speaker-2: Hello folks, welcome to another episode of Human in the Loop. Very, very excited for this month, and I'm not just excited because Chris is not here, but because it does mean I have a new co host. So Ana, can you say hello?
speaker-1: Hi, hi everyone. my name's Ana Welch. I am Chief Technology Officer for a company called Cloud Lighthouse and the technical director for the Center for Trustworthy AI. I am delighted to be here because my mind is just fueled by other people and their ideas and what they think and what they do every day. So yeah, thanks so much for having me.
speaker-2: Of course, and you're gonna love this one then, Ana, because I was so excited to finally get Sean on the podcast. I've been trying for months. So threatening whatever incentives I try to get him to spend time with us. So Sean, welcome and thank you so much for being gracious with your time and and joining us today. For those out there who I don't know, somehow missed you and don't know what you do, can you share with us about you and the world that you occupy in the AI sphere?
speaker-0: Yeah. So my role is I help makers become solutions architects and I provide advanced training and AI Power Platform in Azure. And when I say advanced, I mean like real enterprise architecture, difficult to solve scenarios, the Microsoft Learn can't help you anymore scenarios, the problems that AI can't solve scenarios. That is what I train in. And I'd like to say I, but it's not really me anymore. There's like five of us now. And we solve the problems AI can't and we train you to do that. So you make the most money you can.
speaker-2: I love that. there's a clear value prop there that I I subscribe to. Also, just a personal thank you because I really enjoy your content. I have a neurodiverse brain that is definitely squirrel in terms of attention spam. And you do such a great job at not only producing super valuable content, but you do it in a way that my brain can actually stay focused and understands and you do it in a way that is absolutely wonderful. So a personal thank you for the content that you create and your team creates because I know that I benefit from it.
speaker-0: Awesome. Thank you for saying that. We assume no one has an attention span, and that has worked really well for us.
speaker-2: a good lesson in life. so okay, let's I I always start with kind of a softball question and I'm actually gonna have it to both of you and Ana, you're gonna go first. responsible AI and AI in general, you know, topics that not really new but new in other ways. So I always ask our guests, how do you define responsible AI? And Sean, while you think about that, Ana I'm gonna go to you. What is responsible AI for you?
speaker-1: For me responsible AI or trustworthy AI rather is a the whole base of artificial intelligence. And it for me it spans from is our AI, you know, ethical, reliable, all of the rules that we know, all the way to Do you know what it does and how it does? So are we able to interpret what our AI agents, workloads, automations are doing right now step by step? Are we comprehending what's going on step by step? What is the AI's logic when it produces something? That's also part of trustworthy AI. are we controlling it? all of the time. What level of human in the loop do we do we have? Because like Sean was saying just a second ago, nobody's got an attention span anymore. And so how do you all of the sudden have this requirement that somebody who used to maybe review and approve the work of person or like 10 people per week now approves the work of 10 people and their agents every week. Like how will they just still keep the the focus on it? All of that is responsible AI for me, together with how much money are we spending on AI? Is this still a good return of our of our investment? all of that built into our Models are workloads, the way people work and what their job means, all of that is responsible AI. So it's absolutely massive and the core of my job, I'd say.
speaker-2: I love it. And I it's so interesting because a few years ago, and when I say a few years ago, I mean four or five years ago, when we were talking of ethics and I'm using air codes of ethics, it was like it was like some niche thing that on the side of the table that most people are like, Yeah, yeah, we have an ethics team and now I look at the company I work for and that central stage of like how do you think of ethics, how do you think of unifications, how do you think of everything from safety to engineering to your point, Ana, how do you ensure that what you're doing one is traceable, is manageable, who manages it and so forth. It's a whole big conversation. It's a whole big role. but I'm glad that you guys have a focus on it because without it we we see the examples of when AI goes rogue. Sean What is it for you? How how would you define responsible AI? I either for yourself or for your clients or in general? It's a it's an open question.
speaker-0: Just try not to prevent myself from ever losing the opportunity to be invited on this podcast again with my next answer. I don't find it that important.
speaker-2: Fair, tell us why.
speaker-0: I'm gonna let that sit in the air. I said that. Yeah, yeah. If you understand the actual security of what AI is doing behind the scenes, like what Ana was alluding to with knowing what it's doing, if you understand, let's take power platform. The admin center and DLP. And you realize that data loss prevention does very limited things in terms of security. And there's all kinds of ways to bypass it. If you understand the basics of how network works and subnets work, and when do you need things like VNet, if you understand when purview comes in or not, I don't see the point of diving too deep into AI ethics because AI, if we're staying in the Microsoft space and the graph API and Copilot, it's running on all the security infrastructure you set up. So if you set that up right, your AI can go like it can invent ideas, but it's not going to bypass your own security. So then we come to the second. So I feel like Microsoft has done a tremendous job of locking it down. It's why they trust it. It's why companies don't go to Claude. It's why they don't go to Chat GPT or Grok is because Microsoft has that security nailed. They've always had it nailed and they just plugged AI into it. the second part is the the ethics. I don't like I do questionable things. I'm not perfect. I like don't always tip at restaurants. So I'm not gonna tell people how to use AI ethically. Here's what I will say, and here's the only thing I say. You are responsible for what AI does. It is your employee. No matter what AI did, you can say AI did it. You will take the full responsibility for it. How much you wanna double check its work? That's your reputation and your life and your company. But in terms of security, That's locked down. So the ethics side of it, I'm not a moral ambassador.
speaker-2: Fair. Can I poke some holes though at your assumptions and give some enchant for those who might not watching? He's like fading into background like Homer Simpson in a in a hedge. I l hey, I love it when people have diverse opinions. That's the whole point of this podcast. Let's poke holes at concepts and idea and we don't all have to agree, but let's have a conversation about the points and disagree. It's interesting how you think about it. And I will challenge you a little bit. You talked about security as being, hey, you know, if you start with all the right things in place, be it purview, be it, you know, mitigations, whatever it is, you don't have to care about the ethics. And here's the thing, I partially agree. But, and there is a but for me personally, in order to get to that stage, you have to start with a set of principles that when you're developing this technology, you say we care about security. Because if you don't care about security, you're never gonna get to a stage where you actually do things for security. And I say that as somebody who has spent a lot of time in our responsible AI framework, where security is a pillar of that, privacy is another pillar of that, fairness is another pillar of that. There's five. but unless you know that you want to care about that, you're not gonna develop something towards it and measure if it works or not. So it's kind of a yes, I agree, but in order to even have security, you have to have a point in your company's whatever history development, whatever you wanna call it, where you say, I care about doing security. Because if you don't, you're not gonna. So that's the only point that I would challenge about what you shared is you have to put a stake in the ground whether you're gonna do a thing or not. And you're not gonna do things that you don't care about or are not a business priority.
