when my back is like it’s tired now I’m done and then it go it’s the simple thing You’ like be lean over the sink to go grab something from the sink or something like that and then your back’s like nope I don’t have it
14:00 your back’s like nope I don’t have it anymore we’re done so yeah I could I can relate to your story about throwing your back out on certain random things and it just hurts it just like you just want move well it’s you want to turn left and instead of just turning your neck it’s a whole what huh like the whole body has to shift yeah yeah no I get that we just took the our family just went out and took all of our kids skiing we’re that’s asking for it that’s asking for here so I have been prepping a little bit I know I I realized I’m not as young as I used to be when I was young I could just get up get on my snowboard and just
14:31 just get up get on my snowboard and just go and we could just go snowboarding for a day and come back and no problem I I realize now that in my age I had to prepare myself so I’ve been doing like walking on the treadmill making sure I get enough time in like I got to start exercising a little bit before I go or else I know I’m going to definitely hurt something I bought some new bindings for my snowboard that allows me to like step in and step out so what I found it was great and much easier than what I was doing in the past which was wonderful I have to bend down as nearly as much but you still have to bend down
15:02 as much but you still have to bend down and grab this little lever to like uncheck your foot pull your foot out and my walking on treadmill the exercise I was doing at home there is nothing that was happening in my obliques on the sides so I had to like bend over on the side and like reach really far over and go undo the binding and I was like oh my word that’s a muscle I’ve never used in my life no no and I think for getting older you don’t work your biceps or your chest you work the muscles where if like I I have to bend down to pull tie my shoes that’s the muscle I need to work
15:33 shoes that’s the muscle I need to work that is the most yeah my finger muscles are very strong I can tip really good I’m one of the fastest typers in this side of the Mississippi fastest typer in my family so that I’m good at but not much other exercising or lifting heavy things exercising so needless to say if you see me slightly going farther down falling off the camera Tommy’s just slowly drifting out of frame right that’s funny so yeah but we’re gonna we’re g to make it and you did mention
16:04 we’re g to make it and you did mention an article that another powerbi article that came out which by the way Microsoft way to go in January with all your updates it’s been awesome yeah there’s been it hasn’t been like a flood of things no but it’s been a steady clip of just good releases through this this feu this January whatever it month it is this January so far we gave them a lot of stuff in December going it’s not going to be updated Microsoft liter said hold my water coolant hold my liquid coolant and I will cool your GPU
16:34 liquid coolant and I will cool your GPU that’s what they said and one of the big ones that I’m personally pretty intrigued to see how this goes positive optimism here with a dash a little sprinkle of skepticism but it is co-pilot summaries and email subscriptions all right I’m curious to hear this is this is this is again why wait for de se to show up I don’t come on go ahe Tommy give me what do you think what does this look like so this is simply the the gist of it is you have
17:04 is simply the the gist of it is you have your email subscriptions as you normally do but what we can do with this now is simply say what on this particular subscription when it’s sent out generate based on the data a summary to me this looks no different than the narrative summary that that visual we that you simply give a summary of what’s on the context of that page and it’s simply now rather than being a visual they simply said we’re just plug this into the email so the positive here is I don’t know Mike if you’ve used that
17:35 is I don’t know Mike if you’ve used that narrative summary at all in in production reports rather than just dabbling around I’ve seen a lot of positive reinforcement a lot of positive commentary with clients and even myself going it’s not bad especially right out the gate you can you can always customize it a bit based on what you want it to show but even if you just use it out of the box it’s been honestly I would call it a positive really really helpful now I can add this to my email
18:05 helpful now I can add this to my email subscription and where the dash of skepticism come is I don’t know what it’s going to generate I Can’t customize this I can’t wait things like I can with a narrative visual where I can say hey narrative text we’re going to talk about this we’re going to talk about this measure plug it in here it’s really just going to generate what it generates and that’s a little I think there’s a lot of caution there so Mike where do you land on this at least the release of this announcement yeah this is this is a
18:35 this announcement yeah this is this is a good announcement I think this is a very solid feature I’m I’m interested to see this one again my my hesitation here a little bit is co-pilot is right now the threshold to get to co-pilot is only at the upper echelon of the the SKS that were that I’m playing with right so most of my SKS are in the F2 F3 F2 F4 f8 level we’re talking about getting organization started I do a lot of testing internally on my own stuff I’m I run a consulting firm but I’m not a huge consulting firm I don’t have the money to go out and go buy an
19:05 have the money to go out and go buy an f64 capacity and have that on and testing it all the time so for me to go in and test a lot of these features it’s a barrier to entry for a lot of I would say medium to small organizations larger organizations may want to take note of this where I have found a lot of the co-pilot things so this feature particularly right using this directly in emailing subscriptions and getting summaries of information I have found like the summaries that co-pilot provides about data set so one another place that I found co-pilot
