So What? Marketing Analytics and Insights Live
airs every Thursday at 1 pm EST.
You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!
In this episode of So What? you’ll learn about the different versions of Microsoft Copilot and how each functions. You will understand the critical differences between consumer and enterprise Copilot versions, helping you make informed decisions. You will discover the strengths and limitations of Microsoft Copilot compared to other AI tools available on the market. You will gain practical insights into using Copilot effectively within your organization while addressing common challenges.
Watch the video here:
Can’t see anything? Watch it on YouTube here.
In this episode you’ll learn:
- The different versions of Microsoft Copilot
- The limitations of Microsoft Copilot
- How it compares to other LLMs
Transcript:
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.
Katie Robbert – 00:38: Well, hey everyone. Happy Thursday. Welcome to “So What? The Marketing Analytics and Insights live show.” I can’t believe it’s already Thursday again. I feel like we were just here.
Christopher Penn – 00:46: It does feel like that.
Katie Robbert – 00:51: This week we are doing a walkthrough of Microsoft Copilot, which I’m actually really excited about because I’ve heard a lot about it. But because we are a Google shop and not a Microsoft shop, I personally have not had a lot of opportunity to see what’s inside Microsoft Copilot. So I’m actually really excited to see if there’s anything else that we should be thinking about what it can do, but also so that we can speak more intelligently about it to our customers and our audience. Because one of the things, if you don’t know, Trust Insights, for the most part, we are technology agnostic, meaning we don’t necessarily specialize in one tool or platform. We have a general understanding of as many MARTECH tools as we can so that we can help as many people as possible. Microsoft Copilot being one of those tools.
Katie Robbert – 01:47: So, Chris, where would you like to start today?
Christopher Penn – 01:51: Well, let’s start with the whole genesis for this episode, which really was every single webinar and keynote and stuff. We get asked this question, “Can I use Microsoft Copilot?” And it’s to the point where this is the stock slide that I now use in these talks about whether or not you can use Copilot. Copilot really is the Pepsi of gender vitals. It’s okay. It’s not.
Katie Robbert – 02:21: Wait a minute. Nobody ran this past me. I prefer Pepsi over Coke every day, so nobody asked me. I’m offended. I am personally offended by this slide.
Christopher Penn – 02:34: Well, it was completely personal. It was a totally personal attack. And one of the challenges with—well, there’s a lot of challenges with Copilot. They have the same problem that Google does with Gemini and that Microsoft has gone and named everything Copilot. So you have Microsoft Copilot Pro, you have Copilot for Office 365, Copilot for Office 365 Enterprise, Copilot for Office 365 Government Cloud, Copilot for Sales, Copilot for Finance, Copilot for Marketing, you have GitHub Copilot as a result, although I forgot Copilot for Windows 10, Copilot for Windows 11. It goes on. So like, because Google did the same thing, Gemini is in all this. It’s not the same. Gemini Copilot is exactly the same thing. Microsoft has treated it like Nutella tried to put it on everything and it doesn’t always go.
Christopher Penn – 03:30: So when people ask us about Copilot, they’re typically asking about two major versions. The first is the consumer slash consumer Pro version of Copilot that anyone can access. The free version is, as you can imagine, just vacuums up all your data that you give it and it’s used for training, and it’s okay. And then the Pro consumer version is $20 a month. However, this is not the version that most people are talking about when they’re talking about it to us. What they are talking about is the M365, the Microsoft 365 Copilot, which is a different animal. The 365 Copilot is integrated depending on your subscription level to a bunch of different Microsoft services such as your SharePoint, Word, PowerPoint, Excel, etc.
Christopher Penn – 04:29: Whereas the consumer version is not, believe it or not, the consumer version is out of the box, more capable than the corporate version. For example, there is no deep research field facility in the corporate version, there is a deep research facility in the consumer version. It has actions, which is this form of agents. In the consumer version, the actions don’t exist. They have Office 365 agents in the business version. And the model choices are different depending on your level of subscription in the Microsoft Office 365 version of Copilot or the M365 Copilot. So right off the bat, to get a sense of what capabilities you have, you have to be clear about which version of Copilot you’re using.
Katie Robbert – 05:21: Which like you started talking. I’m like, “Whoa, whoa. What is happening here? Naming products is not that hard.” I don’t understand why these enterprise companies—I mean, John, I can see you’re probably like, “Oh, this is, they’re doing it wrong.” It’s not that hard. Why do they get it wrong? Let’s rant for a second, John.