speaker-1: I also have to say that there were a lot of beefs there, right? If you've done the purview setup and then you've done the DLP and then you know the subnets and all the networking, da, then it doesn't even matter. Well hell I agree, it doesn't. If you're not all of that work and you have dashboards and you know exactly what's going on and you've got pings and signals of your agents. Yeah, fair. But That's a lot of work there, right? A lot of money. Like you when when you were starting with, you know, is this important or not? It sounds like there's some investment there and therefore for some people who are paying it, they may think it's important.
speaker-0: Yeah. And the people I interact with, I guess the I've been lucky where every company I've ever worked with, they start with security. And if they don't, we convince them to go security first. And I I I think we're all in agreement on security is number one and everything else. It doesn't matter if people can steal your data. Where I think I take a different turn is I'm not gonna tell you about ethics and what you do responsibly with AI. I'll make sure
speaker-1: Yes, all I'm saying is that responsible AI is security. I don't think that you can separate it anymore. Like part of trustworthy and s and and responsible AI definitely is cybersecurity also within whilst I completely agree that Microsoft has locked down the enterprise model of how people should use agentic work in a big company. Totally agree with that. I also agree that this is why people do go to Microsoft and they create their big workloads and their big strategy on Microsoft. There's also a massive, massive motion of open source within Microsoft. So what do you do with like the fund reagents and everything that's been created in Landgraft and like connected back in and all of the tools and the S and Psal? So yeah. That's why I feel like trustworthy and responsible AI can be a massive hole in people's plans in creating stuff.
speaker-2: You brought something super interesting, which is like, you know, open source, barcoding, whole movement, there's a lot more empowerment in terms of people being able to create. And Sean, actually I want your take on this because it it's your bread and butter, right? You help people in terms of AI, you help them really think about the future. What what part while playing this? What's your take in terms of, hey, there's and forget about Microsoft, but there are a million different tools out there, be it Microsoft or not. be a Power Platform, what's your advice to folks, especially as they're thinking of their future when it comes to AI?
speaker-0: We follow a process that it's called ACS when it comes to AI. The first one being, and a lot of what you said earlier around trustworthy AI is doesn't it can be that ethical question, but it's also just accuracy. I care about that more than as you can tell, ethics. I care about when somebody types in something, asks AI to do something, it does the exact thing they asked. Yeah. Whether It is Copilot studio or whether you need to move into Foundry, whether you need to move into a different tool. That's number one is accuracy. If you don't have accuracy, it doesn't really matter. the next point is cost effectiveness. Like how much, like, am I spending the most money on this? Am I spending more than I need to? Do I really need a co-
speaker-1: Password AI for me.
speaker-0: Yeah, yeah. for sure. Next point. so yeah. So then cost effectiveness is the next one. And then which like like Azure just is always cheaper than almost everything in Power Platform. However, you have to go with the aspect of now you have to kind of configure it yourself and you take more risk, which goes into the final step of security. And that's now some of people would argue like, don't you need to put security first? And I'm like, Yeah. I mean, I would say that as an architect or someone who's recorded on a podcast, but when I talk to companies, it goes accuracy, cost effectiveness, then security. I didn't come on this podcast to make stuff up. So that's the order we go in. I think the trustworthy AI framework covers all this and way more. I just know when I've explained it to companies in a way that they listen, I just said ACS, accuracy, cost effect, effectiveness, security. And then that that's made it very digestible for them.
speaker-2: I love that and I love the fact that and we've talked about this in on the podcast and in the in the past of like how do you translate these concepts in ways that matter to people, matter to your stakeholders, matter to their bottom line, and I'm gonna steal the ACS because I think it's such a smart, business oriented, everyday framework that people will just naturally understand because it is what they do at the end of the day. What are some patterns or theme that you are say we are seeing around ACS or around AI in general with businesses? What are things that you're like, huh, this is interesting or hmm, we should be ready for this thing or the other?
speaker-0: Yeah. so in this one, the main one I've seen is it's super easy for AI to summarize and get information. It's awesome at that. It's in terms of creating stuff, like especially like creating content, generating emails, all of that. It's it's it's not good at it. It's just it does its best, but that's the one where we train and educate companies that like you need to be realistic with your expectations of like summarizing info. Yeah. Creating stuff, good luck. And that's where if that's important to you, you're gonna have to spend a lot more money. So I'll give a tangible example. an easy one that's argued all the time, Copilot versus cowork. Cowork's amazing. Cowork costs a ton of money. If you need certain things, like I could demo it right now. I have a meeting with Ana and Ioana coming up next week. They're in three different time zones, coordinate the best time zone and create a repeating meeting every week with them based on when they're free each week. So each week it has to figure it out. Like that type of thing. You can try it with Copilot and it will try. And it's gonna, I've done this multiple times. It's gonna send you 11 invites at the same time on the same day, and you're gonna be like, what did you do? however, if you spend the extra money to go to Co-work, it will pull that off. However, what people don't realize is if you don't ask it that advanced of a question and you just say, like, schedule an invite based on this. Also, go ahead and hop into SharePoint, pull the data from here. Move it into this other folder. You can do that with Copilot. You don't have to spend the money on Co-work. So that's the one where like cost effectiveness comes up the most is like Copilot versus Co-work. and security, it's just not even a conversation because it's all you don't have to do anything, it's already listening to your existing permissions. Accuracy is probably the most fun one, and where I've seen companies usually have to pay consultants to really get is if you want to pay. Not even pay. Let's say you want this is something I think could benefit everyone who's listening to this call right now. What if you had an agent that properly drafted emails for you? Like, like nailed it first try. Not a bunch of like long texts and it's guessing or making up information, but like nailed it. Well, the architecture behind that's very hard. You can try to plug in cowork or AI builder or copilot studio workflows. It'll never be that good. It'll never be that good because every time it's guessing, it doesn't learn from itself. So to build an agent that learns from itself, learns from every email you sent, you get into this like concept called loop engineering. So that handles part A, like accuracy. So I think I went a little off track there, Ioana, but I hope I answer your question of like trying to give you a tangible example of what ACS looks like.
speaker-2: Love those. Ana, I'm gonna go to you in a second. I'll throw you a curve Sean. There's one coming, Scout. So you're gonna soon need to have the conversation of Copilot, Co-work and Scout. And I say that as somebody who's been living in Scout for the last four months, and there's a lot of power in tools that are coming up Ana, what are you seeing in your world and what are themes that you're seeing with your clients?