19:35 another place that I found co-pilot showing up was you can always you can build an exploration an exploration is a basically a table with a single visual attached to it and so an exploration is an object you can go save into the workspace there’s a little button that says summarize the data set and it has a little button here that you can like refresh the re you can like refresh the re rerun it rerun it rerun it it will know rerun it rerun it rerun it it will give you different summaries of the semantic model and it does some interesting things like hey the date range starts from here to here this one this column has 10 unique values it does some
20:07 has 10 unique values it does some generic summarization but like that seems to me like why would I need co-pilot to do that it just seems like it’s giving me just some general statistics back about the model and to your point Tommy why wouldn’t I just have users summarizing the model a little bit better and having it give me that context right so I don’t really love these features where they’re trying to get it to Auto summarize to your point I don’t get to control it I don’t know if it’s going to give out a good summary when you’re sending out emails to people I think you’re really looking for like regular known content that
20:37 for like regular known content that people are going to expect I think you potentially could introduce an opportunity here that someone gets an email with some information that is misleading or not quite clear for what
20:48 misleading or not quite clear for what you need to deliver most of my experience is the co-pilot experiences work really well when I’m doing very heavy code things right so to me co-pilot makes a lot of sense in Dax view co-pilot makes a lot of sense in timle view these are places where I think the co-pilot makes a ton of sense because it can read the code that’s there and suggest things that based on top of the information that’s presented so I think that’s a better place for co-pilot I’m more leery about these
21:18 co-pilot I’m more leery about these other Solutions where I don’t really want to set up co-pilot and forget it and just let it run I I’m developing a data analyst Manifesto and it’s a bunch of rules rules that you should always apply in situations and like to your point this is something that could potentially experiment or you that could potentially experiment or look at somebody’s goals a know look at somebody’s goals a different way and one of my Manifesto rules we’ll call it rule 17 don’t experiment with someone else’s kpi and if I’m GNA put this in a CO in a summary
21:48 if I’m GNA put this in a CO in a summary and it’s gonna show something even I’m not even GNA say wonky or wrong but not what someone’s expecting someone’s gonna have an issue with that someone’s gonna go whoa I what do you mean 10% they’re going to have questions that you may not have the answer to because you didn’t know where that name number come from and that to me is just on the road to distrust so manifest to rule 17 don’t mess with someone’s kpis and if I’m looking at the summary looking at the feature here it doesn’t interesting job of like saying here’s the sales overview here’s Revenue by sales page I
22:19 overview here’s Revenue by sales page I overview here’s Revenue by sales page my question here would be is how mean my question here would be is how how simple is this model to understand I how simple is this model to understand again and everything I do with mean again and everything I do with anything that’s AI generated it always seems to make more sense to add additional documentation descriptions making sure your model is well rounded so I think part of this too is and I don’t know how this would be impacted does the co-pilot do a better job summarizing the model when there’s a lot more descriptions and details around where does each column come from what does it stand for what do you mean by
22:50 does it stand for what do you mean by sales region or country or these other details my assumption would be it should make the co-pilot answer better based on those things so again I lean on the fact that most of the models that I see in the natural world that people build they’re not very well documented they don’t have all the information in them that they need so my opinion is like that’s where we should really be starting regardless of whether they’re using co- pilot or not the experiences of using these semantic models across multiple experiences inside power. com
23:20 inside power. com that’s much better so I would I would really like that to have so walking away from this observationally I think you’re going to want to start spending a lot more time documenting your model and making sure your model is topnotch with really good explanations because anytime then you throw co-pilot at it it’ll make the co-pilot answer questions much better better th% so that that’s where I would lean on this one and to your point Tommy like again looking at the output of this right it did like Revenue by open sales
23:51 right it did like Revenue by open sales stage product line overview Revenue by Fork I feel like these are things that I would want to if I’m going to use code co-pilot I would want to prompt it with something like what you would normally do is say hey here’s the co-pilot you’re going to summarize this I need you to produce a summary that has the following information in it like look for this kpi and summarize it by region like let it let it do the summary of that particular detail that’s fine or hey I I have a visual on page one it’s called this summarize this visual
24:23 it’s called this summarize this visual in text form and then send the email so I think we need some initial prompting that would then let you add your own prompt that gets run with the actual prompt to go get the data from the model I think that’s what I would need here to feel more confident about this feature instead of just turn on a switch and say co-pilot it I was about to to say that’s Manifesto rule 24 be wary of simple toggles because the fact that you just turn this on and there’s no alert
24:53 just turn this on and there’s no alert you don’t know what’s going to generate again that can cause a lot of waves that you’re not aware of yes yeah so like also be aware here too there are some considerations of this feature you considerations of this feature this is not available in government know this is not available in government clouds so if you are in government this is not going to be available to you there’s no summary supported for dynamic per recipient subscriptions so that’s actually a feature where you put the recipient and information inside the semantic model and use Dynamic email management and then
