AI is Getting More Expensive: What You Need to Know
Hey, everybody. Welcome to Make Others Successful, a podcast where we share insights, stories, and strategies to help you build a better workplace. If we haven't had a chance to meet yet, my name is Mitch, and I am joined here with Mike to my right and Matt on the other table over there. Hopefully, familiar faces. If not, you will get to know us very soon, I think.
Mitch:New podcast episode today, and I do wanna say it is printed on cardstock, so it is very important. It is what percent thicker is cardstock than normal paper?
Matt:Depends on the card cardstock, but this is probably double.
Mitch:Double. So twice as important as our previous podcast. So you need to listen to the whole thing. We are talking about, believe it or not, Copilot and Cowork and Microsoft and what is changing in this world because there's been a couple little releases, things coming out of Frontier and seeing some changes to pricing models. And it it kind of causes us to think differently about what do we evangelize and what do we say this is worth your money versus this is not worth your money.
Mitch:And given this is all new, it's a little bit unformed. This is our forum for having some of that conversation. I'm excited to talk about it today. So Matt, do you want to give like the overview of what happened recently that that has changed?
Matt:The big thing is that co work, so Microsoft Copilot co work went GA and as part of that GA, they dropped the pricing model.
Mitch:Dropped? You mean introduced a pricing model?
Mike:Not not dropped the price.
Matt:Yes. I meant like mic drop.
Mike:Okay. So something we should probably say first, what is Cowork?
Matt:Yeah. We're going to get to that. So we're going talk about what co work is. The pricing model is the thing that happened just recently and that pricing model gives us an opportunity to talk about the pricing of AI and what how that works, etcetera. So that's the thing that happened.
Matt:I think that we want to do what you what Mike said is get get more into what co work is. Who wants to take that?
Mitch:I'll do co work. So co work, if you are familiar with Claude at all, Claude has the concept of co work and so that might ring a bell. If you are not familiar, the concept is that typically in AI world, you're chatting with an AI model, a large language model, and it's giving you responses back. And maybe it'll do something for you if you tell it to do something. But oftentimes, something that needs to happen involves lots of round trips or multiple, like, spider webs of information that needs to get grabbed and then brought all back together and then acted on and then carried out and then reported on.
Mitch:Like, there's a a life cycle to a lot of the things that have to to happen when you say something like, can you schedule time with me and Matt next Friday? And so co work is the sort of thing that you can ask to do that for you and act on your behalf. So it will go, look at my calendar, look at Matt's calendar, look at maybe our previous events together, see if there's a common time that we have this sort of meeting, look forward, is that same block available? If no, what are other options? If yes, I'm gonna send this request if you have it approved.
Mitch:Otherwise, I'm going to return it here in this window and say, hey, I've drafted up this thing. Does this look good to go? Do want to allow me to do this thing? And it can be really powerful and it can really relieve some of the busy work in someone's day to day if they have recurring tasks and they want to offload some of the the thinking work or just the the busy, maybe menial work that AI can now sort of do for you.
Mike:A simple a simple example just like I wanna restate it really simply is, it used to be you could ask copilot or an agent of some sort, find a time that works for both of us to have a meeting. Now the pattern shift or the paradigm shift is find a time that we both have available for a meeting and book it.
Mitch:Yeah. Act act for me. Right. Yeah. And so the the term co work here is brought over from the Claude world.
Mitch:Copilot co work is using the Claude models generally speaking. I know they announced integrating some some other ones over in the near future, but that's the the parallel here is they had Claude Cowork. Microsoft got it over in their world. Now there's a Copilot Cowork, which is just that concept within the context of your Microsoft Suite and your Microsoft tenant with your Copilot agent. I think I did it.
Matt:Yeah. Got all the high points. Yep. Yep. Do we want to talk a minute about how what, like, what makes that unique and what makes that different?
Matt:You talked a little bit at a high level, but it is kind of an interesting model because it functions a lot more like a person instead of with a goal in mind. It reminds me a lot of what the difference between regular chat and researcher is, right? So researcher has been around for a while, for a long time now and it allows you to create more complicated work product, but it's going to ask you some questions. It's, hey, like I need to get all this stuff done so I can create this thing and make this thing happen. Cowork can do multi step processes.
Matt:It can not just generate one thing but generate a whole host of things. It is, in fact, in many cases, the whole I want to schedule a time, that's almost too simplistic for co work when you think about what you want to use it for. Where it really is useful is to do large and complex work across a number of data sources integrating in multi step processes. Right? Like that that's fair in what what you Yeah.
