Confident, strategic AI leadership

Allegra Guinan of Lumiera helps leaders turn uncertainty about AI into confident, strategic leadership. In this conversation, she brings some actionable insights for navigating the hype and complexity of AI. The discussion covers challenges with implementing responsible AI practices, the growing importance of user experience and product thinking, and how leaders can focus on real-world business problems over abstract experimentation. Featuring: Allegra Guinan – LinkedIn https://www.linkedin.com/in/allegraguinan/Chris Benson – Website https://chrisbenson.com/, LinkedIn https://www.linkedin.com/in/chrisbenson, Bluesky https://bsky.app/profile/chrisbenson.bsky.social, GitHub https://github.com/chrisbenson, X https://x.com/chrisbensonDaniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaLinks: Lumiera https://www.lumiera.ai/Sponsors: Shopify https://www.shopify.com/in?utm_source=podcast&utm_medium=audio&utm_campaign=us-all-na-awareness-1q24-en&utm_term=practicalai&utm_content=practicalai – The commerce platform trusted by millions. From idea to checkout, Shopify gives you everything you need to launch and scale your business—no matter your level of experience. Build beautiful storefronts, market with built-in AI tools, and tap into the platform powering 10% of all U. S. eCommerce. Start your one-dollar trial at shopify.com/practicalai http://shopify.com/practicalaiRegister for upcoming webinars here https://practicalai.fm/webinars!

• 0:03Speaker 00

  Welcome to the Practical AI Podcast, where we break down the real-world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X, or Blue Sky to stay up to date with episode drops, behind-the-scenes content, and AI insights. You can learn more at practicalai.fm. Now, onto the show.

• 0:48Speaker 01

  Welcome to another episode of the Practical AI Podcast. This is Daniel Whitenack. I am CEO at Prediction Guard, and I'm joined, as always, by my co-host, Chris Benson, who is a Principal AI Research Engineer at Lockheed Martin. How are you doing, Chris? I'm doing very well today, Daniel. How's it going? Yeah, no complaints. It's been an interesting week of AI things, with, I guess, OpenAI being open again. I'm sure we'll talk about that on this show at some point. But maybe there's a connection there to responsible practices. But I'm really excited today

• 1:32Speaker 01

  because I've seen a lot of things popping up on our friend's channel, Demetrius over at ML Ops Community from Allegra Guinan, who is co-founder and CTO at Lumiera. Welcome, Allegra. Great to have you here.

• 1:49Speaker 02

  Thank you so much. Great to be here.

• 1:50Speaker 01

  Yeah, yeah. Well, I kind of alluded to responsible AI things, which is certainly something I'm sure we'll get in. But it may be useful just to hear a little bit of your background and kind of how you got into what you're doing now, which I understand is advising business leaders around AI principles and responsible AI practices and strategies and that sort of thing. Correct me if I'm wrong. But yeah, I would love to hear a little bit of that background and kind of how you arrived to doing what you're doing right now.

• 2:26Speaker 02

  Yeah, for sure. I have sort of an unconventional background and path towards CTO and co-founder. It's definitely not something I ever thought I would land at, but happy I did. I got into the tech scene over 10 years ago now. I was living in San Francisco. I'm actually based in Lisbon now, similar cities in some ways. But I was in the heart of the tech scene sort of by default. I had actually studied in the arts and it was a different path than I expected. And I started working at a startup that was doing 3D visualization that was quite ahead of its time. And I was working on the data platform team there, so I worked a lot with backend engineers

• 3:08Speaker 02

  and figuring out what our data architecture would look like essentially for this 3D visualization e-commerce that was working in the interior design space. So that was obviously very new to me as my first tech job. And I had started in more of an ops role and then moved quickly into product management. So I was building out internal tools, which is why I was working so closely with those teams and focusing then a lot on search optimization, figuring out which words would yield which results. You know, couch, sofa, you get the same sort of thing in the end. Seems very basic now especially, but back then it was a lot of work. And I got really excited working with data teams and it kicked off this sort of interest and passion in a way for technical projects.

