Loading summary
Transcript52 lines
- [00:13]
A
Well, hello everyone. Welcome to our third Three Smart Guys podcast, sponsored by Directions on Microsoft, your authoritative source of all things Microsoft enterprise software, cost management and licensing. If you want to optimize your Microsoft footprint, come to us at directions on Microsoft.com. so I'm Barry Briggs. There's Peter O' Kelly over in my top right on my screen, and George Gilbert. Wave, George. And together we've got something like over a century. And I think the majority of that century is on me. Of hard earned experience in enterprise computing. And we are all veterans of the desktop computing collaboration wars. Some of you with gray hair remember that back in the 90s and the 2000s, I personally spent five years or thereabouts as the lead developer on Lotus 1, 2, 3, which if anybody remembers it now, they remember it as. Wasn't that, Wasn't there something before Excel? Anyway, the desktop wars were, as we all know by now, decisively won by Microsoft probably about 20, 25 years ago. The only real challenger today is Google Workspace. And I looked up just before this call and they have somewhere over 20% and they are growing in terms of enterprise market share. So something we'll talk about in the course of our conversation. But what I want to start with is the notion of word processing, presentation graphics and spreadsheet, PowerPoint, Word and Excel. We have become, come to the point where we have taken those for granted. They are the archetypes of communication for us today. Maybe before that were books started by Gutenberg, before that were chiseled walls, before that the oral tradition, illuminated manuscripts and so on. But Today it's Word, PowerPoint in Excel. And they've really gotten me a point where I'd say we just take them for granted. They're just there. Now, if you're a cfo, they're not just there because if you're a cfo, you didn't forget them. Because Microsoft frequently and maybe even capriciously raises prices and changes the licensing terms very frequently over time, over and over. But I think most C levels in it would recognize that Office has become infrastructure. It's cogs, it's the cost of doing business. You negotiated with Microsoft, but in the end you paid. But you know, I want to hear from you guys personally. These days I live in Claude and in Gemini. I, I even know senior Microsoft people. I was talking to one earlier in the week who say they start their day not in Office but in Copilot. So my question for you guys is, you know, are we at an inflection point? Is Office in trouble? Is it Gulp. Is Office dying?
- [03:30]
B
From, from my perspective, I think to your comment on the last 25, 30 years, I think that Microsoft Office and again, we'll just, let's just say Office aka Microsoft 365 until we get to the platform dimensions. And then I think it's facing its biggest competitive challenges since it vanquished SmartSuite and WordPerfect and the other early productivity app vendors. There's no question it's been an astonishingly profitable product line for a long time and as you alluded to, they still have dominant share in the most profitable part of the market. But I think there are a couple of things going on. One, I think in terms of overall seats, there are lots of reports that say actually Google Workspace has surpassed Office. Again, maybe not in the parts where everybody's paying for it, like education, individual users, small medium sized businesses and everything. So I think that that threat never really went away. But I think in the bigger picture, as you also kind of suggested, we're shifting now. We're shifting from artifacts to activities and there's all the talk about Agentic and things related to that. But I think increasingly people are saying I don't want to go launch this tool or this application. I want to get things done. And I think that is going to present a big challenge to the traditional Office applications. So incorporating AI is another thing. And we talked a little bit before in a previous episode about how that could potentially relegate the Office apps to a smaller role because what's going to be the center of the user experience for that activity for focused thing. So yeah, I, I do think that there's revitalized competition in this space. I think ultimately it's going to be very good for Microsoft customers, for the industry overall. And it'll be interesting to see how effectively Microsoft can leverage its resources to take things to the next level.