Matt:Would
Mitch:It can't. Well, you're starting to, like, this door has become like thrown open, I think, by Microsoft because of this pricing increase. And so there's like this world of what can it do versus what should you use it for. So yes, I I would say the stuff that I'm describing is fairly simple. It can definitely do it can go deeper.
Matt:So you think, forget the pricing model, the simple things you still would prefer to use Cowork for over the other tools?
Mitch:Generally speaking, yes. So I've I've had enough instances where I can tell CoPilot on its own does not have enough context to really answer or do what I'm asking it to do. And I know in my brain, you only grabbed back the last five days. I need the last three weeks. And it's like, oh, I can't conceptualize of that or it can't Obviously, gather all that that's that might not be the exact example, but yes.
Mitch:That's interesting. I know co can just throw it at it and it'll do what I tell it to do.
Matt:Yeah. Myself and Cowork, I have found we've been using Cowork for in Copilot for
Mitch:Couple months?
Matt:Yeah. Couple months now. I find myself the smaller tasks, it's a little annoying to to use I find it annoying to use Cowork.
Mike:Faster to just do it yourself?
Matt:Faster to just use Copilot to do a good enough job for what I'm doing.
Mike:To get an answer or
Matt:get To get to get me started.
Mike:Create some content or Another
Matt:thing that is happening with this that has happened before this, but it's related, there is now a bunch of edit functionality within standard Copilot, right? So many times I just can use that. Like, it can schedule meetings for me. It doesn't need co work to do that. I don't need co work to edit an Excel document, I can do that in Copilot.
Matt:What Mitch is talking about is, which is another piece that I have found, which is that I have given prompts into Copilot and asked it to find do things, find things, whatever, and been frustrated with it and then take it to Cowork and it works great.
Mitch:Knocks it out the park.
Matt:Knocks it out of the park. It's amazing. But it's just interesting, when I think about co work, even without the licensing thing, which we're going to get into a little bit later, I automatically think I want to try to avoid the smaller things. That's just been my experience. The extra thinking about it differently isn't worth it to me.
Mitch:But Okay.
Matt:That's interesting to hear that you have the other the other perspective on it.
Mitch:Yeah. There there's a few things we want to get to in this section written here, but I actually I think I want to save those for after of like the question is how are we going to actually use Cowork? How are we going to integrate it into our workflows and things like that? And I think we want some of the the pricing and that like the thought process behind that before we share or think through some of that stuff. So do we wanna jump ahead to
Mike:Sure.
Mitch:What changed? So the thing that changed is today, you can buy a Microsoft three sixty five Copilot license for $30 a month per user.
Matt:The premium license.
Mitch:Yeah. The it's a moving target. What that gets you is access to Copilot chat. So, you know, asking questions about things that are in the context of your Microsoft tenant as well as lots of different agents and things like that, and also access to Copilot in your tools and able to edit in things like Word, Excel, and PowerPoint and things like that. That's been the de facto standard for a little bit.
Matt:Yeah. For a while.
Mitch:And so Cowork came maybe a couple months after Claude. Cowork hit the ground running and it was introduced as a Frontier Preview. Option. Preview. Right?
Mitch:So Frontier is the part of Microsoft that they release the release ring where people are getting early access quote unquote to these tools. And it's where they can test things and get feedback on things from a relatively public or like a if you have a subscription, you can get access to those things. When it was in there, you could do whatever you want with Cowork. You could ask questions. As long you have
Matt:a Microsoft three sixty five license
Mitch:Yep.
Matt:You could do ask it as many questions as you want, do whatever you want with it.
Mitch:Yep. Yep. Exactly. So we were interested, is maybe the the lightest term, to find out co work, when it comes out of Frontier, is gonna cost money. It just got released with a pricing model where they say, here's how much each token costs.
Mitch:Here's about how many tokens these sorts of requests will take. Here's about what that means for your team. And it is not included in your Copilot license.
Matt:But it does require it.
Mitch:But it does require it. Yes. So it is an add on, an added cost to Copilot. You can't just get co work. And so this was one of the first times where we got introduced to the concept of copilot is not everything copilot.
Mitch:There is something beyond there that you can pay in addition to. What did that trigger in your guys' brains when when that happened?
Matt:Lots of things. It all of that is happening in the backdrop of everybody's charging more, everybody's trying to figure out how to turn AI into more money. So there was definitely some thinking about that. But for me, it's kind of it's more of a marketing frustration than an actual frustration with the the cost of it or that it is is going to be an additional cost. You know, Microsoft is marketing Copilot as being able to do all these great things and they add this new thing that definitely is better, but they put up a paywall.