25:23 management and then your your summary if co-pilot decides to slow down and go really slow for for whatever ever reason you may not even get a summary because copilot took too long to generate the summary for the email so interesting apparently this is a an issue that happens on really large reports where it doesn’t actually get to summarize things so again interesting feature I think Microsoft and every AI company is trying to put AI to put AI everywhere this is another place where I feel like we’re trying to spit a square peg in a round hole a little bit we’ll
25:56 peg in a round hole a little bit we’ll see where it goes it it could surprise me I’ll definitely try it out I’ll see how it works but we’ll see how how we can get this to to to run inside our models we’ll see what it’ll be interesting be curious the feedback from other people as well in this feature the feature is also in the the chat window as well in case you want to get to go in there anything else Tommy we should cover off on I think that’s I think this is a pretty good walk to really what we want to talk about today this is and the this topic I’ll I’ll tell everyone who’s listening as soon as it came out Mike and I at the
26:27 as soon as it came out Mike and I at the same time sent the link to each other going oh we got to talk about this I would agree with that one very much so all right Tommy give us a run up what’s our main topic here for today yes so Microsoft and fabric have introduced for those who are struggling or do just simply don’t have the one of the highest licenses that you can get for a fabric an f-64 and above the co-pilot capacity we’ve talked about copilot and Spades you probably like that sounds great but I’m not spending x amount for my
27:00 I’m not spending x amount for my f-64 well I don’t even have fabric well guess what Microsoft has introduced the new FCC license the fabric co-pilot capacity this is a dedicated capacity that is assigned to individual users at your organization who can use co-pilot so right now if you before this feature came out if you wanted to use copilot it was assigned to a workspace because it was part of f64 and above no f64 no Co
27:30 was part of f64 and above no f64 no Co no no co-pilot for you however now with the F6 FCC capacity or copilot capacity we simply purchased the capacity and assigned individual users or teams to that capacity now they can use co-pilot in a pro and I’m you’re gonna have to correct me Mike but I think it’s just a pro workspace or I’m not sure if it’s premium workspaces as well but that user can use co-pilot really wherever they deem fit yeah I’ll jump in there because I I was able to skim me into the article what
28:01 able to skim me into the article what you’re asking for users can be assigned an FCC to utilize co-pilot in the following scenarios scenario number one powerbi reports on a pro workspace so you could use co-pilot with a pro workspace again license that’s licensing based on the workspace level not the user level the second option here is a fabric workspace that is smaller than an f64 so an FCC would allow you to use co-pilot in these smaller use cases and then finally
28:32 smaller use cases and then finally fabric co-pilot scenarios where fabric workspaces are just purely smaller than f64 yeah yeah so I’m seeing here this is qu Q& A that they have on the article the initial launch surprisingly is only powerbi Pro workspaces premium per user will be supported soon as well and they do not have any plans to support the a or embedded workspace capacities yeah this go ahead Tommy no I was GNA I was just going to say like so right off
29:03 was just going to say like so right off the bat they I think they’ve gotten the feedback that yeah there’s a lot of smaller teams to your point that are simply never going to purchase an f64 because it’s just not going to make sense for a lot of organizations that are smaller or midsize but they really and Microsoft’s trying to push co-pilot so that question was how do we get co-pilot out to more people and this is their solution so I’m still I’m what I’m what the article doesn’t really illuminate to me is like where do I buy this where does this get purchased from so I don’t really understand very
29:34 so I don’t really understand very detailed like what is the fabric co-pilot capacity it what I don’t understand is what is this purchased from from Azure is this purchased from the power. com environment like where does this come from that’s a little bit gray to me right now I I’m not quite sure where I find these things and how much does it cost per user like there’s no pricing in here is it a per user pricing is it a total cost is it the same price as an F2 like what does this mean for co-pilot
30:05 like what does this mean for co-pilot that’s a good point and I’m assuming based on the article because there wasn’t a documentation link or at least I missed it it it talks about your FCC can be throttled which is telling me that you have a capacity you can assign as many users as you want and it’s not per user cost but the more users you assign the you’re going to get get to throttling that capacity at least that’s the takeaway because yeah there was really no no information there
30:35 was really no no information there about a per user except by assigning it that way look at this we’ve got the real experts showing up here talking capacities we’re so we’re talking false good we got the article is a little bit gray here and so feedback to Microsoft maybe add some more context here but gy Cube showed up we very rarely have gy Cube so gy Cube welcome to the the podcast we appreciate you be here so I see you purchase a regular fabric capacity and then you designate it inside the settings of the fabric capacity to be a co-pilot capacity so
31:08 capacity to be a co-pilot capacity so the the extent here is you could
31:09 the the extent here is you could purchase an F2 and then designate that F2 as the co-pilot capacity but that’s what I was missing that’s what’s not very clear in the article to me that okay the the order to get this is buy any capacity turn it on as a co-pilot capacity and then you can use that as your capacity piece there as well that’s very informative and the article is not very clear there so maybe some addition would be would be helpful typically a lot of these articles come out with the ability to see okay you’re publishing the feature
