Revolutionizing Daily Tasks and Business Strategies: Insights from Dynamics Minds in Slovenia

Mark Smith: 0:00
Welcome to the Power Platform Show. Thanks for joining me today. I hope today's guest inspires and educates you on the possibilities of the Microsoft Power Platform. Now let's get on with the show. All righty, we'd like to welcome everybody back to the Ecosystem Podcast. A little different today first time in the history of the show, we're all together. Welcome, welcome, hi, welcome Hi. So we are your hosts Mark Smith, andrew.

Andrew Welch : 0:33
Welch William Dorrington.

William Dorrington : 0:35
Anna Welch. Chris Huntingford Hunting Ford, ford, ford. He hunts Fords. We're coming to you.

Mark Smith: 0:41
Wow, that joke has never been made. Behold sports, we're coming to you.

William Dorrington : 0:43
Wow, that joke has never been made. Thanks for the wise words.

Mark Smith: 0:49
We're coming today live and, by the way, if we slur and stuff, it's not because we have drunk anything, because we haven't. We have generally been in transit for the last 24 hours, some of us worse for wear than others, but anyhow, we're here at Dynamics Mines in Porto Rojo. I can't pronounce that, so hence it's been pronounced, and so we decided why not? Let's try a podcast. We've never used this equipment before. We've never used this microphone before or that one. What could go wrong? What could go wrong, but to set the scene. I'm going to hand to Andrew. Go wrong, but to set the scene.

Chris Huntingford : 1:25
I'm going to hand to Andrew Well, so for this episode we have some product placement for Ryanair. Chris Huntingford would like to speak about his experience getting to Dynamics Minds to the conference.

William Dorrington : 1:40
Let me tell you I will never, ever, ever, ever, ever, ever, ever ever fly with that terrible, appalling excuse of an airline ever again.

Chris Huntingford : 1:52
How close were you to being arrested in the airport yesterday, Dude people?

William Dorrington : 1:56
saw the fury. It was actually so bad Like they booked me on three incorrect flights, dude three. I had to go through security twice. It was wonderful they were all delayed and then they wait. The best part is I got left at a services station by the cab driver.

Andrew Welch : 2:15
At the.

Mark Smith: 2:15
Italian border. So just to put this in perspective, I took the same flight earlier in the day. It took me two hours, right as in destination to destination. You flew Brian Air. No, I said I did the same flight, different Chris set out at 10 am and didn't get here until 3 am the next day.

Chris Huntingford : 2:37
I love when Mark says the letter M and also the word dead.

Mark Smith: 2:41
Can we? What Dead? Why did I say M? M? Yeah, all right, also the word dead. Can we? Why did I say im?

Chris Huntingford : 2:45
Im All right. So for those of you who don't know, we're at the Dynamics Minds conference in Porto Roche, Slovenia. We've got I don't think they were expecting this many hosts on the same podcast in the podcast room, but this is.

Andrew Welch : 3:01
A beautiful backdrop.

Chris Huntingford : 3:02
With this beautiful backdrop right here. This is a beautiful backdrop, with this beautiful backdrop right here. But otherwise this is. I think this is the best Microsoft community event of the year it's got to be. I mean the, the.

Andrew Welch : 3:12
Yeah, I'd say ColorCloud in a different way was fantastic, a lot more intimate, a lot more sort of reminding me a bit more of the origins of the community. But yeah, I've never been here before and it's. It's another level, it's blown my mind, quite frankly.

Chris Huntingford : 3:24
Yeah, the energy at this conference right, and the people wandering around and it's the only time I've ever seen sort of an exhibition floor with the displays and the sponsors who are here. That's lively and people chatting and people getting to know one another. It's a fabulous event.

William Dorrington : 3:41
It's amazing, right, and people have made so much effort, so it's not just hey, here's my stand and here's my company. They've got like theme things going on and wizard hats and flaming Ewoks with doom swords and all sorts of weird cool things. There's a flaming Ewok. Yeah, it sounds great. It just sounds cool. Man, we should put some pictures actually on the website from this.

Ana Welch : 4:03
They do have a photographer, and that's amazing as well. We all look way prettier than we are in real life in these pictures. Honestly top quality stuff.

Andrew Welch : 4:13
He's stunning.

Ana Welch : 4:14
That's true. The amount of technology, though and yesterday Andrew and I did an ecosystems one-on-one course was the first of its kind. We learned a lot, mainly that it cannot be done in three hours.