Matt:In some ways, like I said, I can understand why they're doing it, but at the same time, the main tool isn't as good as I want it to be. And so then gating a tool that will do something better than the rest of it, the other tools, that isn't a big thing. Like, if it if it was only the large complex tasks, it's like, okay. But some basic things that I wanted to want Copilot to be able to do, like I said, I couldn't get Copilot to do it but I could get CoWork to do it. Yeah.
Matt:And so I think that's where a lot of my frustration comes from.
Mike:I think it's going to be difficult for them to draw the lines around the things that you're talking about because if they say, well, got to pay more just so it does a better job, what does that really mean? Right? People are gonna have a very difficult time with that. So I think that's a challenge. Hopefully, that that will become more clear as time moves on.
Mike:But the other two things that came to my mind were, well, there it is. Somebody's finally trying to recoup the cost of data centers. That makes sense because there's been a lot of that spending going on. And then ultimately, I think it's going to make organizations think about their spend on real human resources versus augmenting humans with various levels of AI capability. Now that they actually have to spend money on tokens and all of like, whatever that ends up being at the end of a year, if I have, you know, a $100,000 FTE, how much more value can they provide if I spend this much more on the tools that they're gonna have access to?
Mike:And I think that's an interesting question.
Matt:Well, it I think it also bridges the boundaries between what traditionally agents are because agents this is not a new concept. When you think about agents and you think about regular users that don't have the Copilot license using agents, that's all consumption based. That's based on the credits that you're using and it's bridging that bound much like what you're talking about of like, okay, I can build an agent that does something and I'm gonna get charged based on consumption, so people are doing this much. Or I'm gonna build, use co work with someone and they're gonna, you know, be supercharged even more than just regular copilot and I'm also gonna get charged a certain amount, whatever that might be, for the things that they're doing. I think it also complicates the story around telling people what to use.
Matt:I'm very, very interested to see what happens with the default user behavior is to use the most powerful thing all the time. All the time. I don't care if I'm asking it to, you know, what's one plus one? Use the most powerful thing. That is default user behavior.
Matt:And if that's what people are using Cowork for, and I know that's an exaggeration, but I think the the statement sort of holds, that's a lot of money for not a lot of value.
Mitch:I've started to see and hear more narrative from Microsoft employees around, well, you don't need that tool for this thing. You can get by with with Copilot chat or Copilot within the context of one of the tools and they're they're getting ahead of that, I will say, intentionally. Like, I can feel the intent there. And I think the intent is good but it's really annoying. It is all it is doing is creating more of the problem that people have seen of the Microsoft tools feels like this massive toolbox.
Mitch:Things can do the same things, but one is better than the other. Which one do I choose? Which is some of the reason why we exist. But now they're saying, let's add complexity to the AI side and now have people have to decide, can I use Copilot chat for this or do I need to go and chat with one of the Office apps? Or should I use Cowork for this Or should I use Microsoft Scout to do this for me?
Mitch:Which is the the step even beyond, which is not the topic of conversation today. But it's I'll say it's annoying. And let me connect it to our business for a second is we come from a world, an industry that is very used to selling time. And I relate it a lot to this concept of, well, you can get a token and spend a token. What they are introducing is the concept of having someone need to make a conscious decision every time that they prompt.
Mitch:Something like this to spend money. I just earlier today thought I I went and asked co work to do something for me, and I had a subconscious thought of, is Matt Dressel going to see however many credits I use on this thing, and am I going to hear about it, like, for good reason? And then am I going to have to develop some extra layer of thinking where I have to determine, can I use this tool? Is it worth me fighting with IT to use this thing versus not? Am I gonna get my wrist slapped because I'm using $40 of co work credits every month or something like that?
Mitch:I don't know. I think that's the reality we're headed towards. And I I find it to be annoying, especially in the concept of, like I said, our industry, we we sell zero time. We sell deliverables and we sell outcomes. And we we set for a fixed price, here's what you're gonna get.
Mitch:And for up until now, that's what Copilot has been. For fixed price, here's what you get, generally speaking. But now they're introducing this. You get a fixed price plus, run the meter up however high you want, we don't care, but we're going to start the ticker.
Matt:Yeah. I think the problem is that in a software model, you're not selling outcomes ever.
Mike:So there's a distinction there also between because the tokenized consumption based model, it's not that it didn't exist because it did. Like if you're building agents or using AI models for example, like you're tied to a consumption model And so you're gonna pay that way and so you're gonna pay based on usage. But those things I think the thing that we're struggling with is those things are more easily or clearly connected to a maybe a revenue target or a business process or a critical function of the business. I mean, you're like, well, I have to have that. I can justify it.