speaker-1: so first of all I wanted to go to what Sean was saying with the accuracy and I I I love that. I agree with you. That's exactly what we need. Like I want an agent not only to book my next meeting, but I need it to understand from the context of my emails that the meeting's not needed anymore. And to kind of ping me, like according to what you said, it looks like the plan has changed, the leadership has changed, the workshop has changed, you know, the deliverables have changed. Do you still want me to book this? You know? And that's where the power kind of brings, the the power comes from. you know, these agents that we do work on creating very custom to deliver high accuracy and obviously high ROI. and I'll have a question on that in a second, but I'll I'll come back to your to your question, Ioana. What I'm seeing is is so so I come from a pro code world, right? So my customers are like I don't know fast. They move fast. Which means they're not always working under the blanket of like Copilot or even Co-work or even Scout. They go a bit rogue sometimes. Like you're saying that's why people don't go to Claude? I see a lot of people going to Claude, I'm I'm here to tell you. And wherever we lock And I I'm also seeing a lot of organizations lock down Copilot in any of its forms, right? From M365 Copilot to absolutely everything, right? And that's because they don't know how to set up the basis that you were talking at the beginning, you know, with data cataloging and and I don't Defender and and all of the things. And so they are They are worried and they're not unlocking it, which causes their staff to do what? Paste all of this information into a 20 bucks Claude to get what they need. And so and that and that's what we're and that's what I'm seeing. So obviously what we are trying to do is to enable people to deliver that accuracy in whatever medium they are. And my question for you, Sean, would be, have you seen that? And how do you see these jobs going in the future, starting from the person who used to book all of those meetings and had to ration through the logistics of where people are and you know, how would an event looks like and does that meeting event still needs to be there or not, to us creating an agent to do that job for them all the way to these organizations who would where the staff would love to use Copilot or Co-work and they cannot.
speaker-0: Yeah.
speaker-1: What should they do or how do you see these jobs?
speaker-0: Can I ask a couple questions to clarify the question? So one question I heard is like you've actually seen a huge shift in people going to Claude, is one that that
speaker-1: Hundred percent. Yeah.
speaker-0: and I've seen a lot of that as well, counter to what I said earlier, but I'll explain why I didn't mention that one. The other one was a lot of companies also lock down Copilot, they lock down Co-work. and then the third one is like, how do we deal with kind of like shadow IT? When all of this is locked down, how they get around it. And then I think the third question, if I understood correctly, is what is the future of work with all of this?
speaker-1: That's right.
speaker-2: Nice and easy set of questions.
speaker-1: They're all kind of connected if you think about it, right? Yeah.
speaker-0: That's how they worked in my head. So we'll just go in order. First one, Claude's cheaper and more powerful. Like all of Claude's models, I will go on record as say as of today, way better than GPT or anything else. So Claude wins in terms of like price and great models. Every startup I know uses them.
speaker-1: I've got somebody in our company who prefers GPT and will die on his sword for the open AI models. Every There's
speaker-2: There's always an odd liar.
speaker-1: always somebody who manages to like work these models into providing their best accuracy. Let's just see.
speaker-0: Is he a pro coder? Is he writing code with it? Okay. Yes. Yeah, I don't care about writing code, but good for him. So on this point, yeah, Claude Claude's awesome, Claude's cheaper. Last I heard the way Claude stores their data, a lot of European countries aren't comfortable with it. And then the second one is going through Copilot Foundry, you still get access to all of those models in a more governable way. Like, I don't know the governance around Claude because I know it's isn't as good as what Microsoft is doing for Microsoft products. So, like, Cloud is awesome, Cloud is cheaper. I'm going to the first thing you said about trustworthy secure AI. And it's like, now I'm gonna lean towards the companies that want to govern it. If you want to go a little more rogue and pay for Claude and do it that way, like okay, you it'll be cheaper and it will be In a lot of cases, quite accurate. just now governance is your problem. So I I don't want that problem. the other one was point number two, or shadow IT. I don't know why I understand why companies lock down Co-work because the spending can go through the roof. I understand why they would limit Copilot if they don't understand it, if they haven't gone through your AI trustworthy framework. If companies are locking down. Copilot and they're not offering an alternative like Claude which which to me doesn't make sense because now you're like stopping copilot a governable tool and you're introducing a different one to be cheaper and potentially more accurate. Like that is just confusing to me and is a different conversation.
speaker-1: To me they're not. Like they're they're just not offering anything. Right. Like I I have seen organizations just lock down, you know, Co-pilot with the thought of, you know, I'm gonna let that be tomorrow's problem. We've got work to do. Let us fill our Excel spreadsheets, you know, like right now. That's that's what we're doing. And I've seen people then go. away and doing their own thing because everybody wants to get ahead and everybody wants to just keep their job, right? Yeah. It's just the way it is.
speaker-2: Yeah, the reality of that becomes yes, you lock you lock down fearing whatever, but you're just making it so this is where it becomes a not using air quotes, bring your own AI, it's like bring your own beer. People are going to default to AI tools that work to for them. so suddenly you're taking things outside of your secure enterprise environment and putting it in the open with tools that potentially do train. their models back on what is in the consumer or public area because the advantage of enterprise, you can lock down on how that data is used and how that data gets organized used back into the models and retraining. So it's an interesting sort of it's not even a catch twenty two, I don't know what that is, but it's kind of like the you try to you try to do the right thing and I'm using again air codes, but you actually create a bigger problem for yourself because your employees will default what works for them.
speaker-0: Okay, so that that's awesome. Thanks for clarifying it, Ana. So that's the point of some companies are locking down completely. And then I think this ties really well to the future of work for that third group. I'll go on record and say this. If your company is locking down AI completely, I would quit.
speaker-1: Go away. Right. I'm with you.
speaker-0: I think if you're not scared about your job, me, everyone on this call, everyone listening, due to AI, I think you have no idea how crazy powerful it
speaker-1: Delusional.
speaker-0: is. Like GPT Codex, let's reference your boy at your company if it's a boy, I don't know what their gender is. the person at your company who loves GPT, Codex, came out in February, was the first model that wrote itself. Yeah. Can we understand the complexity of that? It it was so Like the most advanced piece of software wrote the most advanced, one of the most advanced pieces of software. Like that's enough that I believe if an AI can get enough data about your business problem, it can outperform any idea you have, any architecture you have. The one advantage we have is that we're humans and we have this like kind of human understanding of other people and business, and we can do that. But overall, it is a time that's never existed like this. I hope it all works out for us. I hope we all get jobs in this AI age. I really do. I also don't understand how, with the way the world works today, how there will be enough work if someone like Ana can do something eight developers used to do when it's just Ana with AI. Where do those other eight developers go? So I believe my responsibility is to get people as trained and knowledgeable as possible. So if this is the Titanic. I can create as many lifeboats as there are.