John Wall – 05:44: Yeah, well, it’s—when we were talking about this show and this came up, I was like, “Oh, he’s obviously going to be talking about Copilot for GitHub.” He wouldn’t be talking about Copilot from Microsoft. So like, what is this all about? And so I’m excited to get a look at the whole thing. But yeah, this is just like a classic branding disaster of, you know, everybody thinks everything’s the same thing and it’s, you know, I guess it’s the same as calling everything AI. You know, it’s just like, “Oh yeah, with AI is baked into everything. Just check out what we’ve got.” So, yeah, I’m glad we’ve got a guide to this dumpster fire. I guess that’s the short version.
Katie Robbert – 06:20: So if I’m following correctly so far, Chris, a lot of our clients will come to us and say, “Hey, I just realized we have Copilot.” But they are likely a Microsoft Shop with 365 Copilot, not the personal version, which is probably akin to a ChatGPT or a Google Gemini that you don’t have to have the rest of the product suite in order to get the free version of this. Whereas 365. And if we just want to, for the sake of the show, I’m just going to call the corporate one 365 and the other one is just Copilot. Is that enough of a distinction? No, well, no, you’re shaking your head. But I’m saying the one that’s baked into 365, when we’re talking about that one, I’m going to refer to it as 365.
Christopher Penn – 07:14: Okay.
Katie Robbert – 07:15: And then when we’re not, we’re talking about Copilot, which I understand—I’m fully aware that’s not the branding. I’m fully aware of that. That’s not what I’m saying at all. I’m saying for the sake of this show and us not to get confused, the one that’s baked into Microsoft is 365. The one that’s not, we’ll just call Copilot.
Christopher Penn – 07:35: Sure, sure. One of the other things is that the consumer version also has labs, which if you’re a user of Google’s Gemini and you have the Gemini consumer version, it has access to more advanced features than Gemini for Workspace because of the restrictions on the Workspace product. And here’s why. The 365 product has to be certified for enterprise, which means it has to pass minimum, some compliance tests, has to have the safety built-in that Azure customers expect, which are Microsoft cloud users that governments and healthcare facilities and all these different groups would need of any software tool. So as a result, the 365 Copilot is much more limited in terms of what it can do because it has to. It typically does lag behind the consumer version in features.
Christopher Penn – 08:40: And unsurprisingly for Microsoft, some of the features that are built into the consumer version, you have to pay extra for in the 365 version. So that off the bat makes it even more confusing. And then, as I said at the top of the show, if you want to build really good AI agents, you can build crappy agents right inside 365 Copilot. But you can also pay extra for Copilot Studio, which will allow you to do things like connect your agents to your SharePoint. And of course, as John mentioned, there is GitHub Copilot, which is owned by Microsoft, which is a coding tool. And then you have 365 Copilot customized agent versions for all of the different verticals that you might want.
Katie Robbert – 09:32: So these are their pre-baked agents. And then the GitHub version, I understand that third tab that you showed that is akin to—you showed this a few weeks ago, and I forget the name, but so obviously N8N does this as a third-party tool, but Google has their version of it. Opal, what is it called?
Christopher Penn – 09:59: Google Opal, yes.
Katie Robbert – 10:01: Opal, yes. So this is akin to Google Opal. You know what?
Christopher Penn – 10:07: Sort of.
Katie Robbert – 10:08: Well, I’m going high level here just to give people context. I’m not saying it’s one for one.
Christopher Penn – 10:16: But so like in Copilot Studio there are agents which are like GPTs, but then there are flows which are like N8N.
Katie Robbert – 10:23: Right, okay.
Christopher Penn – 10:25: And all that’s extra, of course. Okay, so what are the basic features available in the 365 version? Copilot, which is the one that most people are going to have, you’re going to have three basic sets of functions, four sets. You have agents, which are their version of GPTs or Gems or Claude projects, if you’re familiar with those. You have conversations, which are regular chats that you have just like you chat with ChatGPT. And then you have pages, which are an exceptionally confusing, terrible implementation of the canvas that you’re used to in ChatGPT or Gemini. And then you have from in here you have things like apps and stuff that you can create from, either integrate or create from your system. So those are the four major functions inside of the 365 version of Copilot.