31:41 see okay you’re publishing the feature and then immediately it’s followed up by here’s the learn more document I don’t see a a link to learn more so that must be something that’s there also as well well it’s nice to have smart people that are listening so because Microsoft has actually listened to the to the podcast so Mike this feature to me there’s a lot of ways that I would describe it when I first read this article I was honestly like there’s a bit of a few words finally there’s a bit of can’t wait and then there there’s a
32:11 of can’t wait and then there there’s a lot more of what are the restrictions but I’m gonna focus more on that finally feature here because to me for and I’ve heard from clients like oh I love this co- pilot we use it all the time in office I can just use it in powerbi right you’re like yeah well how much do you have on on hand how much scratch you got so got so basically so knowing what I know now right so the fact that I can turn on an F2 capacity and dedicate that as the capacity for the fabric so there’s some things I’m thinking through I’m unpacking this right as we reacting to
32:42 unpacking this right as we reacting to the article I like this feature a lot already now that I know I can turn on something that’s dedicating itself to a fabric capacity Pro have been left out of this whole thing there there is some goodness that comes from fabric capacities anytime that I’m building things in like the power query so in Power query or data flow power query data flows things have need help there and I can link my desktop to co-pilot and have co-pilot run things through desktop so to be able to expose that to more users across my organization I think will be very beneficial there
33:14 think will be very beneficial there is this ability to run co-pilot and make report Pages again if you’re looking for a quick win on a semantic model and build a report page with something that would be helpful turn on co-pilot give me a build me a report page on top of this data and then let it just start building you some things it reduces some friction there for new users to create some goodlook reports so I really like the ability to be able extend this further into all users inside Fabric and not leaving the pro and premium per users out of this well here’s the question so I think we’re
33:44 here’s the question so I think we’re both in agreement that this is obviously a barrier that’s been removed and we’re all agreement like good keep going good job Microsoft but let me ask you here because obviously there are going to be some restrict restrictions simply around you have a capacity so it can only handle so much data or so much processing at once who are you giving this to in your organization and and I I’ll preface it a little here if I have my hardcore powerbi Pro power users I have
34:15 powerbi Pro power users I have people who dabble in building models and reports and we obviously are touching touching fabric I’m thinking here like there’s I can’t give it to everyone or at least a single one without obviously running into some just just capacity throttling MH so who’s the primary audience if you’re going to from an adoption point of view who are you focusing on first what what do you mean which users what user are gonna get the most impact from this from this yeah well I’m I’m going to probably
34:47 yeah well I’m I’m going to probably throw this at people that are brand new to powerbi right this is if you’re an expert you’re probably not me as me as one who’s a pro developer who’s been doing this for a while I don’t spend as much time like mucking around with like the auto generation features of things I just make what I want to make because I know how to make stuff right I don’t I can go to Dax query viw and just make my measures I don’t have have to go to Dax query viiew and explain my measures however a new user would be very interesting to so go into
35:17 would be very interesting to so go into Dax query view use co-pilot and say hey I am interested explain this measure what should I use here what are some you what should I use here what are some what does the summarized columns know what does the summarized columns function do in Dax this is something that comes from co-pilot inside the Dax SC view so I think from that standpoint that’s really interesting to me to be able to have like a coach right so new users need a little bit of a coach next to them to get through these initial like hurdles of building some basic measures right
35:47 building some basic measures right Eugene Midler has done a really good diagram of how he learned Dax and has a really good diagram of like where things are easy so I use it all the time in my my training because it’s so good but the early the easy things are summarization averages like this makes sense on top of a single column but it’s when you get into like filter context transition like how do I do a percent change that’s when people’s heads start melting when you’re like new when you’re starting out you’re like wait what what do I have to do how do I build this what’s the context of my single data point so in in that initial
36:19 single data point so in in that initial stage of understanding and unpacking how to make a Dax measures in the model I really think this co-pilot portion for that would be very useful and it would also be nice if co-pilot could do a little bit of like SQL biing on top of this right you could you could have a a a slow measure or a slow visual again it’s not there today but it would be nice if co-pilot could just take note and say hey this visual slow I’ve run the per performance analyzer or I see
36:49 the per performance analyzer or I see your performance analyzer output here’s the measure that appears to be slow when I look at this measure it’s it has some bad practices you’re using an all St in here you probably shouldn’t do that here we might suggest you rewrite it like this that’s when I think when it can suggest some minor changes that include Pro performance that could be really impactful there I don’t know why but when you said that I think copile will help a lot of new users learn evaluation context my heart rate went up slightly I just I could feel it I’m looking at my watch right now that’s that’s
37:21 watch right now that’s that’s that’s a tough one I’m going to differ with you on the primary audience or the first audience here I like that that one I really do and I think that’s a great one but to me if I’m a brand new user I’m thinking okay how much are you even using copilot today do you even know where to start but to me the primary audience here or who I want to First enable this for I’m not saying the other ones are not important but who am I going to focus on first I’m focusing on power users in powerbi to get them using