speaker-1: I love that. I'm I'm with you and I have one more question and I'll let you be. Ioana she'll she's never gonna have me on her podcast ever
speaker-0: Exactly.
speaker-1: again. We're just like off. We're just like yeah. Okay. So I love that. I believe what what you believe with a caveat. I think that's the big governments, the and the big vendors as well. problem. I think that's the problem that they've created and they should really think very seriously about solving. Governments and big organizations should be solving this problem. But I'll give you like a or real life scenario, right? Let's just say you are a company and I don't know, a year ago you decided that you are going to create a new power platform practice. And you've hired, let's just say, seven individuals, and you love them and they're nice people to work with and they've learned power platform appropriately. And they create all sorts of things for you, such as power pages and you've got a bunch of model driven apps and some canvas apps as well and everything's fine. But Hashtag AI and therefore your website should be changing quickly quicker and your processes should be changing quicker and you're a bit afraid that your competition is gonna go ahead. What do you do with that team of our platform developers, consultants, whatever you want to call them?
speaker-0: how many people do I have on in this hypothetical scenario?
speaker-1: I should say seven. Five, seven, ten.
speaker-0: Yeah. I would say in this case Okay, I'm trying not to say enroll them in my training program so they can also be consultants as well as architects but auto, you kinda like softballed that question to me. so I'll kinda
speaker-1: Thank you.
speaker-2: say that enroll them in Sean's training program because that's exactly what I had in mind as Ana was asking that question. I was like I know this is not for me but the answer is right there. It's upskilling
speaker-1: Let's talk about
speaker-0: It's yeah.
speaker-2: help people learn the
speaker-1: the record, Sean did not tell us to ask him this question at all. Right. But no, seriously. These are the things that are happening every day. How will these people understand that before okay, fine, before I used to do a power automated flow or a bunch of apps and I had like user stories and fields and processes and so on? Now I need to build I'm I don't know your training program. in enough detail, but this is what I would do. Now I need to build connectors, MCPs, things that are able to be used by AI and, you know, accelerate things.
speaker-2: Can you tell us about your training program, please?
speaker-0: Yeah. Speaker of the word accelerate, it's literally in our title. So our
speaker-1: Funny.
speaker-0: Is the AI Architects Consultants Accelerator. Let's just jump to the tech right away, honestly. That's what you were asking about. The first thing we do is we get everybody started in first of all, we explain what I said earlier, which is all these tools are going expire. Anything I teach you is not gonna be really that relevant in a year. It's like you need to set up the coding and software foundation for yourself that you have AI do as much for you as possible. So everything we teach is from VS Coach.
speaker-1: I thought you don't care about you you you absolutely don't care about anything code. I think that's what I heard earlier, no?
speaker-0: I think we can replay the recording because I don't think I ever said that sentence.
speaker-1: Okay.
speaker-0: I said I don't care about being the best pro coder. Even though I'm pretty sure I said it as I don't care about anything code, is were the exact words. My actual meaning is I'm focused less on pro code. So here's my point. When I say less on pro code, I mean getting into the very specific details of like do I need to get into I don't know, like Multitask processes. Here's the point. We do teach Pro Code after everything I just said. So in this case, we show them how to use VS Code and pretty much build all of your power platform solutions from there. The main thing we do is show you how to take all the most complex software problems from data migrations to security to licensing to deployments, copilot, AI, fast apps, performance, all of this. We take all this and we show you how to build it as quickly as possible and as flexibly as possible for the enterprise to large scale. Now, the big difference is saving your time is one thing. Like, sure, you can spend the rest of that time now going on vacation, spending it with your family, whatever you want. We feel like now that you have all that time saved, you need to understand the business side of power platform. You need to understand how to work with stakeholders, you need to understand. Pretty much on all the skills you've built over years of your career of like how do you work with people that are scared of copilot, that want to shut down governance, that don't even understand how to begin and hack that don't know what to do with tech because they're not a tech company. Like, how do you help them? And this is what we train them to do. We believe the future of work is you hop into something like VS Code, it codes up your solution for you. Your whole job is understanding the client. And we teach you under how to understand the client.
speaker-1: B N F D
speaker-2: That is brilliant. I love that. And I love the I love the pairing, right? It's like it's the technical skills, it's the business acumen, it's the AI acumen. But here's the real world stuff. In terms of you can have all of that unless you understand your client and can work with your clients. It's you can have the best coded app in the world. It's not gonna make a difference. I love it. So in the interest of time, I always ask our guest a closing question and today I'm actually gonna ask you both. If you were to give our audience homework, what would that audience what what would that homework be? Ana, you go first.
speaker-1: Find your passion. What moves you when you go into a room? How does that how is what you do change the room? Find your passion truly? I know it sounds privileged, I know because you're doing a job and you're busy and you just want to keep it, but I'm serious. Find your passion because if you find your passion and what you like doing, what sort of problems would you like to learn more about? That's where you're gonna be able to do exactly what Sean was saying, you know, understanding your your customer, being able to quickly enhance their you know, their ideas into in into solutions that are not as our friend Dona says. faster horses, you'll be able to to to reinvent a problem statement and create a car, right? Find your passion. That would be my number one advice. Write it down every single day. What did I like about my day today? And do that.
speaker-2: I love it. Sean, what homework would you give?
speaker-0: Besides follow me and my YouTube channel, besides that home.
speaker-2: Obviously. That is my homework for the audience.
speaker-0: I mean, I really liked Ana's advice, so I'll go off of that. get some rich friends. Make rich friends. All your, you should be the poorest in your friend group. If you're not, I don't understand how you expect to like move up if all of your friends have the same experience and the same amount of salary than you. Now, obviously, you may have some friends who are very poor and have a lot of free time. Well, they're doing something right. So you can learn from that. The point is. besides find your passion, what's helped me find it is hanging out with very, very successful people.
speaker-2: It's funny you say that. I'm going on a tangent, but I was reading this study that influences work. And when you're working with high performing people, it rubs off you. And when you're working with people who are less performing, it also influences you as much as you would want it not to. So selecting your tribe, I'm just gonna call it a tribe. Selecting your tribe and the people that you surround yourself with, super, super important. That's something that I fundamentally believe and have believed for a lot of my adult life is be peaky with the people around you. They're gonna influence you for good or for bad. so it's a choice. Folks, thank you so much. Sean, you're definitely gonna come back and also will you. this podcast is all about, hey, let's poke let's poke holes, let's have different ideas. That's what the beauty of it is. Otherwise we will make an AI generated draft and read it to the audience. That's not what it is about. So folks, thank you. thank you for our audience for tuning in and we'll see you next time for another episode of Human in the Loop
speaker-1: Thank you.