Christopher Penn – 11:20: The thing that most people use most of the time is straight-up chat. Like, how do I have a conversation with this thing and it supports all the stuff you would expect it to? Can I upload stuff from my workplace? Can I upload files? Can I upload images? Yes, you can do all those things. And it can in general do most of the things that any other service does. However, one of the things that’s especially unclear, Copilot in the 365 version does not allow you to immediately select a model in the way that the consumer version does. So in the consumer version, you have three choices. You have the behind-the-scenes Microsoft fee, Microsoft v4 Think Deeper, which is fee for reasoning, and then OpenAI’s GPT5. They don’t tell you that.
Christopher Penn – 12:12: They just tell you fast, slightly smarter, and then actually smart.
Katie Robbert – 12:19: Wait.
John Wall – 12:21: Yeah, I was just like words, words.
Katie Robbert – 12:24: Not even that. If you go back to the consumer version for a second. So in the consumer version of Copilot, one of the models that I can choose is OpenAI’s GPT5.
Christopher Penn – 12:37: That’s right.
John Wall – 12:45: Okay, so you just said there’s a bunch of different flavors on there, but classic UI, like there’s no explanation of what any of those are. Basically, you need somebody to tell you or just get lucky on the roulette spin.
Katie Robbert – 12:57: No, and I get that because that’s the same problem with Gemini. My problem with this is that it’s OpenAI’s model. That Copilot is just like, “Hey, sneaky, let me get someone else’s model in here that they did the work for.” Like, this isn’t even ours.
Christopher Penn – 13:14: Well, Microsoft, as one of the largest investors in OpenAI, has—that was negotiated as part of that.
Katie Robbert – 13:21: Gotcha. It’s just like that one gave me pause. I was like, “Wait a second.”
Christopher Penn – 13:26: Yep.
Katie Robbert – 13:27: But I don’t need to use Copilot to get GPT5. I can use OpenAI to get GPT5.
Christopher Penn – 13:34: That’s correct. That’s correct. But you’ll see the same level of model confusion even in OpenAI. So if I go into ChatGPT, it now has Auto Instant thinking. So you can get, you get the same level of confusion. Like, okay, well, what should I use?
Katie Robbert – 13:51: Right. Because that’s only clear to the developers. Auto Instant and Thinking means nothing to a consumer.
Christopher Penn – 13:59: Right. Behind the scenes, Auto is a model router that chooses the model for you based on your prompt. Instant is GPT5 Mini, and Thinking is GPT5 Medium or GPT5. So GPT5 Medium because Pro is GPT5 High behind the scenes. You can’t see that in Microsoft Consumer Copilot. It’s fee for fee 4 reasoning and GPT5. And in 365 Copilot, you don’t get a choice. Microsoft routes it for you behind the scenes with the exception of the “try GPT5” button, which if you push when you turn that on now, allows you to use GPT5 inside the 365 version.
Katie Robbert – 14:39: Of Copilot, which of course is from a UX perspective on a different side of the screen than every other. Okay. Yep.
Christopher Penn – 14:47: Yes. Because they don’t really want you using that because it costs them money.
Katie Robbert – 14:51: Right. John, I don’t know about you, but I always go into these live streams thinking I at least stand somewhat of a chance of understanding what’s happening. But like this, we haven’t even done anything yet, and I’m—my brain hurts.
John Wall – 15:06: Yeah, no, the mess is all over the place. Well, and the one I really love is when they don’t let you pick. And so it’s basically like if you show up and you’re like, “Well, I’d really like some steak.” They’re like, “Well, you’re only at the hot dog plan.” Like, “I’m sorry, buddy, you’re not getting a quality answer for that.” So the fact that you don’t even have any control about how good or how deep the analysis is going to be, but that’s the trade-off too, right? Is you either have an unintelligible list of options or you have them just make the crappy decision for you. Those are the only two paths really on that.
Christopher Penn – 15:41: Mm. So let’s take a look at the difference between the two of these. We’ll give, we’ll—I’ll turn off GPTF GPT5 first. And we’re going to give this an admittedly absolutely awful prompt. We’re going to say, “Write a blog post about the importance of B2B marketing in 2025 with an emphasis on community building.” Right. So very simple, straightforward prompt that is going to deliver just absolute garbage. And that’s by design. So here’s what the Micro 365 Copilot comes up with. And we’re going to kick off the same thing with the quick response version in the consumer version and we can look at the two side by side. Maybe I think it just evaporated. There it is. Okay. “The new era of B2B marketing in 2025. Why it’s evolved, the power of community.” Four bullet points there, how to build it, examples, and final thoughts.