• 3:53Speaker 02

  And then I moved over into the fintech space. I was working at Chime, which maybe you've seen them in the news recently. I was also working on the data team there and really saw it scale. So I was a part of the technical program management team. So I was managing technical portfolios with a lot of different programs, usually longer scale and multiple year data programs, and figuring out how to get real-time data to those who needed it within the organization. And that often meant ML engineers. So that was sort of my first touch point in the ML space, working a lot on fraud and security within the finance space, which was also extremely interesting. And I got further into this.

• 4:34Speaker 02

  Uh, data engineering world and I built up this dev advocacy with within me and I really loved working with those teams. And so then I shifted again to CloudFlare. So I was working as a technical program manager there. It was a multinational or is a multinational organization, of course, so a bit bigger scale again than I was working on previously. Uh, and I got to then get deeper into a lot of AI initiatives across the enterprise. So that was sort of my path and I met my co-founder at Lumira, Emma, who's the CEO, around four years ago on a beach in Lisbon. And we didn't realize we had this similar

• 5:14Speaker 02

  interest in the AI space and in technology. And then many years after we already became friends, we we realized that. And she had already had an organization around consultancy and business strategy. Uh, and we came together and formed what is now Lumira about a year and a half ago. Um, and really it has grown over time, especially as the the space is shifting so much uh with technology and AI. But as you mentioned, we are really trying to guide leaders in making more responsible decisions around this new era of technology, which is how we got the name, Lumi for Light, Era for Future, so brighter future. And it's something that I'm I'm really excited to be a part of and to be uh leading in. And

• 5:58Speaker 02

  I'm I'm glad that, you know, my voice is sort of starting to come across the ecosystem in in various places.

• 6:04Speaker 01

  That's great. When you were talking, I was thinking back to a conversation Chris and I had, I don't know if it was a couple of episodes ago, but we were talking about the services industry kind of more broadly. And it it does seem like there's such a need for good, you know, responsible strategy and insight from the services standpoint around AI. But also, there's kind of like on the other end of that AI eating up some of the maybe services industry around certainly around whether it's marketing agencies or maybe like. Prototyping new projects and that sort of thing.

• 6:45Speaker 01

  So what is it like, I guess, being, being part of that services industry in this age of AI? And how do you see kind of the role of, you know, advisors, service providers either, you know, shifting or, I don't know, what, what are your thoughts around that in terms of, you know, the best value that even kind of services companies can provide during this time where maybe certain areas of what they provided before is getting, getting kind of gobbled up by some of the things that the AI is providing?

• 7:20Speaker 02

  Yeah, there's a lot in there. So I'll start with sort of the challenge of being in this space. And that is that we're not offering a silver bullet. We're not offering a tech product that will solve all of your problems. We are not just handing this over to you as an answer. We're really developing leadership. So our core product is a executive education program that spans eight weeks. And we are covering all of the challenges that we're seeing right now in AI that are not necessarily technical. It's a very human centered approach to technology. And of course it's easier and people want something that's fast and they just want the answers and what should I build? What should I buy? Like, how do I do this? And then can I pay you to do it for me?

• 8:05Speaker 02

  And we're really offering a counter narrative to that. And instead asking you as a leader, you have already built yourself up. You're, you're leading this organization, however big it may be. What can you do now to make the right choices instead of offloading that onto somebody else? And it's really testing folks and it's putting them up against a wall sometimes to be a better leader. That's a personal choice. Not everybody wants to invest time in being the best version of themselves or maybe they're tapped out and they don't want to do this anymore. That could also be the case, but we're working with the ones that are trying to position themselves as really leaders in this, in this age that we're in and to continue to

• 8:48Speaker 02

  lead and to bring their entire organization around, because I'm sure you're seeing there are a lot of failed projects, a lot of missed returns, expectations that were not met with AI in the past couple of years. And most of that is a human issue or a leadership issue. It's not a technical issue. The tech is there. We are missing this communication and translation layer, and that falls on leadership in our minds, which is what we're trying to address.