Chris Huntingford : 4:32
It turns out it is very difficult to squeeze introduction to ecosystem-oriented architecture into three hours.

Ana Welch : 4:39
But what we did, I think, was structure the whole information in like six major principles, and that worked really well and it felt like people went away, like walked away with something. Plus there was a lot of like pre-reading and stuff like that.

Chris Huntingford : 4:56
I was particularly impressed with the kind of quality of the report outs at the end of the group of the small group exercises right. So what we did is we broke. We had a class of 15, 18 people, something like that was that many, it was a huge like five people. No, we had we had. We had 18 registered and 15 turn up.

William Dorrington : 5:18
Um never a good story of facts, guys guys.

Chris Huntingford : 5:23
At least 50 people At least 50 people in the class right. We broke the class up into three groups and we gave them each a couple of scenarios, and the whole idea was that each group was playing the role of an ecosystem architect. So we had a scenario that we would give them about hey, your organization is trying to build a data platform, or your organization is trying to modernize a bunch of apps. And then each scenario came with three proposals and the idea was hey, an engineer or someone on your team has made these proposals. Which proposal are you, as the ecosystem architect, going to recommend to the CIO or are you going to make your own proposal? And the report outs from the group activity was. I thought they were very strong.

Ana Welch : 6:08
So we recommend it, especially if you have, like, a very hard problem to solve. You just do a workshop, you know, and get a bunch of clever people from around the world in Slovenia to try and solve your problem, because those were some really clever people around the world in. Slovenia to try and solve your problem, because those were some really clever people and I think my biggest takeaway was the fact that actually there were two, but the number one takeaway was the fact that people are really interested in ecosystem architecture and things are moving that way, for sure.

Chris Huntingford : 6:42
Yeah, chris, how was your panel today?

William Dorrington : 6:44
Oh, it was awesome. So I did mine with um fast. Oh, and Victor, they just they're just fun right, Like they just read and it's quite cool because faster is very SharePoint, he very technical, very Debbie, and then you just include SharePoint and technical. Actually, yeah, I know it's normally a sinful thing for me to do. Victor's got great PowerPages experience and he's done quite a lot across the stack. And then I'm obviously exploring the Purview side and everything and it was cool. We got asked some really funny questions around data. So I used your story around Dataverse hydrating the data lake, or one lake, and how the go-to data platform of choice for me right now for transactional data and relational data is Dataverse and why it's important and also why that plugs into Purview right, Like the data mapping and all the security available there. So that came out right and I've actually noticed it's coming out in pretty much every discussion I'm having. So, yeah, especially the ecosystem side. So, yeah, it was good, fun, man.

Chris Huntingford : 7:45
Well, power Platform right now is a. I think that Power Platform is all a data story right now. Yeah, that has been for a while. Yeah, it has been for some time, but I'm still astonished how few people, actually, I think, look at it that way and talk about it that way. But you know, to me, the modern data platform at this point is very reliant on a combination of fabric and power platform, and the reason for that is because of Dataverse. Dataverse is the story here.

William Dorrington : 8:09
Yeah, and what's wicked right is that? So Tonya, one of the ladies in the audience, she's been to a couple of conferences and she always comes to the sessions and she was in one of the governance sessions I did like in Germany and I love the way she was thinking about it. She's like, yeah, this is about data. We should be really thinking about it that way and she said that in the audience. I thought you know what? That's great, because the message is resonating really well.

Andrew Welch : 8:31
It's the key way to leveling up Power Platform in general. If you look at the way that we've gone and the evolution of applications and the complexity and reducing that complexity and reducing it, and reducing it and reducing it, the easiest way to explain the Fabric and the synergies between the two is that's what they've done with data. Right, data was still incredibly complex. They still had to set up a shed, nearly slipped on a shed of infrastructure and various integrations, et cetera, et cetera, et cetera. Fabric removes that now. So you get that democratization of data which allows you to create truly powerful applications. Yeah, and that's why we're seeing that creep up more and more.