- [05:28]
C
Yeah, pick up on that. Just Barry Br Peter, when you talk about on the one hand the Google products have maybe surpassed it in unit share, if not revenue share and you're talking about tools to activities. We got a preview of this when Slack emerged and we had this new work surface. And then Microsoft responded because it wasn't a business model shift, it was just, okay, we'll add teams, you know, and the, the bigger issue I think now is we're seeing there's, there's a potential business model shift, like where you see a hint of it in Cowork from Claude, which Microsoft did their own version of. But the, the problem is cowork gives you just a conversational interface, but if there's an editing surface in there and then Cowork can drive the Office components with open source libraries on the open file formats, then they can work with the Office apps essentially without Office license. That's the business model challenge. And a user would still need the Office apps for now to interact with these artifacts that, that a Cowork type engine might, you know, agent might, might create. But you could see in this editable surface like a future version of MCP that already has a user interface. You could see someone building lightweight components that read and write these file formats that the user would interact with. And then you really are, have gotten around, you know, you've moved the apps to the periphery, then you've created alternative to the apps. You have your own container and you know, so it's a business model and a technology shift if you get that far.
- [07:39]
A
Yeah, you know, I will say, I just want to say, you know, as an anecdotally, you know, there's a lot of, you know, we're talking about a lot of visionary stuff that's, you know, years down the road. I was using Claude yesterday to help me build, build a slide, a PowerPoint slide. And sure enough, it built me a PowerPoint slide. And it was great. But what I discovered was it's its implementation of PowerPoint was, or its usage of PowerPoint was terrible. So I wanted it to draw an arrow between two boxes. Didn't draw an arrow, it drew a rectangle, filled in the rectangle and then put a triangle at the end of that and filled in the fill it. So when I went to, I went to adjust the arrow, none of the stuff aligned. So we'll get there, but we're not there yet.
- [08:35]
B
And George, just to make sure I'm on the same page with you there because there are a lot of different technologies and abbreviations that you ran through kind of quickly there. The best case scenario for Microsoft in the world that you just described is if co pilot pages becomes the center of activity. Right.
- [08:51]
C
Okay. Yeah, yeah, I'm, I'm talking, I was talking about the, the Claude Claude version, you know, and, and, and back ended perhaps by glean instead of the Microsoft graph, you know, for, for companies that have a lot of non Microsoft information they want to organize.
- [09:10]
A
Yeah, I mean I saw, I saw an interesting article this morning actually. It was, it was, it was from some guys of perplexity saying they hated MCP actually because, and my view is they were using it wrong, but they didn't like MCP because it was no substitute for just raw APIs. And they said, and you know, and I think there's a tendency among developers to think, oh, I should use MCP instead of an API. But there's clear nauseous use cases for both. I mean, if you want fine grained access to an application, you use an API. If you just want to know, hey, what do you know about this particular topic? And MCP is a pretty good choice now if you look at, if you look at. I'll just say one more thing about that. If you look at SharePoint's MCP server for example, it's got, it's got maybe 50, 60 tools in it, which very fine grained. And you know, frankly if you're going to go to that level, you should probably go at the API level.
- [10:11]
B
And I think for both of those things you have just said this is not the primary target audience for the next generation of productivity applications. I think if you go to somebody who is comfortable in Excel and PowerPoint and Word and say, hey, I got this cool new thing called MCP which is better than an API, they're going to say when are we eating lunch? I don't want to know that.
- [10:33]
C
But I'm talking about like mcp. Yes. I'm looking sort of down the road where I think Anthropic is trying to turn MCP into something that's not just a shim, that abstracts an API, but they, they want to make it, I think a component model. That's my impression. And that has a, that has a ui.
- [10:57]
A
It's very believable.
- [11:00]
C
Yeah. That they're, that they become the new container.
- [11:05]
B
But MCP is not prominent in that anymore. That's just an extension and mechanism for integrating with other tools. The focus needs to be on the other tools, not the enabling technology.