Christopher Penn – 16:43: And we go in this version, the consumer version, “Why it’s all about community. The shift, transactional relation, communities, competitive advantage,” lots more emojis, “metrics that matter. And final thoughts.” So those are functionally, reasonably about the same. I mean there, there are syntactic differences and stuff.
Katie Robbert – 17:00: Stuff.
Christopher Penn – 17:00: But they’re not wildly different.
Katie Robbert – 17:07: I can’t process that quickly, so I’m just going to take your word for it.
Christopher Penn – 17:12: Like I said, the most glaring thing is the consumer version throws a lot more emoji in it, which I find entertaining. Now, where this starts to get interesting is when you start to use—let’s start a new chat. We’ll turn on GPT5 and we’ll give it the exact same prompt. Now, let’s start over here, start a new chat, switch to GPT5. This should be almost identical because of the nature of the model. So why it’s the new growth engine gets three bullet points up front. Building, benefits building, and final thoughts, why it matters. Benefits building. So structural. There’s, there’s obviously four format differences. But you can see even in this version of Copilot you get all those emoji. You do not get that in here. And in this version, you get citations from where Copilot was trying to fish up data.
Christopher Penn – 18:09: You do not get that in the consumer version. So this one is automatically has a different set of system prompts that creates a different output.
Katie Robbert – 18:19: Okay, but again, not like 365 isn’t available to everyone. If I don’t have a 365 account, I’m not getting Copilot 365.
Christopher Penn – 18:32: Correct.
Katie Robbert – 18:33: Okay, I just want to make that clear because we’re showing them side by side and I just want to be clear that, like, you may not have access to the 365 version. Please stop scrolling. You’re making me very dizzy. The consumer version is the version that you would have access to as a non-365 user. And I think that’s where, for my own sanity, I’m like, “I just need to make sure I understand why we’re looking at the two together.” And one of them might just be forever out of reach for me.
Christopher Penn – 19:09: Right. So when people say, “Hey, can you help us understand how to use AI at our shop?” And we say, “Well, what do you have?” And they say, “Microsoft Copilot,” we do have to be very clear, “Okay, which version do you have?” One of the things that can be challenging with Copilot, depending on the version you have and your subscription level, is that it may make decisions based on what it can pull out of your SharePoint or other document requirements repositories first because it does have the ability to query your enterprise internally if you’re paying for those features. And so you can still, you can get even more different results based on the data you have.
Christopher Penn – 19:49: So one of the things that we always advise people to do with Copilot is make sure that the administrator has done a good job of sandboxing it as to what files it’s allowed to have access to and that the data custodian at your organization spend some time cleaning things out. Give the old SharePoint a spring cleaning because if you have, the Copilot completely just throws up with the door, say, “Hey, Copilot, you can have anything you want.” Inside our company, it may start pulling out data that’s 20 years old.
Katie Robbert – 20:21: That, that is a whole different episode because as someone who has developed in SharePoint almost 20 years ago, I can tell you that, yeah, the way that it’s set up, when someone sets up a SharePoint instance for a company for the first time, they have big aspirations, and this is the folder structure and this is how it’s going to look and this is where stuff is going to kept. And then fast forward six months and it’s a hot mess. We set up our Google Drive in a very clear way and my God, I cannot get anybody to follow said structure and there’s only four of us. I cannot imagine, no, I can’t imagine because I’ve been there how messy it is with an enterprise-sized company.
Katie Robbert – 21:15: A question that just came in is, “But it could still hallucinate with internal company data.” Yes, I assume meaning like if you set SharePoint 3—sorry, Copilot 365 on top of your SharePoint and say only use company data, it could still hallucinate.
Christopher Penn – 21:33: Absolutely, it absolutely can do that. In fact, depending on the quality of the data that you have, it might hallucinate more because under the hood it uses the Microsoft Graph, which is Microsoft’s internal knowledge base of your data, as its retrieval mechanism to figure out what data to pull from. So if your SharePoint is filled with junk, it’s not going to go well. So hallucinations are when a model tries to fulfill a request and it doesn’t have the information, it has wrong information that is statistically probable but not factually correct. Internal data could be worse than what you have externally. A real simple example would be if you, let’s say you have a blog and you have a bunch of blog drafts in your SharePoint somewhere and they all were binned. You’re like, “Nope, these are all terrible. We’re never going to use these.”
Christopher Penn – 22:29: If you didn’t exclude that directory of crap blog posts you were never going to use, and you came to your 365 Copilot and said, “Hey, let’s write a blog post about this,” and it semantically matched some of the stuff in your crappy drafts folder. You’re not recycling parts of your crappy drafts.