Mike:How much does it cost? If the cost goes up, am I generating more revenue? Like, if the math works out, the math maths. In this case, you're tying it to a big unknown, an ex a user that's an expense. Maybe they're a revenue generator but like, it's unpredictable because you don't really know how much they're gonna use it on any given day or any in it within any given month.
Mike:And so then, I think from a, like, IT management standpoint, you get to the end of a month and you look back and you're like, what is our copilot, co work usage and look at this person, they're going nuts. What are they doing? What are we we need to train everybody on, like, it's a whole new thing that somebody has to, like, wrangle.
Matt:We're gonna like, it's it's That's
Mitch:why your people only need Copilot chat and maybe they don't need
Matt:I mean, let's break it down though. So this is going to get a little a little different topic than usual. The Chinese models don't have this problem. Who knows why? Because they want my data?
Matt:No. Because they've been forced to deal with lower power, so they automatically choose cheaper models as a default, always. They've spent invested trillions of dollars to make their models work on lesser hardware. The Western models, Claude, GPT, all these things, they just build bigger. Well, bigger is more expensive.
Matt:Right? And so that model of just building more expensive is not sustainable. It's just not. Right? Eventually, you run out.
Matt:You cannot eat up all of the memory and all of the CPU power for the entire world in AI data centers for seventeen years. I know it's not been that long, but it could be that long in the future and not have something to pay for. Right? It is a it is a cost. Microsoft isn't necessarily investing in that same thing because they're not building models.
Matt:It's a lot of these They're they're not investing they're not investing in the same way
Mitch:that
Matt:No.
Mitch:No. No.
Matt:No. I don't I
Mitch:don't know that we know that, but we have a hunch that they are not.
Matt:I mean, you when you look at what their investment with OpenAI is, etcetera, etcetera, they've invested in other people. They're not their primary They want
Mitch:to be the system, the orchestrator. Correct.
Matt:Right? And so they have a problem. But this isn't unique to them. Because the other thing I would like to point out is you have the same problem with Claude. Claude, you only get a certain amount of credits.
Matt:And that's it. Once you run out of those credits, you're done for the month. Right?
Mike:You can get new ones. But the pricing models are different between those two. One is you you buy fixed price upfront, you run out, you're done. The other one is we'll charge you at the end of the month after we figure out how much you've used.
Matt:Not really. The next level for them is to pay for twice as much for a particular user to buy them the next level of their thing. And then beyond that, it's consumption. If you go to enterprise, if you go to the enterprise level, it's like they have it's all the same kind of thing. Everybody has the same problem.
Matt:Yeah. Is The difference is
Mike:Maybe I'm mistaken, but I thought there was a difference in that, like in the cloud world, there you you are stopped. You have to make another transaction happen to go beyond it, to consume more, whereas in the Microsoft model they will allow you to consume beyond
Matt:Until you until you get to the enterprise level. Yeah. Right? When you get to enterprise, it's the it's a different negotiate, whatever you're like, it's a whole different ball of wax. Right?
Matt:But in the enterprise, you're paying a lot more. Right? A lot more per month per user. It's a significant amount of money. And it's they've just handled it differently.
Matt:And if you think about it and you translate it in the co work, in the Claude model, you are paying that same monthly fee for the same thing. But the only model you get is Claude. Because in Copilot, you're paying the same monthly fee, you're getting access to all of the models at all the different levels in all the different ways, you just don't get co work. Right? Like, is an interesting there's can see why they're doing it, like what you were saying before, I can see why they're doing it and I can see that it is a valuable thing to do for the consumer if they can do it well.
Matt:The thing that I'm frustrated with is the non co work functionality isn't as good as I want it to be. Mhmm. If it was, I'm totally fine with it.
Mitch:Yeah. And I'm not trying to backtrack on what I said. I think what they're doing is justified, 100%. Like, they are investing a bunch. They're trying to make up some of that gap.
Mitch:It makes sense from a business scenario. Please, like, try to make up some of that gap. But at the same time, you guys were talking about, you know, Claude versus that. And I think some people would say what Microsoft is doing is a lot better because instead of having a fixed thing that I'm drawing from, I can just go however much I need.
Matt:And if, like And the credits are shared. So if I pay for Claude and I pay for 10 licenses, each person is getting their own threshold.
Mitch:Right.
Matt:Not I buy five licenses, Not this is my
Mitch:Not a So like I can in some ways I appreciate that of Microsoft. The thing that I don't like is the fact that it is creating another layer of processing for the
Matt:The end user.
Mitch:The end user.
Matt:Yeah. It's really interesting that you say that though. I don't know how many customers we've talked to that have been deep in AI. Like, we come to them and start talking to them and they're certain people, not the whole organization, but certain people, engineering people are deep into it. You know what they want to talk about?