37:51 power users in powerbi to get them using fabric products and services because Mike how many we talked about power query and the solution is from that data engineering point of view that you’re going to be notebooks or you’re going to be doing something in pipelines right or what the applications of gen two are if I’m coming from gen one data flows so I think the ease of use for someone who already knows what they want to do with their data but they’re not you like again they don’t have any skill or knowledge or experience in notebooks
38:24 knowledge or experience in notebooks hello this is perfect for hey we’re going to ramp you up to fabric and this is where you have a coach where I know the questions to ask and I think that’s where I’d be weary with new users as my priority because I don’t even know if they’re going to know what to ask I know we’ve talked about the report like build a visual for me a build a report page but if they’re not aware of the data and they’re just starting off I think they’re going to have not the desired outcome this goes back to Kurt bu’s article on when he used co-pilot if you don’t know how to ask
38:55 co-pilot if you don’t know how to ask the right question you’re probably GNA not get the right answer that’s that’s that’s an article from sqlbi initially and I think Kur on that with the you and I think Kur on that with the but ask better I even told know but ask better I even told my son this I’m teaching my kids this ask better questions I can give you better answers so like I get a lot of questions from the kids like just random hey why did that do that well well don’t just ask me the question why you ask me the question why why is your blinker always on like know why is your blinker always on like because like like that but like
39:26 because like like that but like actually actually specify that is all right so Tommy I don’t know if you’ve been following the comments in the chat here we have to we have to do a little bit of a walk back here so guy in the cube has reread the article that was just recently published and again this is maybe this is Credence to us this article is confusing it’s not we smarter than we think we are okay we were smarter than we thought we were so this FCC capacity would allow you to extend Fabric or sorry co-pilot into Pro
39:57 extend Fabric or sorry co-pilot into Pro and premium that does exist however the FCC switch or toggle will only exist on an f64 and higher so when it’s it correction yeah correction the FC the FCC feature does not exist on an F2 through F32 you can only turn it on an f64 but once you turn it on F at x64 level you can go use it in non Fabric workspaces or elements so it’s basically
40:27 workspaces or elements so it’s basically saying you have co-pilot at an f-64 we want to give that to users in other workspaces so they can leverage it this changes my opinion now because I wanted the former option I want I want it to be added to any capacity so I already have to so just so I heard you correctly let’s let’s unpack this again I just want to make sure the recording we may have to restart the podcast today totally threw us off you have to have an existing f64
41:00 off you have to have an existing f64 even to get started correct there there is no there is no F2 capability this one yeah I’m disappointed because I really I would be willing to try this out if I could do this all the way down to an F2 did we miss that no no it’s in there all right it’s in there I’m I’m gonna search for the word f64 on the article right now and I’ll I’ll read you the line where it says this okay so what is the fabric Co pilot pilot capacity fabric co-pilot is a feature
41:30 capacity fabric co-pilot is a feature that allows users to utilize co-pilot
41:32 that allows users to utilize co-pilot even if their reports are not on the f64 capacity which sounds like what we’ve been talking about yes does the FCC have a size or skew restriction bullet point number three in the Q& A portion at the very bottom it says yeah only a P1 f64 or greater can be labeled as a FCC so I thought that was a typo because I’m like with everything else that was being said nope I it it is still limiting us to the f64 so that’s where this is where I’m
42:03 f64 so that’s where this is where I’m like man like the the wind had just gotten taken out of my sales I’d really like to see this down to the F2 level or even some smaller skew like this doesn’t help me then like you’re still at that barrier yeah now it’s time to go all right Tom Tommy it’s time to go install our own deep seek on this thing and run this thing for pennies on the dollar man what’s going on can I install fabric on my raspberry pie that’s what I want to know so know so yes the very bottom of the article is where this feature like it’s where it
42:33 where this feature like it’s where it calls this out and that should have been like front and center at the top of the article like hey has to be you can’t this is a f64 P1 feature that you can turn on and so let me just unpack the feature again one more time it means you if you have an f64 and you want to delegate it to other non fabric workspaces you can do so so this is giving co-pilot to non co-pilot work only after you’ve reached the f64 level it simply helps organization yeah it only helps organizations who have had or
43:05 only helps organizations who have had or consider the caution of the problem that hey I don’t want co-pilot to eat up all my f64 yeah correct and that’s what that’s doing but for everyone else what we were praising the heavens for still still outside looking in still in the it really feels to like how much co-pilot capacity could be run this is this is where I would argue like I really would want this to maybe run on like are we is it worth the spend right are we going to find that much value from copilot that I’m going to spend five grand
43:36 I’m going to spend five grand and dedicate a single capacity to co-pilot things or does it make more sense for me to say look I’m willing to spend an F2 or F4 level of co-pilot capacity that’s adding enough value for me so this is where we talk a lot in the podcast around adding the balance between value and cost right if you’re if you are using co-pilot and finding an immense amount of value you’ll pay more for it but if I’m thinking about this like if we don’t haven’t tested it out yet you can’t even get our hands on it then how do I test the value all right Mike and I know