Katie Robbert – 22:47: Well, and it’s—that actually sounds like something that could be easily overlooked as, “Oh, we need to clean this out.” Because how many of us have received, you know, a document with the title, “And Then_ Version 1,_ Version 2,_ Final, No, Real Final?” And so. But where it ends from where it started is very different. But all of those versions live in the same place. They all live in, you know, SharePoint or Google Drive or wherever. And if we’re talking about SharePoint, nobody’s going back and cleaning those things out.
Christopher Penn – 23:28: Mm, yep. Yeah. So that’s chat. Pretty straightforward. The next thing would be agents. So they’re not available in, in this version. In the consumer version, right, you can set up actions, but you can’t build agents in the consumer version. In the 365 version, you can create agents, and you can do it in one of two ways. You can do it in the built-in Copilot Studio Mini Agent Builder, which is what you’re seeing here. This is kind of a misleading term. These are not agents. These are, these are just mini apps. They’re like GPTs. So if you’ve built a GPT like you, we built KDGPT once upon a time. That’s exactly what this is. However, there are some restrictions on this version. In specific, the knowledge section here requires publicly viewable URLs as sources.
Christopher Penn – 24:26: So in this version of a Copilot agent, you can’t upload documents to it, which—Right.
Katie Robbert – 24:36: The more I learn, the less I want to know.
Christopher Penn – 24:41: However, if you insert the credit card and you go over to the full version of Copilot Studio and you start a new agent there, the Cop—the full Copilot Studio allows you to add knowledge from things like public websites, your SharePoint, your data versus Dynamics 365, all the different knowledge sources within your organization. So you can in this version or upload files as well. So you can in the paid upgrade version of Copilot Studio, build that kind of agent that would match ChatGPTs, GPT capabilities.
Katie Robbert – 25:19: I would just as a quick plug, highly recommend before you start inserting the credit card and getting all the bells and whistles, run through the 5Ps to see if you even need it, because this is something that Microsoft is notorious for, is: “Okay, you’re a 365 customer, so you get Word, Excel, PowerPoint, and some other OneNote program that hardly anyone ever uses. Oh, but what you really need is Visio. Okay, now you’re going to have to insert the credit card. Oh, Visio doesn’t do what you want. Okay, let’s go find some other product that we have in our suite and you have to pay for that on top.”
Katie Robbert – 25:57: I really would encourage you to first run through the 5Ps to see if you even need agents the way Chris is talking about them, or if you’re actually just looking for, your custom GPT, your gem, your artifact, whatever it is.
John Wall – 26:16: Well, and so, and these are not agents as we’d seen them defined on other platforms where they go out and grab data. Right, like this. You’re still limited to what you put in there.
Christopher Penn – 26:24: No, these are GPTs. I mean, that they just. Microsoft couldn’t call them that because OpenAI kind of cornered that particular term.
John Wall – 26:34: Okay, well, that’s good. They took another term to flood it up.
Christopher Penn – 26:40: Exactly. So the third section is Pages. So pages are a really bizarre, not great implementation of the canvas. So if you’ve used Canvas in Gemini, if you use Canvas and ChatGPT, if you’ve used Artifacts in Claude, Copilot has Pages. However, the way they work is completely unintuitive and doesn’t make a whole lot of sense. So here’s an example. We’re going to go right back to where we started. Say, we’re going to say, “Write a blog post about the importance of B2B marketing, put the content in the page, and let’s have it do its terrible thing.” When you work with pages, it allows you to edit directly in the page itself, which can do in the others, but it doesn’t always follow instructions. So here it has decided instead.
Christopher Penn – 27:41: I’ve made a Word document for you, but you can, if you wanted to, edit that by manually adding it to the page and then you can manipulate it there.
Katie Robbert – 27:59: I started this episode saying I was excited to see what Copilot was all about and I regret everything. I want to just Men in Black my brain and forget I’ve learned any of this.
John Wall – 28:12: Because closing.
Katie Robbert – 28:14: Exactly. In all seriousness, it’s helpful to understand because it is so backwards from what everyone else is doing. I can now have more empathy for users of Copilot who are like, “This is all we have to work with,” and they’re seeing all the bells and whistles and shiny things of like Gemini and Claude. And that’s just out of reach for them. So this is actually incredibly helpful to get a better understanding of a question from a LinkedIn user, “Does Copilot 365 and/or consumer support projects similar to ChatGPT?”