• 9:15Speaker 03

  It's really refreshing to hear that you're, that you're trying to change the narrative, because I mean, I would imagine, you know, I know that Daniel and I are constantly bombarded and, you know, with different companies that are out there telling us that, you know, they have the solution. It's AI and it will solve everything having to do with that. And that's so common. And so, um, kind of hearing it being grounded in leadership, I guess, kind of going toward a question here, I would imagine it is very hard for leaders or aspiring leaders to, to kind of process the never ending rapid change that's occurring with

• 9:55Speaker 03

  this space. Because, you know, over the, over time, having seen decades of business and stuff, this is much faster cadence than it has ever been. We've seen change over those times, but it's, but now, you know, literally, you know, every week, there's a collection of new things to consider that are hitting you up, you know, and whatever you're at. So how do you, how do you deal with that? I would imagine that there, that your leaders are coming in with some level of anxiety and some level of uncertainty of like, you know, and maybe even some fear of making decisions that are going to bite them on the rear end by, you know, not, not very far down the road by making choices now. How do you get through things like that? How do you get through that kind of fear and anxiety?

• 10:41Speaker 02

  Yeah, I mean, you called out some of the main challenges that we're seeing. And through many conversations with many leaders across the board, we're hearing the same thing. So one is this like noise exhaustion and information overload of trying to keep up. Another is fear of missing out and getting left behind. So you either end up in this position where you're sort of paralyzed and you're not sure what move to take. And so you're getting left behind in a sense, or you fear that you are. Or you're moving really quickly, you're making a lot of decisions, but they're not necessarily the right ones. They're not grounded in anything. It's just based off of this reaction to what you're seeing around you. And that could be a very narrow echo chamber of information that you're being exposed to.

• 11:21Speaker 02

  Maybe you only check Twitter for your updates, or you only check one sort of newsletter to get your information. And you're not creating this landscape of multiple sources of information to decide which move to make. So yes, there is this stress, anxiety and exhaustion that we're seeing, which is also what we're trying to address. And we do that by not focusing on every latest model, what the latest hype is. We're focusing on what your challenges are as a leader in your organization. And that is not going to change every single week, you know, in theory. Maybe it is for some folks, but if you're a mature organization and you have strategic goals, probably you know what they are, or you should know.

• 12:01Speaker 02

  And then you can start to address what the technology is that would help you achieve what your goals are related to those goals, related to those challenges, sorry. And so it doesn't matter if there are 10 new models that came out last week. You don't need to know what all of them are right now as a senior leader or as an executive. You need to understand what numbers you're trying to shift, what kind of transformation you're trying to move forward within your organization, within your workforce. And then you can find, you know, iteratively, the best solution technically for that. And so we're trying to help people. Build that mindset and that scaffolding to understand the ecosystem. So we do have a section in our program. We have it split up into three foundations.

• 12:45Speaker 02

  The first is is on confidence, which I can go back to in a second. But the second is around action and that's understanding risk. And it's understanding the industry and industry radar. And it's about setting your your vision for AI, not for your organization, but as yourself. What is your personal stance as a leader? What do you care about? Is it security, privacy, transparency? What are those principles that really resonate with you? That you can then use to make your decisions. Once you have an understanding of how to evaluate risk, once you understand what's sort of out there in a general sense of capabilities, not every single minute detail, but a general understanding. Then you can start to think about what opportunities you have in front of you as far as use cases.

• 13:26Speaker 02

  So we really do focus on that rather than trying to put a lot more information in terms of like technicalities in front of you. And then just going back to the the confidence portion. So we have that as our first foundation because we want people to develop this mindset as leaders of, OK, I already have a lot of strengths to move forward with. I've already built myself up and my organization up. Understanding and knowing every new model drop is not going to be the differentiator here. It's how I communicate with my workforce, how I keep people engaged, how I can manage everything that's transforming in front of us and keep people, you know, excited to be here and to be a part of what we're building, whatever that is. And so you need to have that resilience as a leader and that confidence in yourself

• 14:11Speaker 02

  and to be informed before you can start taking action, before it even makes sense to start reading all of the latest news, because it won't mean anything to you unless you have that personal understanding.

• 14:30Speaker 04

  Well, friends, when you're building and shipping AI products at scale, there's one constant: complexity. Yes, your regular models, data pipelines, deployment infrastructure. And then someone says, let's turn this into a business. Cue the chaos. That's where Shopify steps in. Whether you're spinning up a storefront for your AI-powered app or launching a brand around the tools you've built, Shopify is the commerce platform trusted by millions of businesses and 10% of all US e-commerce, from names like Mattel, Jimm Shark, to founders just like you.