44:08 the value all right Mike and I know some people don’t really care sometimes when we have we’ll say less than optimistic takes on things but this is one of those times yeah I got a problem I got I got I got a big problem with this because when we’re dealing with llms and that’s what co-pilot is and that’s what the agents are copilot Studio that’s text based I am failing to see how this is I’m failing to see how much of a problem this is for this to get integrated into a model that already
44:39 get integrated into a model that already has my measures right I I’m not asking co-pilot to look at a billion rows I’m looking at to look at my measures where the thing that takes CPU or agents the time to generate is really over text it’s over text based applications and and yet co-pilot agents if I were to develop co-pilots I’m doing that right now for on a bunch of projects yep the cost is anywhere close to a f-64
45:09 the cost is anywhere close to a f-64 license or anywhere close to what a co-pilot capacity would be I can have a 10 co-pilot agents right now and I’m not even sniffing the cost of here how can I have co-pilot for whatever is it n $20 a month for my entire office which is can go through all my share points okay and and and like it’s going through all my documents which has to not just look at all my thousand of freaking files in SharePoint but then it has to actually look at the files in the word documents and it goes quickly so this
45:39 documents and it goes quickly so this this is my point here than your point Tommy like this is where I think the co-piloting piece is confusing to me because there’s licensing at many different levels I I’ve bought the office co-pilot license yeah how do I use it all the time every day nah not really do I use it sometimes yes I have it rewrite some emails or some things so it’s valuable to me saves me some time in certain areas but you some time in certain areas but it is it is it life shattering for know it is it is it life shattering for me is it something like I really need this one I’m not sure it’s that compelling yet like when I’m generating
46:10 compelling yet like when I’m generating like marketing content or images or things like that dude I will pay for AI all day long like I use other creative programs to generate stuff if I could get if I could feed it more like thumbnails right hey we’re going to build more thumbnails or write descriptions for this video or that’s where I think a lot of this AI stuff for me adds a lot of value do I find an immense amount of value for it rewriting my emails n not so much but to your point Tommy if I’m already paying 20 bucks why isn’t that 20 bucks also being paid for by co-pilot
46:42 20 bucks also being paid for by co-pilot in powerbi that seems like a place I would want to use this more often and now let’s let’s talk about GitHub co-pilot goodness good which who owns who owns GitHub it’s free come on man the GitHub co-pilot is free again granted it’s slower it won’t go as fast but they they throttle you a bit there but like that portion of the copilot piece is free so interesting so I feel like this has been an emotional episode for me I I got really hopeful when I
47:12 for me I I got really hopeful when I heard tell me the the the feature was coming down to a lower skew level and I I feel like the wind has really been taken out of my sale knowing that it’s going to be only at F 64 level the irony of the GitHub co-pilot mic is I can’t use co-pilot in the service but if I use GitHub I can use get copile all over my timle my timle files everything I want yes that’s true like that’s like hack in the system but and Mike to to the point of the general co-pilot you buy an office if you really want it to be impactful you got to create an agent you have to create a in copilot Studio because that can actually
47:43 copilot Studio because that can actually look at a SharePoint folder again going back to what we’re talking about in a specific data model or data set and here’s the thing too with co-pilot in fabric it’s not necessarily going through at least if I’m using the notebook book one I’m not asking it to hey I want you to look at all of my models in this workspace and then do something it’s helpful for I am trying to do this I want to add a column you to do this I want to add a column it’s very still General so yeah I’m know it’s very still General so yeah I’m I’m I’m I’m with you Mike that’s that’s my rant I’m gonna pause there for from a
48:15 my rant I’m gonna pause there for from a skep or pessimistic point of view but come on there’s been a lot of news announcements in general around AI based things and so one of them is there was this whole idea of like co-pilots are not going to be actionable they’re going to do actions with things right so that was was there’s like a a open AI just released like they had their $20 subscription for like co-pilot or or not co-pilot they have their own version of open AI generative stuff right so they have a $20 a month fee for that now they have a $200 a
48:47 fee for that now they have a $200 a month fee for like another AI but it’s it’s based on actions like it can do things for you I want you to go open this website and do these things and like that’s the to me that’s where I’m starting to see things are going to be very effective in the co-pilot space right when can co-pilot start making a vis a single visual or hey co-pilot make me five variations of a bar chart using these this and this field that’s the stuff that I want to start seeing to do and generate because that
49:17 seeing to do and generate because that would that would make more sense because then it could just again the the thing I think is important here is co-pilot can then have some pre- prompting like an agent Tommy like you were saying right build an agent around a visual builds right and that agent is able to quickly go through and give you here’s six different visuals I’ve built you different styles this one has titles on this one has titles off and it even could tell you I built this visual to be more minimalistic I built this visual to have like a dark theme I built this visual to have a light theme and it’s
49:48 visual to have a light theme and it’s more ver brose like I added data labels here it should just give you an array of things to pick from and say which one do you like and you should be able to hit it and boom it saves have you ever watched the guy he’s on YouTube how to powerbi is is the YouTube channel that narrows it down what he say so the I think is his yeah I know the B but that’s from the 80s that’s different different this guy’s he’s current so he does a lot of how-to videos and like how to explain how to build a visual and his visuals are beautiful by the way I