Christopher Penn – 28:59: The answer to that question is, depends on your subscription level. If you have a Copilot 365 plan that includes Copilot notebooks, you can use a Copilot notebook and this account does not have access to that in a way that is similar to a project, but it is not the same. Copilot notebooks are kind of like imagine the least competent parts of Notebook LM with the least competent parts of a ChatGPT project. It’s like the worst of both worlds. That’s kind of what a Copilot notebook is.
John Wall – 29:36: I’ve got to run out and use that.
Katie Robbert – 29:37: Yeah, this is, I mean, this just gets better and better. All right, let’s keep going.
Christopher Penn – 29:44: And so those are the major sections. Pages, agents, and conversations are what’s available now in the agents. The one thing that is somewhat nice is that there are a bunch of pre-baked Microsoft-made agents that you can import that you don’t have to spend a whole lot of time configuring.
Katie Robbert – 30:05: And this is in the 365 version.
Christopher Penn – 30:07: This is the 365 version. So you have ones that might be built by, shared within your organization. You have pre-baked coaches and stuff like that. So you have some things that you could import. For example, there’s an Asana agent you can bring in, which by the way, agents, depending on the agent from the agent store, might also cost you extra money. You survive. And then you have Copilot variants that can live inside office apps. So if I were to open up Microsoft Word and create a blank document here, it’s me. And now I have a Copilot ins—a Copilot variant inside Word that where if you’ve used Gemini inside Google Docs, it’s identical. It is just as dumb as Gemini and Google Docs. There’s a version for this that works with Outlook that is just as terrible as Gemini and Outlook.
Christopher Penn – 31:16: So the Copilot here is different than the Copilot here. So as before, these are all things that are. They slap the label Copilot on everything, but it’s not the same models, it’s not the same infrastructure.
Katie Robbert – 31:39: What about images?
Christopher Penn – 31:42: Copilot does support the ability to create images, I believe. Let’s give it a try.
Katie Robbert – 31:48: Last and again, you’re looking at 365.
Christopher Penn – 31:51: Yes. So, okay, let’s create an image of a gray Newfoundland sitting at a modern office desk, reading the news and sipping an espresso. Now let’s see if it can create an image inside the 365 version. Here it comes. Sorry, I can’t do that.
Katie Robbert – 32:17: Well, what about the consumer version?
Christopher Penn – 32:22: I was going to say, let’s try the.
Katie Robbert – 32:23: Consumer version now with Gemini, for example, or even I think with OpenAI, there’s like a toggle that you can hit to say create image versus just giving it a regular chat. Like if you want to create something in a canvas, you want to create an image. It sort of gives it that little bit more of a direction. Whereas it looks like Copilot doesn’t have that kind of functionality.
Christopher Penn – 32:50: Not in the 365 version. That’s correct. It really does not want to do an image of a gray Newfoundland. I don’t know why.
Katie Robbert – 32:57: But in the consumer version, does it have that sort of.
Christopher Penn – 33:02: Not really, no. No, no.
Katie Robbert – 33:06: Okay. But that’s also helpful to know. Well, you can go ahead.
Christopher Penn – 33:18: You can tell based on the way that it is rendering this that it is using one of OpenAI’s image renderers. So there’s your sipping espresso.
Katie Robbert – 33:36: Yes, that. I’ll take it. That’s possible.
Christopher Penn – 33:40: That is passable. But 365, for whatever reason, just can’t do that.
Katie Robbert – 33:44: But that’s also helpful to know because that’s going to be a constraint for a lot of 365 Copilot users because all of these other tools. See right there, create an image. It is very easy UI, right.
Christopher Penn – 34:02: In fact, I wonder, just out of curiosity sake. Well, that’s different.
Katie Robbert – 34:09: I also, and this again, this is maybe a question that I would pose in our free Slack group Analytics for Marketers, has anybody noticed that Gemini, their responses have gotten a lot more positive and helpful and supportive? So that sounds like a delightful image. Here you go. I was using Gemini the other day and everything I said it was like, that’s an excellent suggestion. And it was like, “Wow, you’re really like doing some great things for the confidence.”
Christopher Penn – 34:40: One of the interesting things is, and this is kind of an aside, but this happened after ChatGPT moved from 4.0 to 5.0. A lot of people complained very, very loudly about it, about the change. They said it felt like, which is alarming, they felt like a friend had died and was being mimicked by a zombie version of their friend because it was no longer nearly as psychophantic as it used to be. People really like the whole being very positive and reaffirming and all that stuff as opposed to a very fact-based tone like, “Here’s your image, this is what you asked for. Here it is.”