Chris Huntingford : 9:09
Because you want to level up, you want to get to that enterprise scale applications, you need to plug it into fabric I'm still hearing, though we're a year on, so I think fabric I forget the exact date, but it was released basically a year ago today or a year ago last week and it went GA yeah, I think in November. But I'm still surprised by how many organizations from an investment perspective and how many architects are still hesitant because they feel like, oh, they want to see the technology be proven. Folks, fabric is a mature technology, built on mature technology. There's more capability. I think that for a long time, for a year, we didn't have the ALM, the Application Lifecycle Management story in Fabric that we wanted. But go watch the build recordings from last week.

Andrew Welch : 10:00
But there's something that comes up all the time, and there's always something that comes up Any new tech. There's always a common theme of something, and fabric seems to be cost at the moment, and I think it's because it's brutally expensive, but is it? And the reason why I say but is it? Because we all know it is, but is it? So? It's democratized it to a point that it just looks like it's ingesting your data.

Andrew Welch : 10:26
You don't see all the infrastructure it's spinning up in the background that you'd normally pay for and you go look, we've had to spin that up and we paid this much. And then this, then this. It's doing that for you, but all you're seeing is the consumption of data flowing. For you know what this is, just capacity. It's a lot more than just capacity. It's all your sparklings, is everything else that goes with it, and I think that's what people miss out, and scott sue does a brilliant explanation of that and why people perceive it to be expensive, when actually, if they really see what they're getting because they're not seeing behind the scenes they go actually, that's reasonably priced.

Chris Huntingford : 10:48
Let's get a shout for Scott Sewell. Follow Scott Sewell, if you're not already.

Andrew Welch : 10:52
He's Mr Fabric himself. That's what I call him.

Ana Welch : 10:54
Yeah, absolutely. I guess what I want to point out is also the fact that people miss out on the manpower that it would take to build all that, and more than the actual skill to build all that. Try and convince a bunch of data people to talk to and work with a bunch of power platform people and then with some devs, and see how long it will take you. Fabric not only hides all of that work away from you, but it makes it easier for everybody to work together.

Chris Huntingford : 11:28
I think this is a good lead-in to the work that Mark has been doing on the. What is it? The fallacy of free? Mark is here.

Mark Smith: 11:39
|Yeah, this is me, you're welcome. Yeah, so the fallacy of free, no, and so this this is a really interesting concept, right, and that it is that people think that if they do more and they add more bodies and there's kind of no factoring of the owner, the cost of those people, so we just add another person to the project, it adds up and this whole concept of the total cost of ownership, people. So we just add another person to the project, it adds up and this whole concept of the total cost of ownership never gets recorded. And so, rather than using a best-in-breed like uh, managed, managed environments, for example, and everyone's like, oh, yeah, it costs premium though, but like, does it really? Like? Yes, it does, but does it?

Mark Smith: 12:21
What I'm saying is that if you mix that in the the total cost of your ownership of the project, the delivery, the outcome, let alone let's not talk about governance, security and everything else you get with it. It's a no-brainer when you compare a real apples with apples scenario, as opposed to what we're just incrementally adding to projects over time and not recording against it. And it applies in the same example that you've just covered. We've just come off Build, a week of Build and yes, I'll say, the word co-pilot was mentioned once or twice, a few times, and with that the word AI. How far into this did we get without actually?

Chris Huntingford : 12:57
saying AI Not too bad.

Mark Smith: 12:59
We've been recording for 14 minutes, but one of the interesting things that comes up is that a lot of people have a lot of apps built on SharePoint and I don't think they're actually aware that all that data sitting on those apps have now just been made available to the office graph. If you're starting your AI journey, any thoughts on that?

Chris Huntingford : 13:52
Yeah, this is something that we've been talking about for a long time, right? Why is premium power platforms such an incredibly good investment? And I think that you know we can make. We can make such a case around the value that you extract. And if you don't believe us, go read the Forrester report from maybe a year and a half ago.

Chris Huntingford : 14:13
But one of the things that I think has been underplayed is the risk that organizations take on when they over build, using free capabilities or capabilities. You know the basic power platform capabilities that come with Microsoft 365, right? So the phenomenon here is that a lot of these organizations they we've already got SharePoint, we've already got these basic power platform licenses. We're going to build these apps on SharePoint and, by the way, this predates power platform. There is a whole generation of the SharePoint app, but the security model in this has often been security by obscurity. Right, our data is secure because no one knows where to find it. Well, congratulations. Ai knows where to find it. And if you've overbuilt these applications in SharePoint, right, as Mark said, that data is now wired into the Microsoft Graph and that data is now wired into the Microsoft graph and that data is now hydrating. M365 Copilot.