50:19 his visuals are beautiful by the way I know what you’re talking about he’s great phenomenal he’s great really really good Creator I love the stuff that he comes out with I I I learn a lot from him but every time I look for this I look exactly at that feature and I go to myself I’m like literally counting the number of clicks that he had to do to get to this thing it’s like click this adjust that change this set this do this this and we’re probably like 20 clicks deep every single so now me as a user I have to memorize 20 clicks to get to a bar chart that now looks like a
50:51 to a bar chart that now looks like a box and whisker chart yeah like or I have to do all these settings to get to a chart that looks like this there’s so much friction there’s there’s a to me that’s the barrier to entry right that’s where co-pilot should be finding and fixing these things right we should be feeding copilot a whole bunch of like defined here’s a bunch of stylized visuals I should be able to say I need a box plot it say look I can style it for you here you go here’s the style for you like that’s the I think that copat would really make easy for us
51:21 copat would really make easy for us you’re you’re you’re on to something here because I think Microsoft greatest I think asset right right now I think is something that they’re working on is that ability where like with a open AI or other products you have to really build it up like it’s it’s not out of the box going to work with the general co-pilot with fabric like I can just create a database right away awesome to your point about the model idea would you rather than with co-pilot in in an ideal world like open AI is
51:53 in in an ideal world like open AI is doing right now they have all these different types of models one better at
51:57 different types of models one better at tasks one uses Advanced reasoning One’s best at reasoning one are better for searching the web I I would almost love rather than a single fabric co-pilot and I know there’s multiple co-pilots that do other things but rather than purchasing a single co-pilot everywhere why could I have a coding co-pilot in Fabric and a model co-pilot where hey I I want a co-pilot just to look at models so I can build my reports other ones just help me with code just help me when I’m in a notebook on what the right
52:29 I’m in a notebook on what the right function is or help me create a function to do X I don’t need you to look at the entire Lakehouse on all the data in there would you rather that or do you want them to keep going with the singular singular copilot if you were to purchase individual co-pilots for coding for visuals for model building I don’t want to spend money on this really what I want I me honestly right okay sure is it isn’t the whole concept of my understanding and this could be my lack of not building them myself right so
52:59 of not building them myself right so agents right so if you’re talking about co-pilot agents or you’re building your own agents right the whole my whole understanding of an agent is you’re building the agent in order that you’re able to predefine some prompt then in like you’re basically predefining a prompt hey this is the format I want you to think about this is the output I want you to give only right answers like you’re you’re pre- prompting the model and basically prompting it a couple times before you actually give it your final prompt so
53:30 actually give it your final prompt so the agents are pre- prompting or adding additional text or context to what you’re going to ask the final question to right I saw a little YouTube short some people were discussing careers in the future right and they said what do you push your children into like what what what research areas do you put your kids in right do you have them become lawyers no a lot of the lawyering and all the searching through text and writing a lot of that’s being handled
54:02 writing a lot of that’s being handled more and more by AI like large language models are doing a lot more of that do you tell your kids to go be like all these different careers and a lot of these other like do you tell your kids to go be like a marketing a copyright or not copyright but making marketing copy right build images build words build marketing they’re now finding that companies are now using AI to generate marketing copy very cheaply and it’s you marketing copy very cheaply and it’s millions of dollars less than it know millions of dollars less than it would have been paying a third party organization to do this one so
54:32 organization to do this one so people are going to get wind of this you people are going to get wind of this third party organizations are going know third party organizations are going to use AI to build your marketing copy for you and you’re GNA be like that was just AI generated did you even read it like is there even involved here and so I think I think what’s Happening Here is we’re seeing a huge shift and again watching another interview with Sam Alman he’s like in the near future not one year but like five years 10 years down the road this AI stuff is going to greatly shift which type of skills people need in order to do the next line
55:04 people need in order to do the next line of work so I I think this is going to really shift things here and to your point Tommy back to your question right do you do we build hyp specified models I don’t want to pay for them I want a I want a library just same way I go to like GitHub models you can go to GitHub models today and you can run models right off of GitHub in preview just go there and run models llama open AI version like all chat GPT stuff like there’s a whole bunch of other things that are directly on GitHub you can just go use that’s
55:35 on GitHub you can just go use that’s what I think it should be more like it should be more like hey co-pilot is exposing all these libraries of large language models to you and you should be able to pre- prompt it and so Tommy to your point if your organization has three or four models you should be able to to write a agent to say look these are the three models I want you to care about St up on these models I’ve severely documented them I’ve added a bunch of descriptions I’ve added titles I’ve added measures I’ve done all the things to let the large language