Matt:How much does a token cost?
Mitch:Mhmm.
Matt:What's your what model are you which model are you using so that I can because I can buy these direct from so and so get a cheaper The value of a total. Like, they're already there. Yeah. They've been there for years.
Mitch:Yeah.
Matt:They've been there for seven, eight years where they're everything is about, oh, the new model, how much is it costing? Oh, they changed it by half a cent? Oh my goodness, that's going to cost me x amount, so I should change like, they've been there for a
Mitch:long time.
Matt:This is just the first time, not the first time, but probably a very prominent time that it's being raised up to the more regular user who's using things. And I I don't know whether or it's a net positive because everybody's going to learn about the the the cost of doing these things and the cost of your choices. They're like, we were just doing a training yesterday, two days ago, where they were talking about somebody asked about the auto choice, right? And some people were like, oh, I like to choose and I'm like, I don't ever want to choose. I wanted to choose the cheapest way that because I'm never going to be good enough to figure out what the right model is to do the I just want auto to work really, really well.
Matt:Yeah. And maximize my expenses, credits, whatever you want to call it. That's what I want.
Mike:That's interesting. I do think that something you just said, we're at a unique time when this is being now applied in at the user level, right, at the the The regular user. Knowledge at worker level, back office user level. And I think honestly that's probably one of the big reasons for the the timing of the price increase is because there is enough critical mass with the tool now where they've realized, oh, it actually can provide some real value, so we got somebody out to pay for it. Like, I think that's actually what is happening because if you think back, you know, certainly four months back, like, the tools weren't there.
Mike:Like, always I would struggle to create like a presentation or like a Word document the way that I wanted it to be and like all it just they was horrible.
Mitch:It's funny because I, at the same time, I think it still has quite a ways to go before it feels Good job. Yeah. In some of those But Yeah.
Matt:It's an interesting it's a very, very interesting thing and it's very interesting to think about a year in the future what that looks like. Because I think the reason why I liked the idea of using Cowork is because of some of the innate functionality to create workspaces where this all the content around the conversation gets stored and doing that not on a local machine because that's what how Claude co work works, right? And I think that that's really interesting and has some really powerful options, but then they gated it and it's like, I don't know whether or not I'm gonna miss out on some opportunity there because I'm too afraid of the cost that that might create. Yeah. And so there's a whole interesting thing about missed opportunity potentially where people could leverage co work to do really great things, but this model creates a risk reward scenario where I don't even want to try it because I'm worried about how much it's going to cost.
Mitch:Yeah. I think like my crystal ball is they have been rushing to get this thing out. They announced, here's our pricing model so we can start get the money trickling in. And then most people, by default, coworkers turned off in your tenant, most people are just going to leave it off.
Matt:Yeah.
Mitch:It's too much Too hard. To figure out beyond that, especially when people already don't use their Copilot license. Like the statistics on that is pretty unfortunate, which is why we try to help people make use of them.
Mike:Do think Microsoft will build in some tools to analyze the actions that your users have taken over the course of the last month and tell you how much that would have cost you in Cowork to have it just taken care of?
Matt:I don't know. I I they're doing a lot to
Mike:Like they do they do something like that in Power Automate. Right?
Matt:Yeah. They're doing a lot. They do a lot in their capacity model and stuff. But like you mentioned that, that's that was, I think, of my other frustrations. You know, we're in the TAP program when we learned about this.
Matt:One of the first questions was how can we learn how much we would have spent? Right. They didn't have an answer. They were like, well, the capacity reporting is going to come out in GA. And I'm like, what in the world, guys?
Matt:Like, we aren't going to be able to tell how much we would have spent in the stuff that we done have done already until after we have to pay for like, what in the world is going up?
Mitch:Yeah. Which Yeah. They give like some credit estimates of this type of task will take
Mike:They're this just building it while they fly it.
Mitch:Oh, which is For sure.
Matt:And that's Good. Okay. Yeah. I wouldn't go good.
Mitch:Okay. There is a bad version which is this has to be perfect before anybody can
Matt:Yeah. A 100%.
Mitch:Hands on.
Matt:A 100 a 100%. But it's yeah. It's a whole it's a very interesting turn of events. I can say, I think a lot of people are would be are are shocked and and will continue to be shocked a little bit that that is how it went. And not so much that that's how it went, but that nobody knew about it until it went GA.
Matt:It's like, why couldn't you say this in
Mitch:Oh, it was January. I'm sure it was internal politics.
Matt:Yeah. Yeah.