Mark Smith: 15:09
And you ought to think about that data, because we talk about PI, or PI personal information as opposed to personal identifiable information, right, pii? And there's a distinction here, because you might think that you're not presencing that data of that individual, but the thing is so many data points are coming together that it's quite clear who it is. I heard of a case recently in a police situation where they want to put the badge number in of an individual. We're not putting their name or anything, but that's identifiable information that AI can light up if it has access to other data sets in your network. So I think you've got to be really careful in that the solutions that you might have built you might. Yeah, it doesn't have somebody's name in there, but is there enough information that pulls together to make it clear that other colleagues know who you're talking about?

William Dorrington : 15:59
Oh, yeah, yeah, it's interesting, right, because I've said to people all the time.

William Dorrington : 16:03
I say this in my session now I'm like ai acts as a loud hailer for badly treated data absolutely it really does, and if you're taking relational data or transactional unstructured data and then contextual data and you're welding those together through an ai platform, all of a sudden, that that data has got superpowers, and if you don't respect that data enough, man, that's, that's on you, right, as a as a business. It confuses me so much because, like I had this conversation with you guys already Like I have a social media profile. I don't let everyone access my social media profile. Like I have protection in there and there's certain things that I don't want people to see. Like I obscure my addresses and things like that. But that's me. And if data is a digital representation of you, your business and your affiliates, why do we not give a damn about it? Like, why do we just say, oh, it's too expensive to do this?

Andrew Welch : 16:49
But yet they will say it's the most valuable resource they have. Every time I'm super confused People.

William Dorrington : 16:58
I'm just extremely confused by that. And then you know you're given these amazing tools by Microsoft to actually solve the problem, like tools such as Purview and Sentinel and all of that stuff, but we're like, no, it's far too expensive to pay that license. We would rather have our data not looked after. Yet it's our most valuable asset. I am completely, am I just really?

Andrew Welch : 17:16
stupid. And then you get a 4-3% turnover leak thanks to GDPR. Well done.

William Dorrington : 17:22
Is it just me, though? Am I missing something here? No, no, no. To me it seems super obvious.

Mark Smith: 17:27
I think it goes back to what Andrew said, that you had a really good term for it. Then it's about you'll have to correct me Security by obscurity, security by obscurity. I think that's the issue in most organizations, because no one's been able to access it to date. But you know, with these language models, right, you know, in that scenario I can say listen, what's the probability of this being XYZ employee?

William Dorrington : 17:51
Because it's now got all that data right.

Mark Smith: 17:53
All of a sudden you can light up those scenarios which might have been hard to write the query. But for an LLM, writing that type of query is going to be a cinch.

William Dorrington : 18:07
Yeah, I agree, I think, now that we have this ability, it kind of terrifies me a bit, but also I think it's forcing organizations and us to actually respect the data that we have, because right now agents aren't out yet completely. We have a prompt box that we talk to, but in two years' time we may not even have a prompt box. It will just do things right. And that's the part that freaks me out the most is that, if I know my data is not being looked after in a business and all of a sudden I have a bunch of agents or AI running around in the background just using my data like an RPA would have, I guess. Yeah, that panics me a lot.

Chris Huntingford : 18:35
Can we talk a little bit about, because I think that most folks may not have yet heard about or thought much about this idea of agents, and there was a lot of talk about this at Build, so does someone want to take that on?

Andrew Welch : 18:47
I just wanted to comment just before it goes, though, like we always, we're very good at explaining the concerns and the negatives that that brings, but actually there is a positive to all this, which is it's going to drive change. It's going to drive better data hygiene in the long run. Otherwise, the companies, the organizations that most need, for example, co-partners from M365, won't be able to adopt it until they've done it. So they're going to be forced to do it, and that's a good thing in itself and that's what comes with change. So I just wanted to point that out. So there is a positive spin on it.

Ana Welch : 19:17
Yeah, no, absolutely, and the only thing that you need to make sure is that you're not one of those organizations who aren't actually, unfortunately, looking after their data.

Chris Huntingford : 19:34
I mean, this is the AI strategy right White paper, right there. You know those organizations that are getting kind of where many organizations that are investing here are pulling away from the pack, but the ones that are not are not going to realize the error of not investing right now until it is potentially too late for them.