- [11:17]
C
Yes,
- [11:20]
A
I mentioned CFOs at the top of the podcast and I think both of you noticed that Microsoft announced Microsoft 365 E7, you know, where E7 has Copilot and Agent 365 built into it for the low, low price of $99 per user per, per, per month. And you know, I think nobody believes that they're actually going to have to pay $99, you know, per seat, per month, you know, in this. But that having been said, what do you guys think of that? I mean, I'll tell you my view, my view is it feels like they're flailing a little bit and they're trying to extract that last little 10% of the revenue market or the revenue space that's out there. And I'm not sure they're going to get it. I'm not sure they're going to get it. Any thoughts?
- [12:23]
C
Let me just comment on that because there's a big difference between your advertised price and your realized price. And, and we know like it's one of those lessons of software history that it's very hard to raise a price for a SKU over time. It's much harder to start with a declared price that you, you know, on sort of unofficially discount heavily. But as you add value to it, you can. Move back to your list price. It's a difficult way of saying like Microsoft heavily discounted Azure when they were trying to attract new customers and then when the customers renewed and they expected the new discounted price that Microsoft was like, well no, because you're on the platform now. And I mean that's just one way of looking at it. I'm not disputing, disputing that they're, you know, that they're flailing a little bit, but I think they're, they're, they're thinking, I would imagine is that they're going to add a lot more value at this SKU level and they want a price point that is a bucket for all that. Yeah.
- [13:41]
B
And I think there are a couple goals that might be in some respects at odds. One of them is maximizing profit, as you said. But another one is they need to get every enterprise as deeply entrenched with the whole collection of Microsoft 365 services as soon as possible because of, among other things like we were talking about, if there's a chance that Word and Excel and PowerPoint are going to get relegated to more like viewer editor component roles, then the importance of going out and getting everybody committed to the full content collection of services behind it. It starts to sound a little bit in the extreme like IBM with fud, fear, uncertainty and doubt where it's like, hey, you know, it would be a real shame if AI sort of burned your house down because you didn't have sufficient controls on it. And the only way to get sufficient controls is to go with the whole suite here. And I, and I think the, the last minute addition there in conjunction with the E7 launch of the anthropic partnership based Microsoft copilot cowork is kind of telling that maybe they're looking for additional compelling things to add to this to get people over the $99.
- [15:00]
A
One of the things that confused me a Little bit was my colleague, one of my colleagues, Lane Shelton says that, you know, they're, they're putting both platform and features into this E7 thing. And in particular, you know, Microsoft, sorry, Agent 365, which is a control plane for agents. Right. And to me, you know, I would have thought that the control plane for agents, which is effectively, you know, the entra ID level, you know, manager of what your agent, who your agents are managing, the identity of all your agents and what they're allowed to do and role based permissions and all that should go as part of the cheapest possible tier of Microsoft 365. So that no matter what your agents are always under control, no matter how you deploy them and on what tier you deploy them. And yet they put it in the most expensive tier, which I just found that very surprising.
- [16:01]
B
Well, I got to assume that their ambition is just like with Apple just works services, that everybody should just sign up for everything because it's the easiest. The historical challenge is getting everybody to get to E5. I think their ambition has to be to get everybody on E7. So I think they're taking the shot saying this is you need all of this. And when you realize you need all of it, it is cheaper to buy it with one E7 seat than it is to buy five or six separate services.
- [16:32]
A
I have to say that the whole thing reminded me a bit of the directory wars guys. Remember those, you know, the Notes name and address book and Novell and Active Directory and so forth. And you know, I think the big lesson coming out of that is, you know, if you own the enterprise identity, the Enterprise Directory, you own the enterprise. I mean, yes, it's kind of a, kind of a full stop that's it's so deeply entrenched. So you know, to me, the, you know, again, making these kind of identity identity level things as entrenched as possible in the enterprise seems like that would be the goal.
- [17:08]
C
But just elaborate for those of us who, you know, either didn't live through it or only have a vague memory of why that's the central leverage point is that because everything becomes an adjacency if you own the, the, the security, you know, for, for accessing a service.