Katie Robbert – 35:25: Interesting. Whereas, yeah, I mean I never noticed before that it was very fact-based, but I’m also sort of very much that way. So when it started being more like, “That’s an excellent suggestion, kiddo,” I was like, that’s different.
Christopher Penn – 35:40: Yep. So there’s GPT5 version of the same thing.
Katie Robbert – 35:46: Interesting.
Christopher Penn – 35:47: So okay, that’s the current state of affairs in the land of Microsoft Copilot. You have seven major versions based on Copilot: Copilot Pro, these are the consumer versions; Copilot 365; Copilot 365 for Government. For large enterprise there’s a security version. Then there is Copilot for Security, Copilot for Sales, Copilot for Finance, GitHub Copilot, and Copilot for Windows. So there’s so many of these different variants of Copilot and they all have different workings under the hood. The other thing that’s worth noting is that you have no control over the routing in 365. You can’t say you—other than the “try three,” the GPT5 button. You can’t control the routing, which means that you have to get very specific in your prompting to try and force the Copilot router to send it to the appropriate model.
Christopher Penn – 36:57: So if you have a prompt that you know is going to be a really challenging problem, you have to literally tell it. You have to think hard. You have to think very hard. You have to think extra hard as part of your prompt to get it to route correctly to the underlying model. And sometimes it works, sometimes it doesn’t work. But you’re not given those choices.
Katie Robbert – 37:22: I’m trying to. And I can’t type and think at the same time. Are there? No, see, I don’t want that. Are there? Is there like a single—I’m trying to find—is there a single page, a website, a whatever that’s like, “Here’s all the crap, the AI that Microsoft makes, it’s AI related?” Like, “Here’s the Microsoft parent umbrella.” And now I’m going into the product marketing of like, “Here’s like we have Copilot 365, we have all these different things.” Like.
Christopher Penn – 38:17: So yes, inside of Azure, which if you are an Azure person, you know this portal, extremely painfully you can see all the different Azure services that are AI related. So there is, there are literally dozens and dozens of these things that all this is a product marketer’s nightmare.
Katie Robbert – 38:38: Yeah, it really is. Because I was going to say, you know, does—and we don’t need to get into it now—but like, does Google have the same thing? Because I know there’s like Wisp and Collab. But they don’t all live in the same place. And that’s the thing that frustrates, I think, the common consumer is we don’t know where to look to know that these things even exist.
Christopher Penn – 39:00: Exactly. Yeah. So there is none of that. The closest thing that I can offer you is on the Trust Insights website, which I just put this up. When did I do this? I find the URL, I put it up. Trust Insights Downloads, and it is Instant Insights. It’s our Instant Insights page. But it’s a specialized version that has a feature-by-feature comparison. I just have to remember what I called it. Vendor Compare Copy or product marketing at its finest. Huh? Exactly. Well, I take after the best.
Katie Robbert – 39:44: He doesn’t mean me.
Christopher Penn – 39:48: So on our website, and we’ll put this URL in our free Slack group, you can see feature by feature what the different systems have. So for example, probably has its own browser called Comet. Most of the other services except Google do not. OpenAI has a browser agent, and literally no one else has that besides OpenAI. That’s, that’s Agent mode. And so this looks at feature by feature where what the different things are called.
Katie Robbert – 40:19: Which this is helpful because this at least gets you part of the way to know that it exists. But like it doesn’t then get you there. So for example, in—no, but if you just scroll so I can see all of—all of Gemini.
Christopher Penn – 40:36: Oh, okay. All of Gemini. There you go.
Katie Robbert – 40:38: Yep. So for example, I’m looking at this and you know where is Collab on here, right?
Christopher Penn – 40:46: Collab is not part of Gemini, but.
Katie Robbert – 40:49: It’s part of Google. Yes, and that’s my point is, so like, that’s the thing where I’m just like, this might be one of my least favorite episodes ever. This, this show has given me a headache. Not because of what you’ve presented, Chris, but because that this is the way it’s existing. And, you know, if we collectively as Trust Insights, if this doesn’t give us empathy for people who aren’t like, well attuned and in it every day. I don’t know what will. Because this, like, it’s a nightmare to try to navigate if you don’t have someone holding your hand to say what is it you’re trying to do? Okay, let me bring you to the right thing. Like that’s just not available to everyone.