Mark Smith: 19:49
So when Andrew talks about the white paper, you can find that on Cloud Lighthouse, which leads me to I'm going to come back to agents. But why do we take this moment to say do you want to welcome me?

Chris Huntingford : 19:59
I'd like to welcome Mark Smith as a partner at Cloud Lighthouse Boop boop. Oh, finally, the industry's worst kept secret he's official.

Mark Smith: 20:09
I'm official. I'm official. Anyhow, back to agents. Tell us more about.

Chris Huntingford : 20:16
Mark.

Mark Smith: 20:17
Tell us more Honestly. I think agents is the most honestly enough, it is the most exciting thing I see in the future as in. You know, I went to a session at MVP Summit and it was the R&D team at Microsoft and how they were thinking about this concept of agents and they had some demos that were. You know, let's say, you took a C++ agent that was just an absolute guru on C++. Did you ban the word guru on the podcast? No, gurus, okay, it was the preeminent expert on the topic, preeminent expert Of coding in C++, right. And so you know, like we've known for years, a developer should never be able to critique their own code as in as the final release, right, you should always have a tester, qa, that type of thing is natural. Then you have an agent that's an absolute, preeminent expert on on testing c++ code and these two agents, one hands the code to the other, the other validates it, hands it back, but then you take out the whole lag of communication, how quickly that solution can be developed out through multiple could be 50 iterations in one hour as an example, right, which you'd never get. That kind of time in the real world and that speed. See, I gave you that example that they gave there.

Mark Smith: 21:36
But now let's take any part of any business. A marketing department you can only afford one marketer or no marketers. A marketing department you can only afford one marketer or no marketers. Now you have the ability, let's say, to hire 50 marketing agents, with each one of those agents being an absolute guru just in one part of the marketing. Oh sorry, permanent expert In one part of the story is, I mean, I just think that the ability for small business, smc, to scale at a low cost, I think it's going to create a lot of new economic paradigms.

Andrew Welch : 22:10
I think it's going to bring a lot more benefits to even sort of us at home chilling out, right. So when we say agents, we refer to larger genetic models, large action models. It's going beyond just the ability to generate, generate, you know, text and images and videos and diffusion models and multimodal aspects. And it's actually the bit that I get really excited about as you probably tell, I've come, I've come alive and I'm very tired is when we start reaching out into the both digital and physical environments to do more. So let's take a really simple example. Let's remove the, the tech aspects and like, even like.

Andrew Welch : 22:42
So I want to book a trip to Slovenia to come to Dynamics Mines, I need a hotel and I need some flights. So if I speak to my agent and I say I'm going here, I need this, my agent from previous trips will know my likes, my wants. I like being near the beach, I like being near a shed ton of rum and I don't like to fly too far because I don't like flying. So I want a short, direct flight. Get me there as quick as I can and avoid so it's not my flight from yesterday. Yeah, yeah, so not your-. I would have just been tense the whole time and then ran off Making sure, we're very clear so what that's done is.

Andrew Welch : 23:14
I ping that across, say, and it goes Slovenia on these dates and it will go all of a sudden it'll go yep, fantastic, I know you like these things. Blah, blah, blah, blah. And it is pre-qualified to go off to bookingcom, to Skyscanner, and it will have my access to my bank to certain values, and it will say I found this. Are you okay to book? Absolutely Go off. And it does that all for you. And that's when we start really seeing how powerful these things are. So we can talk about the code aspects as well. Really important, but actually applying it for customer service, field service, everything else.

William Dorrington : 23:43
Dude, I think that I cannot wait right, because, can I tell you, life admin in my world is freaking hard, like it's really hard, and that going away or being made easier would be unbelievable.

Mark Smith: 23:58
But you take that to the next, like that one scenario that he just gave, that all of a sudden it notices that if you flew a certain class you'd go up a tier on your frequent flyer program. It would also note that halfway on your journey there there was a rum distillery that has the rum that you like, because it went and looked at that peripheral information and while you're away it would know the ages of your kids, family, that type of thing. And that gift was going to be perfect. They were going to have it all ready. It could be delivered to your hotel room, like all that kind of wraparound stuff that we think about or don't get time to think about. It's taking care of all those intricate details that make up a massive.

Mark Smith: 24:37
Yeah, and log the expenses.

Chris Huntingford : 24:41
One of my favorite things about receiving a gift from Mark Smith is how thoughtful it is. Yeah, every time, just put so much thought into it.