- [17:29]
A
Yeah, I mean I think that I, I think the real challenge was the real realization that Microsoft had in those early days was that Active Directory was infrastructure. We thought of Notes as an application, right. And the Notes name and address book was part of that application. And so, and you know, but you know, that led to all sorts of clunky things like having to build directory, synchronization between the two and all that kind of stuff. And it was just so much easier for people to say if you have Active Directory, using Exchange is a no brainer.
- [18:00]
B
Correct me if I'm misremembering. Didn't Active Directory essentially get spawned out of Exchange? I mean, was it. It wasn't a standalone service at the outset. Right? It kind of came along with Exchange.
- [18:11]
A
That's right. Yeah. You're really forcing me to go back into some rarely traveled caverns of memory.
- [18:21]
B
I know I've got my land manager stuff around here somewhere, but Microsoft was
- [18:25]
C
originally trying to build a standalone directory before Exchange and I think that was the part of Cairo. But since they couldn't build Cairo, they needed a standalone directory and they took it out of Exchange. In other words, you write. Yes, they knew that that was central infrastructure for everything, but. But it took them two tries to get it right.
- [18:53]
A
Let's circle back to the present.
- [18:54]
C
Yeah,
- [18:57]
A
this is fun. And for our listeners, we do this a lot because we've all been around the industry for a long time. Let me just say what is in the news this week is Mustafa Suleiman. If you've followed his career at Inflection and Google and then two years ago became one of the CEOs and in parentheses, I track all of the Microsoft officers and it's just a list that just keeps growing and growing and growing. I Forget how many CEOs they have, but they have a lot of them. Mustafa was the CEO, or I guess he still is the CEO of Microsoft AI, but he seems to have had his responsibilities significantly reduced. What do you guys make of that?
- [19:48]
B
I think there are a couple. I think that combination of the Satya Nadella and Mustafa Suleiman memos that were published by Microsoft are really good case studies in corporate choreography. So it's like, yeah, yeah, this is all net positive and we did it on purpose. It's probably been in the plans for a long time. But I agree with you that it's sort of stands out when you say your job's just been cut in half or more in terms of the overall responsibilities he has now. It is possible and probable that being responsible for super intelligence and the frontier model from Microsoft is an all consuming job. So perhaps Suleiman did in fact say, no, I don't want to do this other product team cajoling stuff anymore. But yeah, I think it's a pretty big milestone and a surprise. So two years ago today was the announcement of the Aqua Hire from Inflection, as you mentioned. So yeah, I think this is something to watch. It's going to be a leading indicator one way or another.
- [20:52]
C
So let me take a different take or give a different take on this. No question that, you know, he's lost some product responsibility and he's focused now on the models. But I wouldn't interpret this as meaning that the model effort sort of is, it's flailing. Because if you go back and you listen to what Nadella said on the BG2 podcast, the first time he was on, it was really interesting. He talks about how distillation is like piracy. He's like, once features come out in a model, it's easy to kind of copy behavior, relatively easy. And he's like, know how kind of moves through the industry and with hires from, you know, vendor to vendor. And what, what I don't think he came out and, and said explicitly, but what I inferred was something that Gates used to say about Windows and Macintosh, which was, it's fine to have someone at the frontier and it's fine to have someone who's, you know, a year behind it. And that's going to be the mainstream of the, of the market share. In other words, in a world where you're going to have multiple models building all your in, in all your systems, you don't need to have the Frontier. You're not going to only have a Frontier model. You're going to have models that are cheaper, faster, specialized. And I think what they're saying is, and they don't want to say this because it's, it hurts the brand, but they don't need to be at the frontier. They just need something that's 6 to 12 months behind so that their cost of goods is zero. You know, the, the marginal cost of adding that intelligence is zero. They'll pay for the frontier intelligence where they need it, but they need something less for most use cases or for the complementary intelligence.