Christopher Penn – 41:38: Yeah. So this is the other version which takes a feature by feature and says, “Okay, if you’re looking for a workflow designer, what is it called?” So this is in Copilot, it’s called Studio Flows. In Gemini, it’s called Opal. And the others don’t have that. Right.
Katie Robbert – 41:51: Another question, Chris, does 365 and consumer support custom instructions at the account level similar to GPT at the account level?
Christopher Penn – 42:03: In the free version, for sure it does not. In the 365 version it does. In here. So you have custom instructions and your Copilot memory. Again, it’s built very much like ChatGPT.
Katie Robbert – 42:23: What are your questions, John? Because I got nothing.
John Wall – 42:27: It’s funny you mentioned, you’re like, “Oh my God, this is horrendous.” The punchline for me is just like, “Hey, I’m good. Yeah, I don’t need this. I’m going to stick with the stuff I’m using. I’m ready.”
Christopher Penn – 42:37: I mean, that’s one of the reasons why it’s so challenging to work with these tools is they have so many variants and it’s a lot of the capabilities are not even going to be the same from institution to institution. So we ran into this recently at a client conference in the education space where different school districts have Copilot, but they have different IT teams that enable or disable different features. And so one person might be able to upload documents to their Copilot, another person can’t. And so it’s impossible even at that level to be able to say like, “Here’s how you do this thing,” because it just doesn’t work for everybody. So the long and short of it is that Copilot is what a lot of people have, whether or not they like it.
Christopher Penn – 43:29: And it’s one of the reasons why all these other companies are like, “Oh, you can also buy, you know, ChatGPT or whatever.” And a lot of people, we even heard from a client earlier today on a call saying they were paying for ChatGPT out of pocket personally because Copilot was so bad at their institution. And I was like, “Wow, with these tools, the gold standard format is still plain text.” So at the very least you can do text-based tasks reasonably well. And when they have access to models like GPT5, as long as you know how to prompt them well, you can at least get some stuff done. So of the seven major use case categories like extraction, summarization, rewriting, classification, application synthesis, question answering generation, you can do most of them with Copilot.
Christopher Penn – 44:17: You will have to jump through some hoops based on the restrictions that are built into your version of it, but you can still get benefit out of it if you have nothing else. And the one major reason to use Copilot if you are an institution that is working with sensitive or protected information, if your IT team has set up Copilot and configured it correctly, you can safely use protected data within your institution’s Copilot instance that you might not be able to with like a ChatGPT or whatever because it might not have the same level of compliance that a Copilot would.
Katie Robbert – 44:55: Get all that, John?
John Wall – 44:58: Yeah, again, a good question though, out of that is I’m like, I’m getting this feeling like I just don’t need to go over here or play around with this. Is there any situation where you’d be like, “No, you actually, for this is one place where you should come over here and mess around?”
Christopher Penn – 45:13: There’s literally nothing in Copilot that someone else doesn’t do better.
John Wall – 45:17: Okay, so in other words, if it’s in your org and it’s free, that’s why you’re using it, right?
Christopher Penn – 45:22: Yeah, if it’s any org, it’s free or if it’s in your org, your org requires you to use it for data security practices. That’s the reason to use Copilot. If you have a choice, anything else is pretty much a better choice.
Katie Robbert – 45:34: Okay, all right. I think that this is incredibly helpful though because what we do know about the marketing audience or basically anyone who works in an enterprise-sized company, the likelihood of them being asked to use Copilot is pretty high. Enterprise-sized companies tend to be Microsoft shops. They kind of go hand in hand. So it makes sense that we collectively as Trust Insights. But also, our audience should get at least familiar with what it is so that we can help support those who don’t really have the choice they have to use it. And they’re seeing, this whole AI world go by of this is everything that’s possible, so we can at least have more empathy for those who are being asked to use Copilot.
Christopher Penn – 46:25: Exactly right. So that’s it. If you have comments and questions and stuff, pop on by this the Slack group, but otherwise we will see you all next time. Thanks for watching today. Be sure to subscribe to our show wherever you’re watching it. For more resources and to learn more, check out the Trust Insights podcast at TrustInsights.ai/TIpodcast and our weekly email newsletter at TrustInsights.com. Got questions about what you saw in today’s episode? Join our free Analytics for Marketers Slack Group at TrustInsights.ai/analyticsformarketers. See you next time.
|
Need help with your marketing AI and analytics? |
You might also enjoy: |
|
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday! |
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new episodes every Wednesday. |
Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.