Mark Smith: 24:48
Okay, okay, at the end of the day, right, if I gave you a billion dollars, do you care how much thought I put into it?

Chris Huntingford : 24:56
Yeah, I would rather have a good gift that you put no thought into than a crappy gift that you put lots of that.

Mark Smith: 25:02
you put all the heart, you, you want me to hand draw your card. Put little kisses on there. Oh, that's important.

Chris Huntingford : 25:08
But oh no, the billion, no alexandra, mine and anna's two-year-old did hand draw cards for her, for her grandmother's, for mother's day, and I thought that was nice, it wasn't?

Andrew Welch : 25:18
very intelligent, but it was very cute it was very cute, it was very cute it was very cute it was a bit like you saw the monster.

Mark Smith: 25:24
Yeah, it was very cute, exactly.

Ana Welch : 25:25
Honestly, the whole concierge service stuff, but to the next level, it's fantastic. Andrew and I did use a lot of AI to book our honeymoon. It had to be like around the world trip almost. It had to be booked all in like two weeks and we had a very limited budget because we were in an interesting transitional period of our lives, you know. And we did all that, but using multiple agents, right through prompt engineering, through going Right now, I can already see how AI is evolving. Now I can already see how AI is evolving. For example, the other week I bought some glasses for my mom and as I was just getting ready to check out, I was like, hmm, go, pilot, can I get a better deal? I got a 50% off the same thing. Wow, wow, wow. It was incredible.

Andrew Welch : 26:26
What's scary about this is if you think about when bidirectional encoder model came out generative pre-trained transformer architecture. A couple of years on top of that, we then suddenly got diffusion models. We then got whisper models. On top of that, we then got Lumiere, we got the Saurus model. Then, a few days after that, we got the alpha geometry and science and the mathematics. Then after that we're starting to move on to now actually agents. This is a matter of days, right? How far from sort of dare I say, artificial general intelligence are we, and how do we define that based on you know what we're saying.

Mark Smith: 27:05
He doesn't want to hand me the microphone because he worries that I might go on a conspiracy.

Chris Huntingford : 27:07
It turns out Mark Smith is a conspiracy theorist.

Mark Smith: 27:09
No, I'm not, I'm not, I love it, I love it, I love it, I love it. You know the pyramids.

Chris Huntingford : 27:30
Anna? What was the conspiracy theory? He was peddling on us the other day. Do you even remember? It's almost like Mark said. I'm going to go freak my friends out at Dynamics Mines with every conspiracy theory in the book.

Andrew Welch : 27:44
I can't believe it. I love treating you, Mark.

Chris Huntingford : 27:48
I'm not even kidding, but, anna, what was it that he was going on about yesterday? It is irrelevant.

Mark Smith: 27:55
What I was going to say is the universe is made up of both magnetic and electrical states, and it's going to affect a lot about the lack of gravity we're going to have shortly. So, that aside, I've been watching the Joe Rogan show. So what I was adding to what Mr Dorrington was saying Adding to what Mr Dorrington was saying and the reason I say adding to because my mind is trying to catch up with the conversation is that I was reading the other day a book which was called Superintelligence, which was written in 2014. Amazing stuff Oxford professor, you know and in it he talked about in machine learning, it used to take like 30 days to train a model. That same model now is down to less than two minutes when you talk about the speed, a like for likeness, and you know.

Andrew Welch : 28:59
Not just reinforced learning, but a direct policy optimization.

Mark Smith: 29:03
It's getting a lot quicker and that's what enables it.

Andrew Welch : 29:06
They can do self-guided training as well. It's a whole different world. So we talk about these big developments, but we forget that it's the little developments that make the big developments happen.

Ana Welch : 29:13
Yeah yeah.

Andrew Welch : 29:14
Thanks for attending my TED Talk.

Mark Smith: 29:16
Yeah. So the other thing I just wanted to touch on is like I'm saying it a lot lately is this concept of friction in AI, and I feel like it's a bit clunky. The starting out of AI and I'm talking about how I feel like AI is creating more friction in the admin life. Admin at the moment and I give an example Email is something I've been harping on about is that I would like my email to become a lot more friction free.