speaker-1: Hi, hi everyone. my name's Ana Welch. I am Chief Technology Officer for a company called Cloud Lighthouse and the technical director for the Center for Trustworthy AI. I am delighted to be here because my mind is just fueled by other people and their ideas and what they think and what they do every day. So yeah, thanks so much for having me.
speaker-2: Of course, and you're gonna love this one then, Ana, because I was so excited to finally get Sean on the podcast. I've been trying for months. So threatening whatever incentives I try to get him to spend time with us. So Sean, welcome and thank you so much for being gracious with your time and and joining us today. For those out there who I don't know, somehow missed you and don't know what you do, can you share with us about you and the world that you occupy in the AI sphere?
speaker-0: Yeah. So my role is I help makers become solutions architects and I provide advanced training and AI Power Platform in Azure. And when I say advanced, I mean like real enterprise architecture, difficult to solve scenarios, the Microsoft Learn can't help you anymore scenarios, the problems that AI can't solve scenarios. That is what I train in. And I'd like to say I, but it's not really me anymore. There's like five of us now. And we solve the problems AI can't and we train you to do that. So you make the most money you can.
speaker-2: I love that. there's a clear value prop there that I I subscribe to. Also, just a personal thank you because I really enjoy your content. I have a neurodiverse brain that is definitely squirrel in terms of attention spam. And you do such a great job at not only producing super valuable content, but you do it in a way that my brain can actually stay focused and understands and you do it in a way that is absolutely wonderful. So a personal thank you for the content that you create and your team creates because I know that I benefit from it.
speaker-0: Awesome. Thank you for saying that. We assume no one has an attention span, and that has worked really well for us.
speaker-2: a good lesson in life. so okay, let's I I always start with kind of a softball question and I'm actually gonna have it to both of you and Ana, you're gonna go first. responsible AI and AI in general, you know, topics that not really new but new in other ways. So I always ask our guests, how do you define responsible AI? And Sean, while you think about that, Ana I'm gonna go to you. What is responsible AI for you?
speaker-1: For me responsible AI or trustworthy AI rather is a the whole base of artificial intelligence. And it for me it spans from is our AI, you know, ethical, reliable, all of the rules that we know, all the way to Do you know what it does and how it does? So are we able to interpret what our AI agents, workloads, automations are doing right now step by step? Are we comprehending what's going on step by step? What is the AI's logic when it produces something? That's also part of trustworthy AI. are we controlling it? all of the time. What level of human in the loop do we do we have? Because like Sean was saying just a second ago, nobody's got an attention span anymore. And so how do you all of the sudden have this requirement that somebody who used to maybe review and approve the work of person or like 10 people per week now approves the work of 10 people and their agents every week. Like how will they just still keep the the focus on it? All of that is responsible AI for me, together with how much money are we spending on AI? Is this still a good return of our of our investment? all of that built into our Models are workloads, the way people work and what their job means, all of that is responsible AI. So it's absolutely massive and the core of my job, I'd say.
speaker-2: I love it. And I it's so interesting because a few years ago, and when I say a few years ago, I mean four or five years ago, when we were talking of ethics and I'm using air codes of ethics, it was like it was like some niche thing that on the side of the table that most people are like, Yeah, yeah, we have an ethics team and now I look at the company I work for and that central stage of like how do you think of ethics, how do you think of unifications, how do you think of everything from safety to engineering to your point, Ana, how do you ensure that what you're doing one is traceable, is manageable, who manages it and so forth. It's a whole big conversation. It's a whole big role. but I'm glad that you guys have a focus on it because without it we we see the examples of when AI goes rogue. Sean What is it for you? How how would you define responsible AI? I either for yourself or for your clients or in general? It's a it's an open question.
speaker-0: Just try not to prevent myself from ever losing the opportunity to be invited on this podcast again with my next answer. I don't find it that important.
speaker-2: Fair, tell us why.
speaker-0: I'm gonna let that sit in the air. I said that. Yeah, yeah. If you understand the actual security of what AI is doing behind the scenes, like what Ana was alluding to with knowing what it's doing, if you understand, let's take power platform. The admin center and DLP. And you realize that data loss prevention does very limited things in terms of security. And there's all kinds of ways to bypass it. If you understand the basics of how network works and subnets work, and when do you need things like VNet, if you understand when purview comes in or not, I don't see the point of diving too deep into AI ethics because AI, if we're staying in the Microsoft space and the graph API and Copilot, it's running on all the security infrastructure you set up. So if you set that up right, your AI can go like it can invent ideas, but it's not going to bypass your own security. So then we come to the second. So I feel like Microsoft has done a tremendous job of locking it down. It's why they trust it. It's why companies don't go to Claude. It's why they don't go to Chat GPT or Grok is because Microsoft has that security nailed. They've always had it nailed and they just plugged AI into it. the second part is the the ethics. I don't like I do questionable things. I'm not perfect. I like don't always tip at restaurants. So I'm not gonna tell people how to use AI ethically. Here's what I will say, and here's the only thing I say. You are responsible for what AI does. It is your employee. No matter what AI did, you can say AI did it. You will take the full responsibility for it. How much you wanna double check its work? That's your reputation and your life and your company. But in terms of security, That's locked down. So the ethics side of it, I'm not a moral ambassador.
speaker-2: Fair. Can I poke some holes though at your assumptions and give some enchant for those who might not watching? He's like fading into background like Homer Simpson in a in a hedge. I l hey, I love it when people have diverse opinions. That's the whole point of this podcast. Let's poke holes at concepts and idea and we don't all have to agree, but let's have a conversation about the points and disagree. It's interesting how you think about it. And I will challenge you a little bit. You talked about security as being, hey, you know, if you start with all the right things in place, be it purview, be it, you know, mitigations, whatever it is, you don't have to care about the ethics. And here's the thing, I partially agree. But, and there is a but for me personally, in order to get to that stage, you have to start with a set of principles that when you're developing this technology, you say we care about security. Because if you don't care about security, you're never gonna get to a stage where you actually do things for security. And I say that as somebody who has spent a lot of time in our responsible AI framework, where security is a pillar of that, privacy is another pillar of that, fairness is another pillar of that. There's five. but unless you know that you want to care about that, you're not gonna develop something towards it and measure if it works or not. So it's kind of a yes, I agree, but in order to even have security, you have to have a point in your company's whatever history development, whatever you wanna call it, where you say, I care about doing security. Because if you don't, you're not gonna. So that's the only point that I would challenge about what you shared is you have to put a stake in the ground whether you're gonna do a thing or not. And you're not gonna do things that you don't care about or are not a business priority.