Mitch:Yeah. I don't have a piece of curtain there.
Matt:We don't have any idea, really. No.
Mitch:No. But I imagine there is opposing forces. Can we do this and make it still a $30 per user product that is really great that
Matt:Okay.
Mitch:Gets people in?
Matt:Okay. Let's talk about what we want. I I've been thinking about this. Here's what I wanted it Right? To wanted it to just pull from my Copilot because there is a limit to the number of times you can eat with the premium license, there is a limit to the amount that you can use it.
Matt:It's not all you can eat, but it's really, really Can
Mitch:you get it? Yeah.
Matt:No. It's really, really high. Yeah. Right? So, just pull from that.
Matt:And when I run out, then go to the consumption model. Right? Like, that makes if I'm sitting if I'm a user that's constantly using co work all the time and I'm eating through credits, okay, that's on me.
Mike:So are you saying they're they essentially are creating two two pools of tokens? Two levels, two tiers of tokens?
Matt:They they are. Yes.
Mike:Yeah. And so
Matt:They have for a long time.
Mike:Like And you're simply saying, if you want to use co working, you still have tokens in the first tier
Matt:Yeah.
Mike:Why not them? Use them. Yeah.
Matt:And if and if and if that uses three times as many, so then you're using them up so so much faster, okay. That's that's what you wanted to do. Okay. You you go go pay for the consumption based. Right?
Mike:And you're and you're talking about a pool of tokens that are prepaid essentially.
Matt:Yeah. I mean, I don't I've not actually and I'm gonna be really clear. I've not the licensing on Copilot is very, very, very, very, very, very, very complicated.
Mike:The Copilot premium license.
Matt:I know for a fact that there are credits associated to it. And for example, on things like agents, you can't just use a specialized agent, like a custom agent forever and ever for free. If you there are still limits in regards to if you wanna connect to third party data sources, if like, there are they've already got a complicated model is what I'm saying.
Mike:Well, you you can assign How
Matt:much does it add this to it?
Mike:You can assign, like, a specific agent, for example, you can assign credits
Matt:Credits to it. To your
Mike:environment and then the agent can use them.
Matt:But but if you are paying from a premium license, they don't charge those credits until a certain point in many cases. Like, you just get it for free because you're paying for the premium license.
Mike:Yeah. To me, it sounds a little bit like they have a deficiency in their premium license that first tier where they're not actually tying, like, they're trying to sell something at a flat fee, $30 a month per user, and they probably went way too high on the credits they allowed
Matt:Oh, that's that's possible
Mike:under that plan. Yeah. That's possible too. And now they're like, oh crap.
Matt:I I don't know why, but I can tell you from an end user's perspective, much like what you said, I could envision a world where you simplify this and it is way is way more palatable to everyone, which is to simply say, this uses credits, you get a certain amount, and after that, you gotta start paying. To get people to be able to use it, be like this this light, the weekly status update one looks like a 100 to 300 credits. Like, I should be able to do that if I'm paying for the premium license.
Mitch:Mhmm.
Mike:Yeah. Like, I see all kinds of logic and benefit to allowing somebody to use it at a small at at least an entry level Yeah. To whet their appetite and go, hey, this is really worth it. Let's go. Yep.
Mitch:Yeah. I'm a little bit cynical, but I imagine when they did they crunched the numbers on that, that did not make them very much money.
Matt:No. Of course. Because 90% of the people wouldn't fall into that category and then they wouldn't be paying.
Mitch:Yeah.
Matt:Yeah. I I like I under but just like you said, I understand what it is and that might be a result of what Mike said, which is like maybe they really should have come down on the what what Copilot Premium actually does across the board, but then that hurt a lot of people because they're like, we're still trying to sell that and now I'm going have a hard time. Like, I get I get it. I get that there's a bunch of competing things, but this is it's disappoint it's gonna create more people just because there's been a sea of people just doing co work, just doing Claude co work and being like,
Mitch:Oh, yeah.
Matt:It's just not good enough. Well, I'm going to just that's what my that's where I'm putting my eggs. And I
Mitch:think We have clients who say, I'm not fooling with Co Pilot. Yep. I'm just going to Claude. It's great. Yep.
Mitch:I'll pay for it.
Matt:I'll accept the risks and the additional administration and the additional, like all the other challenges with doing that and they'll just like, I'll accept it because because my people love it. My people want to use it and they find value from it. So, I don't know. It's a it is an interesting challenge.
Mitch:Yeah. I want to kind of start to close this up here and talk about how we think about it in our business. And I want to pose the question first to our let's call him our CFO. Mike, what what would you think if I told you I have some tasks that I like to do on a recurring basis and it costs $7 every time?