Mark Smith: 29:45
Yeah, with the use of ai, I don't want it to just draft me a response, I want it to intelligently understand my email over time, with the goal of making it friction free. So not just drafting, but categorization, what's important, yes, all that type of stuff. And at the moment, the way ai is interacting, it's like um, so you start typing the sentence and then, once you've given me enough information, I will construct your email. I'm like, wow, that's more friction. And I'm like I feel we've got to bear with it for a little while. But I'm in everything you know, because people like, ah, that was too difficult, right, I see even peopleing and they prompt like it's a search engine query they're doing, rather than actually it's a backwards and forwards engagement.

Andrew Welch : 30:32
What's funny about this is I look at all of it and I think actually it could do a load of that, and I do think it's drip fed a bit to us for adoption. I might just say, oh no, that would freak them out, but what we will do is screenshot everything they do on their computer every second to do a whole recall. That was ridiculous.

Mark Smith: 30:48
I'm like how was that given to Satya as a great idea, especially that Mac? You know, Mac had the Wayback Machine back a while so not the Wayback Machine, it was called Mac has had the speech of some time.

Chris Huntingford : 30:59
Time Machine Time Machine, time Machine.

Mark Smith: 31:01
That of sub-time Time machine. Time machine, that's what I was, and I was like did no one think that there's people that have been on Mac will go oh, isn't this time machine the point I wanted to?

Andrew Welch : 31:08
make. There, though, is because you're kind of describing, at the same time, sort of personal large language models, and we all, you know there was a room that Apple was developing this as their race to the market, and it's going to be able to be really good at scamming, and this is the same issue with that recall, but I think there is a route to success there.

Mark Smith: 31:29
But also he dropped in the word edge in the conversation and as someone who's in the ecosystem, knows kind of what they're talking about. When he said that first up I'm thinking web browser. But he was not talking about web browser at all. He was talking about language models at the edge on your hardware, on your device. I think they could have been a bit more explaining about or clarifying what the edge meant.

Chris Huntingford : 31:55
So I have to look this up. Speaking of Microsoft products, here we'll say something intelligent.

Andrew Welch : 32:02
While I find this, what I'm seeing on your phone there is rather worrying.

Chris Huntingford : 32:15
So I was preparing the course material for the session that Anna and I taught yesterday, and one of the things that we wanted to do is we wanted to share some technologies that people should be learning more about. So I asked Copilot to write me a one-sentence description for each of the following Microsoft technologies, and I've got OneLake and Purview, azure, ai, search. I've got a list here, and Copilot did an admirable job of summarizing these technologies until it came to number eight, and that was I had asked for co-pilot studio, and co-pilot replied back co-pilot studio is a fictional term and does not correspond to any known microsoft technology co-pilot studio is such a confusing thing to search on as well.

Ana Welch : 33:00
Oh my god, whenever, because it's a, it's a thing right, and people use it a lot and it's very, very useful. And use copilot studio in conjunction with, I know, customer service like it's a. It's very, very important. You cannot really find any relevant information about it because it just gets mixed with everything else that.

Andrew Welch : 33:23
That's the answer to why it struggles, because all it's doing is using that to predict the next word when they have these awful naming conventions, which is let's just call everything. You don't know what you're talking about. Let's call everything Copilot and let's see how it works.

William Dorrington : 33:36
This is brought to you by the same people that made something Teams, teams, teams.

Andrew Welch : 33:40
Teams.

Chris Huntingford : 33:41
We needed Teams, teams, teams.

Andrew Welch : 33:42
Teams.

William Dorrington : 33:42
Teams, Teams yep and.

Mark Smith: 33:48
Yeah.

Chris Huntingford : 33:49
Well, so this my my debate with Copilot continued when it got to Azure AI Studio, and it said the same thing Azure AI Studio is not a recognized Microsoft technology. It may refer to a fictional or future product, but finally it came to fabric and it said. Copilot said fabric is a term that can refer to various Microsoft services, but it is not specific to a single technology, which, well, probably that is true. But come on, copilot, now you just you know what I want, you know what I want you to tell me.

Ana Welch : 34:22
It's just lazy, that's lazy, yeah, exactly. I think that exactly what we were saying before the fact that it's a conversation with Copilot. It's a series of prompts, and then you asked it a question and you were unhappy with the response, and then you broadcasted it to the entire world. That's just not okay. No, no.