speaker-1: I also have to say that there were a lot of beefs there, right? If you've done the purview setup and then you've done the DLP and then you know the subnets and all the networking, da, then it doesn't even matter. Well hell I agree, it doesn't. If you're not all of that work and you have dashboards and you know exactly what's going on and you've got pings and signals of your agents. Yeah, fair. But That's a lot of work there, right? A lot of money. Like you when when you were starting with, you know, is this important or not? It sounds like there's some investment there and therefore for some people who are paying it, they may think it's important.
speaker-0: Yeah. And the people I interact with, I guess the I've been lucky where every company I've ever worked with, they start with security. And if they don't, we convince them to go security first. And I I I think we're all in agreement on security is number one and everything else. It doesn't matter if people can steal your data. Where I think I take a different turn is I'm not gonna tell you about ethics and what you do responsibly with AI. I'll make sure
speaker-1: Yes, all I'm saying is that responsible AI is security. I don't think that you can separate it anymore. Like part of trustworthy and s and and responsible AI definitely is cybersecurity also within whilst I completely agree that Microsoft has locked down the enterprise model of how people should use agentic work in a big company. Totally agree with that. I also agree that this is why people do go to Microsoft and they create their big workloads and their big strategy on Microsoft. There's also a massive, massive motion of open source within Microsoft. So what do you do with like the fund reagents and everything that's been created in Landgraft and like connected back in and all of the tools and the S and Psal? So yeah. That's why I feel like trustworthy and responsible AI can be a massive hole in people's plans in creating stuff.
speaker-2: You brought something super interesting, which is like, you know, open source, barcoding, whole movement, there's a lot more empowerment in terms of people being able to create. And Sean, actually I want your take on this because it it's your bread and butter, right? You help people in terms of AI, you help them really think about the future. What what part while playing this? What's your take in terms of, hey, there's and forget about Microsoft, but there are a million different tools out there, be it Microsoft or not. be a Power Platform, what's your advice to folks, especially as they're thinking of their future when it comes to AI?
speaker-0: We follow a process that it's called ACS when it comes to AI. The first one being, and a lot of what you said earlier around trustworthy AI is doesn't it can be that ethical question, but it's also just accuracy. I care about that more than as you can tell, ethics. I care about when somebody types in something, asks AI to do something, it does the exact thing they asked. Yeah. Whether It is Copilot studio or whether you need to move into Foundry, whether you need to move into a different tool. That's number one is accuracy. If you don't have accuracy, it doesn't really matter. the next point is cost effectiveness. Like how much, like, am I spending the most money on this? Am I spending more than I need to? Do I really need a co-
speaker-1: Password AI for me.
speaker-0: Yeah, yeah. for sure. Next point. so yeah. So then cost effectiveness is the next one. And then which like like Azure just is always cheaper than almost everything in Power Platform. However, you have to go with the aspect of now you have to kind of configure it yourself and you take more risk, which goes into the final step of security. And that's now some of people would argue like, don't you need to put security first? And I'm like, Yeah. I mean, I would say that as an architect or someone who's recorded on a podcast, but when I talk to companies, it goes accuracy, cost effectiveness, then security. I didn't come on this podcast to make stuff up. So that's the order we go in. I think the trustworthy AI framework covers all this and way more. I just know when I've explained it to companies in a way that they listen, I just said ACS, accuracy, cost effect, effectiveness, security. And then that that's made it very digestible for them.
speaker-2: I love that and I love the fact that and we've talked about this in on the podcast and in the in the past of like how do you translate these concepts in ways that matter to people, matter to your stakeholders, matter to their bottom line, and I'm gonna steal the ACS because I think it's such a smart, business oriented, everyday framework that people will just naturally understand because it is what they do at the end of the day. What are some patterns or theme that you are say we are seeing around ACS or around AI in general with businesses? What are things that you're like, huh, this is interesting or hmm, we should be ready for this thing or the other?
speaker-0: Yeah. so in this one, the main one I've seen is it's super easy for AI to summarize and get information. It's awesome at that. It's in terms of creating stuff, like especially like creating content, generating emails, all of that. It's it's it's not good at it. It's just it does its best, but that's the one where we train and educate companies that like you need to be realistic with your expectations of like summarizing info. Yeah. Creating stuff, good luck. And that's where if that's important to you, you're gonna have to spend a lot more money. So I'll give a tangible example. an easy one that's argued all the time, Copilot versus cowork. Cowork's amazing. Cowork costs a ton of money. If you need certain things, like I could demo it right now. I have a meeting with Ana and Ioana coming up next week. They're in three different time zones, coordinate the best time zone and create a repeating meeting every week with them based on when they're free each week. So each week it has to figure it out. Like that type of thing. You can try it with Copilot and it will try. And it's gonna, I've done this multiple times. It's gonna send you 11 invites at the same time on the same day, and you're gonna be like, what did you do? however, if you spend the extra money to go to Co-work, it will pull that off. However, what people don't realize is if you don't ask it that advanced of a question and you just say, like, schedule an invite based on this. Also, go ahead and hop into SharePoint, pull the data from here. Move it into this other folder. You can do that with Copilot. You don't have to spend the money on Co-work. So that's the one where like cost effectiveness comes up the most is like Copilot versus Co-work. and security, it's just not even a conversation because it's all you don't have to do anything, it's already listening to your existing permissions. Accuracy is probably the most fun one, and where I've seen companies usually have to pay consultants to really get is if you want to pay. Not even pay. Let's say you want this is something I think could benefit everyone who's listening to this call right now. What if you had an agent that properly drafted emails for you? Like, like nailed it first try. Not a bunch of like long texts and it's guessing or making up information, but like nailed it. Well, the architecture behind that's very hard. You can try to plug in cowork or AI builder or copilot studio workflows. It'll never be that good. It'll never be that good because every time it's guessing, it doesn't learn from itself. So to build an agent that learns from itself, learns from every email you sent, you get into this like concept called loop engineering. So that handles part A, like accuracy. So I think I went a little off track there, Ioana, but I hope I answer your question of like trying to give you a tangible example of what ACS looks like.
speaker-2: Love those. Ana, I'm gonna go to you in a second. I'll throw you a curve Sean. There's one coming, Scout. So you're gonna soon need to have the conversation of Copilot, Co-work and Scout. And I say that as somebody who's been living in Scout for the last four months, and there's a lot of power in tools that are coming up Ana, what are you seeing in your world and what are themes that you're seeing with your clients?