Mike:How many of those tasks do you have and how often is the insurance?
Matt:You guys are gonna get a peek behind the
Mike:He's asking the CFO who wants to crunch the numbers.
Matt:Well,
Mike:want it to
Mitch:run every weeknight. Every week.
Matt:$7 a day.
Mike:$7 a day? Yep. What's what is it?
Mitch:Do I have to like am I defending myself or are you just curious?
Mike:I don't know. For $7 a day, would ask the question of whether or not it provides value to the business.
Mitch:It does. Okay.
Mike:If it provides value to the business? Probably.
Mitch:Alright.
Mike:It's probably okay. But $7 a day for every employee.
Mitch:I think all of our employees should do that as well every night.
Matt:It's very interesting because I I know what he's talking about doing and
Mike:Oh, so your compliance officer already has an answer.
Matt:You you know too. You just aren't putting it two and two together.
Mitch:So $35 a week per employee.
Matt:That's incredibly expensive. Yeah. That's pretty expensive. It's incredibly expensive. Okay.
Matt:That has to provide it has to provide hundreds of dollars of value.
Mitch:Mhmm. Okay. I'm I'm playing like devil's advocate a little. I'm not actually vouching for that. I'm just it's interesting thought exercise, I think.
Mitch:Because that's what they're Yeah.
Matt:I know. So here's the thing is, like, if you told me that that reduced my that saved you a quarter FTE, no brainer. Right. If you're saying it makes me happier as an employee
Mitch:You see my tag
Matt:think that's team? I think that's a little harder to quantify for the for the bean counters and the compliance officers of the world. Right. Right? Mostly because you have that idea and you think everybody should do it, so does every other employee we have.
Matt:And we got 10 employee or 12 yeah, 10 employees
Mitch:13.
Matt:Effective well, we're gonna I'm gonna exclude us.
Mitch:No, we gotta get credit. We're at 13
Matt:Five, thirteen. Now I'm now I'm to 13 times 37 a day a week.
Mitch:Like we're about to be 14.
Matt:That's ridiculous. Multiply
Mitch:times one
Matt:more. Absolutely ridiculous. Right? Like, there better be some real value in that.
Mitch:Yeah. I saw But
Matt:that value can come from lots of places. Did that generate four new projects? Yeah. Did that like like, there's so it's such a that's the problem with this. And it's and it's really, really interesting because it's the problem of shadow IT.
Matt:Like, it's the same thing. Right? Yep. When I go and build my own personal system on the side to do a thing better and it makes my life easier and it costs me a certain amount of money, right, to do that thing either in time or in real licensing and I didn't talk to IT and I didn't have to justify it to anybody, that might be okay. But when you start to look at it in a broader context, maybe it's not okay.
Matt:Right? Like, what is the right thing? Mhmm. I would tell you for us, generally speaking, we're going to be figuring out how to manage and moderate our expenses because expenses are a challenge. And you can do a lot with Copilot itself and we're part of the TAP community.
Matt:And so I think the best when you think about it from a high level, our best thing is to encourage Microsoft to make the included Copilot good enough. Yeah. And leverage co work where it really provides value.
Mike:Yeah. I I think the the big, like, as I'm just thinking through and listening to that, the problem is gonna be Microsoft licensing for all of their other stuff, the basic stuff is fixed cost.
Matt:Mhmm. And
Mike:it's very easy to deal with a fixed cost even though now we have monthly subscriptions that we're tied into for years at a time, but it's still fixed. And so you can manage that from a business standpoint. You can make The minute that enter this level of variability into it, it, like, completely is a huge shake up and and I think that variability is gonna be the thing that like, people just choose Claude, it's because they have less variability.
Matt:I'm gonna I'm I'm gonna
Mitch:You can put controls in place to be clear. You can say But
Mike:now I have to think too hard
Mitch:about Yeah.
Matt:I'm gonna put on Mitch's hat. I'm gonna put on Mitch's hat for a second. This is k? Ready?
Mitch:It's gonna be great. Please pay someone.
Mike:Are you implying that Mitch only wears one hat?
Matt:Yes. You guys, all of you, both of you can only wear one hat at a time. So we pay someone effectively to fix a limitation of Microsoft's meeting recordings. Right?
Mike:Or we or we do ourselves To
Matt:make sure the meeting recording is on every meeting. Yep. To make sure it gets moved into the right location, that it's shared, etcetera. Right? Okay.
Matt:I the only thing that's stopping us from using co port co work to manage that right now is there's some APIs that Microsoft doesn't expose yet. What let's say that they do today. Right? You can get out the transcripts, you can get out the files, you can get out the like, all the things we need, you can get out them all. Right?