56:05 the things to let the large language model understand them so study these three models this is what I want you to be really good at and I think that potentially will give you a lot more Runway right Tommy is going to roll out co-pilot for his organization he’s gilot for marketing data yeah yeah tomy’s going to roll out this coil for marketing data he’s going to attach it to this workspace it’s going to run on this capacity and Tommy’s already pre- prompted it Tommy’s done some agent work with it and by the way it’s it’s more catered to that need
56:36 it’s more catered to that need or that team and let me give you another picture of this right let’s imagine we’re a data team and we’re we’re sucking in a whole bunch of like transcripts from like conversations on phone calls right so you’re using one somewhere else there’s an AI That’s transcribing voice conversation into transcripts there there’s large language models you can say you feed at the conversations and say Here’s the conversations my customer is having give me like act as if you’re my customer let me ask questions to my customer so there
57:06 me ask questions to my customer so there should be an a a model and I think this should be supplied through co-pilot right this would make sense bring the co-pilot experience let Tommy build the voice of the customer agent land that inside my powerbi reports that’s what we should be doing and Co funding all of that cost or running that’s what think things make sense and I’d be willing to pay for agents specific to fabric and I know you see you don’t want to pay for anything Mike I don’t want to pay for the line of my I don’t want to pay for the line in my TV but it’s bothering me
57:36 the line in my TV but it’s bothering me and I gotta fix it if I’m already buying fabric just let me like it’s up to me to decide to to to like the same thing happens with like SQL server or the fabric SQL so SQL inside Fabric or pipelines versus data flows Gen 2 everything uses compute so why do I care what I’m using it should just part of the package well I think we’re realizing still the whole let’s go back round trip to this GPU thing I don’t think these co-pilots are cheap at all for Microsoft to do in in terms of also how
58:07 Microsoft to do in in terms of also how much this probably I haven’t heard from a client yet I don’t know if you have but I haven’t heard of how much co-pilot sucks up in a f64 capacity if it’s used on a daily basis it could be like where it’s like whoa I can’t do anything else but co-pilot this is insane again and I’m saying this where it would make sense for it to be a separate payment I I do where we are with technology today I can understand that because it it’s intensive it’s labor intensive to get
58:37 intensive it’s labor intensive to get your something working like that but at the same time well shoot like you said don’t scour my entire universe of data let me choose what I want let me be hyperfocused and then we’ll go as a pay on pay as you go for an agent point of view if you’re already doing that where I guarantee you most organizations have more more files than data and I can do this on a co-pilot agent in in the co-pilot Studio my friend we have a we have a model we have a way to work that way and I will end there okay so good episode
59:10 I will end there okay so good episode sorry if this episode was confusing so people hopefully listen through the whole way through on this episode because there’s there’s a lot of things to unpack in this episode let’s just quick so I’m not going to do final thoughts I’m actually just going to summarize what we talked about today because we waffled a little bit on the conversation so let me just resmar what we heard here and we article all right so the how do you use fabric co-pilot capacity FCC all right so you if you have an f64 skew and you want to use portions of that f64 skew or P
59:41 skew or P skew turning on co-pilot for other parts of your organization you can do so so what this does this is a feature that allows you to go to your f64 or P1 skew capacity in the admin settings turn on a setting that allows users to use this go to the capacity setting and turn on this feature once this feature is turned on other users can be assigned to that capacity to run just co-pilot meaning you have a powerbi pro workspace a premium per user workpace is not
60:11 premium per user workpace is not supported yet but eventually it’s on the road map if you want to use it on an F2 capacity so the usage of co-pilot can be given across more of the organization but you’re still required to start with the f64 capacity is that a good summary time did I miss anything I I think I think kopil would be happy with that summary okay so we should run this back through we’ll run it back through the co-pilot and we’ll say what did you think of Mike’s summary at the end did he get it right was the feature correct so that being said that
60:42 feature correct so that being said that sounds like the feature that we’re talking about here we do appreciate your time thank you for hanging out with us as we unpack these co-pilot and large language models inside powerbi and usage let us know your experience have you turned this on are you would you want to use this feature please let us know down in the comments I’d love to hear more from you in the community on where and if this is a good feature for you as your organization asking for this feature to be existing with that being said Tommy we really appreciate everyone listening to the podcast please
61:12 everyone listening to the podcast please share with somebody else but where else can you find the podcast Tommy oh my goodness you can find us on Apple Spotify or wherever you get your podcast make sure to subscribe and leave a rating it helps us out a ton and please share with a friend since we do this for free do you have a question idea or a topic that you want us to talk about in a future episode head over to power. podcast leave your name and a great question and finally join us live every Tuesday and Thursday A. M Central and join the conversation on all power. tips social media channels you heard it
61:43 tips social media channels you heard it here first we’re confused about this new feature hopefully this adds some clarity to what you’re talking about have a great week and we’ll see you next time
62:18 night [Music]
Thank You
Want to catch us live? Join every Tuesday and Thursday at 7:30 AM Central on YouTube and LinkedIn.
Got a question? Head to powerbi.tips/empodcast and submit your topic ideas.
Listen on Spotify, Apple Podcasts, or wherever you get your podcasts.