- [22:59]
B
I think that's a bet the company kind of proposition at this point and you know, they are saying that their ability to flexibly work with multiple models is a feature instead of a bug. I think if you work at Google and you look at like integration from the data center and the TPUs all the way up into workspace, and you say maybe, you know, some would argue that the race is really down to two players between Anthropic and Google. Anthropic, which Microsoft has an investment in and just did a product partnership with for copilot cowork. So I think that that's, it's a plausible bet. But if the bet's wrong, then they're going to spend a phenomenal amount of money and end up struggling to remain. You know, like look at Meta as a leading indicator there for how to waste tens of billions of dollars and then say, oops, it didn't work the way we hoped.
- [23:51]
A
Yeah, but George, George has a point which is that, you know, Microsoft has historically, you know, been successful being a fast follower. Right. And so, you know, maybe what they said is, hey, it's, it's, it's too hard being on the leading edge and it's too risky being on the leading edge. And to George's point, you know, maybe, you know, six to 12 months behind the leading frontier models is, is just fine. That's all I need for copilot and I want to come back to Copilot in just a second. But maybe that's, you know, that's all we need.
- [24:21]
B
Yeah, but I mean that gets back to the theme of the Microsoft good enough moat. And the good enough mode is not for them to determine, it's for their customers to determine. And if there's open competition and others are adding interesting features like Notebook LM or whatever, Nano banana, other stuff, Microsoft.
- [24:38]
C
Peter, I'm saying something slightly different. I'm saying any system is going to have a constellation of models in it, one of which is a frontier model and other models are on the performance and cost elsewhere on the performance and cost curve. And then there might be some fine tuned models where the frontier models you might not have a license to, you know, do RL training on them.
- [25:04]
B
Right.
- [25:04]
C
And I'm saying they want to, they want to get those other ones and they'll, they'll license, you know, or pay full freight for the, for the frontier model.
- [25:16]
B
So I think these are separable issues. One is that neural symbolic and the rest of it there will be multiple model types. The other issue is is there a commoditization of frontier models which are infinitely expensive and move very quickly. And I'm saying if they bet wrong on the second one, they have a problem precisely because their arch competitor, going back to the productivity application space as well as in the cloud space increasingly is Google. And Google happens to have one of those leading frontier models. And at least at this point Microsoft does not. Which takes us back to maybe that's why. Wait, wait, wait, wait, Peter.
- [25:52]
C
No, that I'm not being clear. I'm saying they don't need to build one of the frontier ones, all they need is access to.
- [26:01]
B
No, I hear you and I'm saying I'm disagreeing with you. I think that if one of your competitors is building it and can fully integrated at the deepest level across their product lines and that matters to customers for differentiating features, then being able to say we're multimodal and we can work with whoever you want is sort of like their database platform approach. Saying you want Mongo?
- [26:23]
A
Let me, let me jump, let me, let me jump in here. I want to switch gears. I want to talk about Copilot. Everybody's favorite topic and you know how I think about it is, and this gets back to our conversation about, you know, is Microsoft, you know, fast follower or on the. Or are they, you know, a first mover? And it seemed like for a while they were a first mover, right? I mean, you know, you know, on November 30, 2022, ChatGPT was announced. And you know, within a couple of months, you know, Microsoft had made this big deal with OpenAI and Satya Nadella had even saved Sam Altman's job. And, and you know, it looked like, you know, you know, Google was confused and AWS was and is, you know, flailing a bit and Meta is wherever meta is. But it'd be, you know, I think you're right, Peter. It's coming down to a two horse race, you know, and is, you know, I guess here's, here's the question though. I mean we're talking about, if we're talking about Co Pilot, which is what is the right user interface model here? What Microsoft has done is, I'm going to be somewhat harsh and say they bolted AI onto their existing applications. With Copilot, you still have Word, you still have PowerPoint, you still have Excel. We bolted them on. They had to teach the copilots all about the internals of each of those applications and so on. That's what they did. Alternatively, if you look at an application like NotebookLM for example, from Google, you don't start with any of the tools. You just say, here's what I want to do. You gather up all the information and then based on whoever your target audience is, you can build a spreadsheet, you can build a PowerPoint and the presentation metaphor becomes an afterthought. It's really about the research and the, and the report creation piece of this. And the output part of that is sort of the last step rather than being intricately interwoven with the task. I mean, do you agree with that analysis and do you think that which one's going to win?