speaker-1: so first of all I wanted to go to what Sean was saying with the accuracy and I I I love that. I agree with you. That's exactly what we need. Like I want an agent not only to book my next meeting, but I need it to understand from the context of my emails that the meeting's not needed anymore. And to kind of ping me, like according to what you said, it looks like the plan has changed, the leadership has changed, the workshop has changed, you know, the deliverables have changed. Do you still want me to book this? You know? And that's where the power kind of brings, the the power comes from. you know, these agents that we do work on creating very custom to deliver high accuracy and obviously high ROI. and I'll have a question on that in a second, but I'll I'll come back to your to your question, Ioana. What I'm seeing is is so so I come from a pro code world, right? So my customers are like I don't know fast. They move fast. Which means they're not always working under the blanket of like Copilot or even Co-work or even Scout. They go a bit rogue sometimes. Like you're saying that's why people don't go to Claude? I see a lot of people going to Claude, I'm I'm here to tell you. And wherever we lock And I I'm also seeing a lot of organizations lock down Copilot in any of its forms, right? From M365 Copilot to absolutely everything, right? And that's because they don't know how to set up the basis that you were talking at the beginning, you know, with data cataloging and and I don't Defender and and all of the things. And so they are They are worried and they're not unlocking it, which causes their staff to do what? Paste all of this information into a 20 bucks Claude to get what they need. And so and that and that's what we're and that's what I'm seeing. So obviously what we are trying to do is to enable people to deliver that accuracy in whatever medium they are. And my question for you, Sean, would be, have you seen that? And how do you see these jobs going in the future, starting from the person who used to book all of those meetings and had to ration through the logistics of where people are and you know, how would an event looks like and does that meeting event still needs to be there or not, to us creating an agent to do that job for them all the way to these organizations who would where the staff would love to use Copilot or Co-work and they cannot.
speaker-0: Yeah.
speaker-1: What should they do or how do you see these jobs?
speaker-0: Can I ask a couple questions to clarify the question? So one question I heard is like you've actually seen a huge shift in people going to Claude, is one that that
speaker-1: Hundred percent. Yeah.
speaker-0: and I've seen a lot of that as well, counter to what I said earlier, but I'll explain why I didn't mention that one. The other one was a lot of companies also lock down Copilot, they lock down Co-work. and then the third one is like, how do we deal with kind of like shadow IT? When all of this is locked down, how they get around it. And then I think the third question, if I understood correctly, is what is the future of work with all of this?
speaker-1: That's right.
speaker-2: Nice and easy set of questions.
speaker-1: They're all kind of connected if you think about it, right? Yeah.
speaker-0: That's how they worked in my head. So we'll just go in order. First one, Claude's cheaper and more powerful. Like all of Claude's models, I will go on record as say as of today, way better than GPT or anything else. So Claude wins in terms of like price and great models. Every startup I know uses them.
speaker-1: I've got somebody in our company who prefers GPT and will die on his sword for the open AI models. Every There's
speaker-2: There's always an odd liar.
speaker-1: always somebody who manages to like work these models into providing their best accuracy. Let's just see.
speaker-0: Is he a pro coder? Is he writing code with it? Okay. Yes. Yeah, I don't care about writing code, but good for him. So on this point, yeah, Claude Claude's awesome, Claude's cheaper. Last I heard the way Claude stores their data, a lot of European countries aren't comfortable with it. And then the second one is going through Copilot Foundry, you still get access to all of those models in a more governable way. Like, I don't know the governance around Claude because I know it's isn't as good as what Microsoft is doing for Microsoft products. So, like, Cloud is awesome, Cloud is cheaper. I'm going to the first thing you said about trustworthy secure AI. And it's like, now I'm gonna lean towards the companies that want to govern it. If you want to go a little more rogue and pay for Claude and do it that way, like okay, you it'll be cheaper and it will be In a lot of cases, quite accurate. just now governance is your problem. So I I don't want that problem. the other one was point number two, or shadow IT. I don't know why I understand why companies lock down Co-work because the spending can go through the roof. I understand why they would limit Copilot if they don't understand it, if they haven't gone through your AI trustworthy framework. If companies are locking down. Copilot and they're not offering an alternative like Claude which which to me doesn't make sense because now you're like stopping copilot a governable tool and you're introducing a different one to be cheaper and potentially more accurate. Like that is just confusing to me and is a different conversation.
speaker-1: To me they're not. Like they're they're just not offering anything. Right. Like I I have seen organizations just lock down, you know, Co-pilot with the thought of, you know, I'm gonna let that be tomorrow's problem. We've got work to do. Let us fill our Excel spreadsheets, you know, like right now. That's that's what we're doing. And I've seen people then go. away and doing their own thing because everybody wants to get ahead and everybody wants to just keep their job, right? Yeah. It's just the way it is.
speaker-2: Yeah, the reality of that becomes yes, you lock you lock down fearing whatever, but you're just making it so this is where it becomes a not using air quotes, bring your own AI, it's like bring your own beer. People are going to default to AI tools that work to for them. so suddenly you're taking things outside of your secure enterprise environment and putting it in the open with tools that potentially do train. their models back on what is in the consumer or public area because the advantage of enterprise, you can lock down on how that data is used and how that data gets organized used back into the models and retraining. So it's an interesting sort of it's not even a catch twenty two, I don't know what that is, but it's kind of like the you try to you try to do the right thing and I'm using again air codes, but you actually create a bigger problem for yourself because your employees will default what works for them.
speaker-0: Okay, so that that's awesome. Thanks for clarifying it, Ana. So that's the point of some companies are locking down completely. And then I think this ties really well to the future of work for that third group. I'll go on record and say this. If your company is locking down AI completely, I would quit.
speaker-1: Go away. Right. I'm with you.
speaker-0: I think if you're not scared about your job, me, everyone on this call, everyone listening, due to AI, I think you have no idea how crazy powerful it
speaker-1: Delusional.
speaker-0: is. Like GPT Codex, let's reference your boy at your company if it's a boy, I don't know what their gender is. the person at your company who loves GPT, Codex, came out in February, was the first model that wrote itself. Yeah. Can we understand the complexity of that? It it was so Like the most advanced piece of software wrote the most advanced, one of the most advanced pieces of software. Like that's enough that I believe if an AI can get enough data about your business problem, it can outperform any idea you have, any architecture you have. The one advantage we have is that we're humans and we have this like kind of human understanding of other people and business, and we can do that. But overall, it is a time that's never existed like this. I hope it all works out for us. I hope we all get jobs in this AI age. I really do. I also don't understand how, with the way the world works today, how there will be enough work if someone like Ana can do something eight developers used to do when it's just Ana with AI. Where do those other eight developers go? So I believe my responsibility is to get people as trained and knowledgeable as possible. So if this is the Titanic. I can create as many lifeboats as there are.