Matt:And I could, in twenty minutes, write a co works prompt that would every day or every hour even, run and figure out whether or I have any meetings that have meeting recordings and put them in the right spots. And let's say that cost $37.
Mike:I'll still go build a Power Automate to do it
Matt:if the APIs are there. Very interesting. There's no way I would. It's very interesting.
Mike:It's it's it's repeatable and structured enough to to warrant either an agent or an agent flow or a flow.
Matt:That's really interesting. And the cost is what's stopping you.
Mike:Yeah. Because I can control the cost. It's it's much more fixed. Just need maybe a premium license for Power Automate if there's a connector.
Matt:That's very interesting. I might
Mike:use Copilot to help me build that flow.
Matt:It's very interesting.
Mitch:As you can see, our thoughts are still evolving on this topic, but
Matt:I mean, they always will. Like, it's it's yeah. They always will.
Mike:That that co work thing, it also depends on where that co work thing gets built, right? Like, when we think about automation and personal automation versus organizational automation, if I can build that thing in a way that I know is gonna always produce a guaranteed result and not miss a meeting, not miss something Mhmm. Is always gonna work and it's something that will live long after the person that built it departs Sure. From the company. Yeah.
Mike:Like, we can talk but, like, I know I can already handle that well with Power Automate.
Matt:Very interesting. Very interesting.
Mike:Or workflows, like whatever the pick your flavor of
Matt:It just takes a lot of effort. The one can be done in an hour and the other one's gonna take four weeks.
Mitch:No. I think something that you said we have to wrap up. But something you said that was interesting to me is when something gets to the level of everybody should have their co work or maybe their Microsoft scout doing this every night, How soon does that become or graduate from an individual's
Matt:Thing.
Mitch:Personal assistant?
Matt:That's what he's saying. That's what he's
Mike:That's what exactly what I'm getting. It's kinda like not really co work.
Matt:Something else. Yeah. The problem is the problem is, again, that the way Microsoft gives access to data, it's very difficult to do in a system wide level. Many of these tools are individual and so you'll end up having the same problem.
Mike:But that is but that is is in itself a problem because then you have to like guarantee that it works the same way for everyone.
Matt:I know, but I can't fix that problem. It's not a
Mike:meme, that's Microsoft problem. But that's a question that you need to ask is like, how do I need this to behave? Does it need to be absolutely guaranteed every time? And if it if something fails, central office knows about it. Or is it just good enough if I give everybody the prompt and they can just make it work and follow-up on it and like, depends on what it is.
Mitch:We will report back and our thoughts further on this.
Mike:Can you turn on the cost tracking mechanism for all of the AI stuff that we're doing so we can report It
Matt:already is.
Mike:Great. I think something that might be useful
Matt:Probably bring a laptop into No. Could
Mike:three months from now, we could have another discussion about how much it How much cost
Mitch:did I spend today?
Mike:For co work and was it worth it? What did we do? Was it worth it?
Mitch:K. Everybody set a reminder for three years from now.
Matt:So Three years. Today No.
Mitch:No. Three years
Matt:would be three months in We're the AI gonna
Mike:have a whole different conversation about a completely new platform
Matt:from now.
Mitch:Next June is it June, right? June 18 Yeah. Of 2027, come back and see what we say.
Mike:Actually The podcast. Co work. Can you schedule a podcast recording on this We
Mitch:need to have co work listening in the office. We are digressing and that's when I know it's time to wrestle.
Matt:It's late in the day.
Mitch:Hopefully this is, if anything, a small amount of entertainment to hear us kind of fuddle through these concepts. And like Matt said, it is something that is constantly evolving. And so we know that hopefully we can stay one step ahead of you all so that you can take some of the breadcrumbs we're leaving behind and follow them as we attempt to create a a clear path for these things. But we wanted to give a little, you know, glimpse into what we're actively thinking about here at Bulb and how we're reacting to some of the changes in the Microsoft world. So that's all we have for today.
Mitch:We'll see you next
Mike:Good discussion.
Mitch:Thanks for listening. Thanks, guys. Hey. Thanks for tuning in to Make Others Successful. If you enjoyed this episode, we'd love to hear from you.
Mitch:Is a couple ways you can do that. One, we'd love for you to rate our show on your favorite podcasting app. And then if you have feedback or topic ideas or suggestions for future episodes, head over to bulb.digital/feedback and let us know. Your input is super valuable and we'd love to hear from you. Don't forget to subscribe and share this podcast with others who care about building a better workplace.
Mitch:Until next time, keep making others successful. I'll see you.