- [28:43]
B
So I think we're partway along in this. So I think most people today, when they think about Copilot or they think about Gemini in workspace, they're thinking, help me use this artifact generating application better. They're not thinking today like, you know, there are some compelling demos that say, hey, go prepare for my management meeting, or, you know, help me catch up to all the things I missed and so on. And that's more, you know, building on the underlying graphs and so on. But I think today it's more about help me generate artifacts than it is about help me do activities that are beyond that scope. So I think we're still very much in early days. I agree with you, however, that it looked like initially it was, let's bolt AI as a feature onto everything instead of standing back and saying, how could we reimagine productivity and maybe accelerate this shift from artifact centricity to actually doing activities that are not one to one mapped to artifacts? I think if they had a first mover advantage there, they blew it because Copilot is not. Does not have impressive market share. And there are reports that even Nadella said back to his product team, this just doesn't work. So let's go back to the drawing board.
- [29:59]
A
Interestingly enough, I was using it this morning actually in preparation for this, for this pod today, and I was using it to write a document. And yesterday as I was using it, I thought, hey, you know, this is pretty good. It's actually, it was actually writing it, the writing the word document exactly the way I wanted, exactly the tone I wanted it. And this morning I went to pick up the work that I was doing and Copilot was down. So I don't know. That's just a, you know, Data point of 1, but I don't know.
- [30:32]
C
Well, this. When, when you're using Copilot, where are you using. Is there this task kind of frame that. That becomes the new container? Is that because that's what we're missing? Or you were you, you were. Were you working in copilot in word,
- [30:49]
A
Copilot in word and with a copilot pane off to the right. So what do you think? What do you think, George? I mean, do you think is office is off, is offices. Are offices. Is offices days numbered or are they numbered? Or, or does it have a. Or.
- [31:06]
B
Or.
- [31:06]
A
Or will. Or will office absorb AI and retain that? You know, the retakes. Retain being the primary metaphor, the primary archetype by which we create content.
- [31:20]
C
I I do think, I do think we'll see something that's more like a a blend of Notebook LM glean to organize things on the back end and something like co work with an editable surface where I mean Cowork and Notebook LM almost like somehow to merge because this idea that you have information core and then it can be presented in any modality is really compelling. But I think what Notebook LM is is missing is that give me the automation across, you know, across all these these artifacts. That's where, that's where cowork would come in. But but to bring it back to Office, I still think it's not going away anytime soon. It's the risk is they just get pushed more to the periphery and they become components, you know, and then the components are easier to replace over time. That's, that's the I think the, the risk
- [32:32]
A
we are at time Any any closing comments before we before we wrap up?
- [32:39]
B
One one brief one. I promise I will keep it brief. That I think is going to be interesting in this is I do think that the combination of AI with productivity apps may also lead to a revitalization of the power user database tools, both because the AI capabilities will make it possible for more people to actually use those features. And also having done that and gotten your like we were discussing in our last episode, getting your data house in order. Now AI can do a whole lot more with it, but that's another one where Microsoft really needs to figure out how they're going to marshal and kind of consolidate their customer value proposition for that. But I think Son of Access definitely a big task. Yeah.
- [33:24]
A
Anything else, George?
- [33:27]
C
Nope.
- [33:29]
A
Well, thank you all very much for watching and hope you enjoyed our monthly podcast. And again, if you have questions about anything having to do with Microsoft technology, pricing, cost management, licensing, come to direct us on Microsoft.com thank you all very much.