speaker-1: I love that. I'm I'm with you and I have one more question and I'll let you be. Ioana she'll she's never gonna have me on her podcast ever
speaker-0: Exactly.
speaker-1: again. We're just like off. We're just like yeah. Okay. So I love that. I believe what what you believe with a caveat. I think that's the big governments, the and the big vendors as well. problem. I think that's the problem that they've created and they should really think very seriously about solving. Governments and big organizations should be solving this problem. But I'll give you like a or real life scenario, right? Let's just say you are a company and I don't know, a year ago you decided that you are going to create a new power platform practice. And you've hired, let's just say, seven individuals, and you love them and they're nice people to work with and they've learned power platform appropriately. And they create all sorts of things for you, such as power pages and you've got a bunch of model driven apps and some canvas apps as well and everything's fine. But Hashtag AI and therefore your website should be changing quickly quicker and your processes should be changing quicker and you're a bit afraid that your competition is gonna go ahead. What do you do with that team of our platform developers, consultants, whatever you want to call them?
speaker-0: how many people do I have on in this hypothetical scenario?
speaker-1: I should say seven. Five, seven, ten.
speaker-0: Yeah. I would say in this case Okay, I'm trying not to say enroll them in my training program so they can also be consultants as well as architects but auto, you kinda like softballed that question to me. so I'll kinda
speaker-1: Thank you.
speaker-2: say that enroll them in Sean's training program because that's exactly what I had in mind as Ana was asking that question. I was like I know this is not for me but the answer is right there. It's upskilling
speaker-1: Let's talk about
speaker-0: It's yeah.
speaker-2: help people learn the
speaker-1: the record, Sean did not tell us to ask him this question at all. Right. But no, seriously. These are the things that are happening every day. How will these people understand that before okay, fine, before I used to do a power automated flow or a bunch of apps and I had like user stories and fields and processes and so on? Now I need to build I'm I don't know your training program. in enough detail, but this is what I would do. Now I need to build connectors, MCPs, things that are able to be used by AI and, you know, accelerate things.
speaker-2: Can you tell us about your training program, please?
speaker-0: Yeah. Speaker of the word accelerate, it's literally in our title. So our
speaker-1: Funny.
speaker-0: Is the AI Architects Consultants Accelerator. Let's just jump to the tech right away, honestly. That's what you were asking about. The first thing we do is we get everybody started in first of all, we explain what I said earlier, which is all these tools are going expire. Anything I teach you is not gonna be really that relevant in a year. It's like you need to set up the coding and software foundation for yourself that you have AI do as much for you as possible. So everything we teach is from VS Coach.
speaker-1: I thought you don't care about you you you absolutely don't care about anything code. I think that's what I heard earlier, no?
speaker-0: I think we can replay the recording because I don't think I ever said that sentence.
speaker-1: Okay.
speaker-0: I said I don't care about being the best pro coder. Even though I'm pretty sure I said it as I don't care about anything code, is were the exact words. My actual meaning is I'm focused less on pro code. So here's my point. When I say less on pro code, I mean getting into the very specific details of like do I need to get into I don't know, like Multitask processes. Here's the point. We do teach Pro Code after everything I just said. So in this case, we show them how to use VS Code and pretty much build all of your power platform solutions from there. The main thing we do is show you how to take all the most complex software problems from data migrations to security to licensing to deployments, copilot, AI, fast apps, performance, all of this. We take all this and we show you how to build it as quickly as possible and as flexibly as possible for the enterprise to large scale. Now, the big difference is saving your time is one thing. Like, sure, you can spend the rest of that time now going on vacation, spending it with your family, whatever you want. We feel like now that you have all that time saved, you need to understand the business side of power platform. You need to understand how to work with stakeholders, you need to understand. Pretty much on all the skills you've built over years of your career of like how do you work with people that are scared of copilot, that want to shut down governance, that don't even understand how to begin and hack that don't know what to do with tech because they're not a tech company. Like, how do you help them? And this is what we train them to do. We believe the future of work is you hop into something like VS Code, it codes up your solution for you. Your whole job is understanding the client. And we teach you under how to understand the client.
speaker-1: B N F D
speaker-2: That is brilliant. I love that. And I love the I love the pairing, right? It's like it's the technical skills, it's the business acumen, it's the AI acumen. But here's the real world stuff. In terms of you can have all of that unless you understand your client and can work with your clients. It's you can have the best coded app in the world. It's not gonna make a difference. I love it. So in the interest of time, I always ask our guest a closing question and today I'm actually gonna ask you both. If you were to give our audience homework, what would that audience what what would that homework be? Ana, you go first.
speaker-1: Find your passion. What moves you when you go into a room? How does that how is what you do change the room? Find your passion truly? I know it sounds privileged, I know because you're doing a job and you're busy and you just want to keep it, but I'm serious. Find your passion because if you find your passion and what you like doing, what sort of problems would you like to learn more about? That's where you're gonna be able to do exactly what Sean was saying, you know, understanding your your customer, being able to quickly enhance their you know, their ideas into in into solutions that are not as our friend Dona says. faster horses, you'll be able to to to reinvent a problem statement and create a car, right? Find your passion. That would be my number one advice. Write it down every single day. What did I like about my day today? And do that.
speaker-2: I love it. Sean, what homework would you give?
speaker-0: Besides follow me and my YouTube channel, besides that home.
speaker-2: Obviously. That is my homework for the audience.
speaker-0: I mean, I really liked Ana's advice, so I'll go off of that. get some rich friends. Make rich friends. All your, you should be the poorest in your friend group. If you're not, I don't understand how you expect to like move up if all of your friends have the same experience and the same amount of salary than you. Now, obviously, you may have some friends who are very poor and have a lot of free time. Well, they're doing something right. So you can learn from that. The point is. besides find your passion, what's helped me find it is hanging out with very, very successful people.
speaker-2: It's funny you say that. I'm going on a tangent, but I was reading this study that influences work. And when you're working with high performing people, it rubs off you. And when you're working with people who are less performing, it also influences you as much as you would want it not to. So selecting your tribe, I'm just gonna call it a tribe. Selecting your tribe and the people that you surround yourself with, super, super important. That's something that I fundamentally believe and have believed for a lot of my adult life is be peaky with the people around you. They're gonna influence you for good or for bad. so it's a choice. Folks, thank you so much. Sean, you're definitely gonna come back and also will you. this podcast is all about, hey, let's poke let's poke holes, let's have different ideas. That's what the beauty of it is. Otherwise we will make an AI generated draft and read it to the audience. That's not what it is about. So folks, thank you. thank you for our audience for tuning in and we'll see you next time for another episode of Human in the Loop
speaker-1: Thank you.