Mark Smith: 34:42
But that's the thing. Right is that if it's a conversation and you're having a conversation with you know a friend and they said something that you knew was out of line you don't go. Oh it I'm out, exactly goodbye. Can you believe that? I can't believe you said that, like you wouldn't take that indignant type, you know attitude, you'd go buddy, you're you know you would say you're full of it, mate.

Mark Smith: 35:06
That ain't the truth, right? Or what are you being smoking? Or something like that? Right, you would call them on it, you. You would develop the conversation further.

Andrew Welch : 35:12
Is that what you said to the chat?

Chris Huntingford : 35:14
Yeah, yeah, yeah, yeah, yeah, yeah. Yeah, I typed my own one-sentence description. Nice, you did that was myself.

William Dorrington : 35:23
Considering you are an author. Thank you.

Chris Huntingford : 35:25
That's nice of you to say. It's such a pleasure.

Ana Welch : 35:27
But I think it was actually very interesting because we assembled all of that information about technology for the scenarios that we had in the course, but people came up with so much more technology, also very suitable. So that was very interesting how ecosystem enablement actually means so much more than the tools that we would expect.

Mark Smith: 35:53
yeah, yeah I'm surprised with that new functionality. Back to the microsoft. What is it called? Uh, recall, yeah. So I put a tweet out and said, okay, I'm going to deal with the elephant in the room here. Um, what if there's stuff that I don't want to add into my digital record? You know, content I might look at from time to time.

Mark Smith: 36:13
I know you're all acting like you don't look at this stuff. But let's say, you know, is there a pause button like on it? Is there any kind of well you know what? This window of next three seconds? I don't want to record it Three seconds. Yeah yeah, 30 seconds Okay cool.

Chris Huntingford : 36:34
Incognito browsing is totally secure.

Andrew Welch : 36:37
But there's still screenshots.

Mark Smith: 36:38
That's recording my screen.

Andrew Welch : 36:39
We can end on that.

William Dorrington : 36:44
I'm literally not commenting.

Chris Huntingford : 36:49
So back to Build last week. What were some of our favorite build announcements? There were some good ones, but what were we?

Mark Smith: 36:58
into.

Chris Huntingford : 36:59
Copilot PC. Tell us more.

Mark Smith: 37:02
It costs about $4,500. That's when I've specced up the machine. So if I get it right, right and I don't know that I do if I get the processor, the RAM, that type of thing, my understanding is it's running small language models right on device. So it's going to do a lot of the AI heavy lifting for speed and a lot of what you do direct on hardware as well as interface back to the cloud. That's my understanding of it.

William Dorrington : 37:27
Okay, so yours is different. No, no, no, I didn't understand it at all in the beginning. You've cleared it up for me, yeah. I was just wondering why I use a MacBook.

Mark Smith: 37:37
Because of speed. The biggest lag on using an LLM is internet latency, right, the speed that it can get there If it's going to become full. Conversational in other words, no lag, right, shucks, okay, I get it Right. Conversational in other words, no lag, right, shucks, okay, I get it right. So the idea is, if I because we're all ultimately going to go to voice as a interface that we're going to use all the time right, so you're not going to stop, right, yeah. So you imagine, if you said, uh, you know you asked a question or having a conversation, the other person went quiet for a period of time. Well, I thought about what the fuck are they? Sorry, what are they going to say, right? Um, that would be painful. So what? The whole idea now is that, yeah, is that they're going to process a bunch on device and then hand to server, etc.

William Dorrington : 38:20
That makes total sense that's, that's my understanding.

Andrew Welch : 38:22
Final word, final word from the most impressive thing about that, though, is it used to take such extreme power to run these models and such large volumes, and now you know this goes back to what I was saying about the waves of technology crashing again closer and closer and closer together. We've gone from we need all the power in the world to hey, why don't you just put it on this computer and just run your own little personal lmao?

Chris Huntingford : 38:42
we're back to on-prem, you know it's great, reminded us all why we went with the name Cloud Lighthouse because of the waves.

Mark Smith: 38:51
Yeah, the waves, yeah, and that device 850 grams. It's 850 grams. I'm American. That needs to be edited out what he said there. Thank you, hey, thanks for listening. I'm your host business application MVP, mark Smith, otherwise known as the NZ365 guy. If there's a guest you'd like to see on the show, please message me on LinkedIn. If you want to be a supporter of the show, please check out buymeacoffeecom. Forward slash NZ365 guy. Stay safe out there and shoot for the stars.