• 15:04Speaker 04

  With literally hundreds of ready to use templates, powerful built in marketing tools, and AI that writes product descriptions for you, headlines, even polishes your product photography. Shopify doesn't just get you selling, it makes you look good doing it. And we love it. We use it here at ChangeLog. Check us out, merch.changelog.com. That's our storefront and it handles the heavy lifting too. Payments, inventory, returns, shipping, even global logistics. It's like having an ops team built into your stack to help you sell. So if you're ready to sell, you are ready for Shopify. Set up now for your $1 per month trial and start selling today at Shopify.com/PracticalAI.

• 15:48Speaker 04

  Again, that is Shopify.com/PracticalAI.

• 15:57Speaker 01

  Yeah, Allegra, it's really encouraging to hear your perspective. I can second Chris there. I, of course, when we're working with customers, when we're talking to people on the podcast, when we, you know, interact in our companies, all the time hearing like, "Oh, this new model came out and, you know, now OpenAI has open models and I need to, you know, should I switch?" And yeah, I think having the sort of internal piece that like even if no one ever released another model, you have more than enough to be like very transformative in your organization. Like, don't, don't worry about it. There's a long way to go there. Yeah. So I think that that's really interesting.

• 16:41Speaker 01

  I love also the perspective on leadership because one of the other things that I think we're seeing a little bit, and I would love to get your perspective on this, is kind of the executives in a company kind of dictating.

• 16:57Speaker 01

  Like we are now going to transform with AI, right? And everyone in the company really not understanding what that means practically. Or, you know, leaders like, oh, I'm a manager in an engineering team and I want all of my developers to be more efficient, you know, so I, you know, dictate to them, all of you need to be using these AI tools, right? And really, like, no one ends up using them or there's like, everybody, everybody kind of has the workflows that they're used to. And so there's really not that kind of trickle down transformation that that happens. I'm, I'm wondering about your perspective on that.

• 17:40Speaker 01

  And maybe even for me, as a, as a leader of a team, you know, in, in my company, I really want to understand that element better because I want to both lead by example, but also understand, to your point, how to lead well, my team forward in a way that is embracing the right AI technology and, and being transformed. But I know that I also can't just walk in one day and be like, everybody use more AI. And then I go sit at my desk and even whether I'm using more AI or not, it really doesn't, doesn't matter.

• 18:16Speaker 02

  Yeah. I mean, this is one of the critical mistakes that we're seeing now. And there have been some recent news stories coming out of leaders sort of having to roll back their AI first organizational approach because they weren't expecting the backlash that they got. They sort of assumed everybody was thinking about AI the same way as they were, uh, which is not the case. Everybody's coming to this from a different level of literacy, from a different perspective. Everybody has a past relationship with how they view this technology as it relates to complexity, if it's actually more useful or not. So you can give people, you know, any tool you want, you can give them a stipend, like free money, go try whatever you want. But unless you help them understand why or what it would help them solve and really bring

• 19:01Speaker 02

  up that level of AI literacy across your organization, it won't make a difference because people won't understand what you're trying to do here. And unless you also communicate that clearly as a leader, you're going to come across as very prescriptive. I think, you know, especially in the engineering space, we all know that that's not ideal. We don't like when people are super prescriptive and just tell us what to do, right? We like to explore and to do research and to get there on our own. And what's interesting about AI is that this is really coming from the bottoms up in a lot of ways. Three times more uh employees within organizations are using AI than their leaders think. That was from a recent report this year. Uh so it's not that people are not ready or can be engaged.

• 19:45Speaker 02

  But it's meeting them where they are and having a conversation that's very honest. So what are you using AI for? It shouldn't be stigmatized. If you want to encourage usage, then help people understand that it's okay to share where they're using AI and why they chose to do it that way. Uh have, you know, open sessions where you're sharing with one another. Establish this like sort of AI champion culture and a fail forward culture as well. You have to invest time in experimentation and research and know that it's not all going to be perfect. And to make people feel like that's okay. And that they can try these things and share openly, because if you don't do that, then it doesn't work. Like, people won't use it if if they're not part of it and if they're not involved. Um and they already are using it, that's the thing.

• 20:29Speaker 02

  Most people are using something to the side of their work, whether or not you put it in front of them, you know, purposefully or not. They're using ChatGPT or Claude or they're coding with something on the side. They have a AI-driven IDE. Something is happening uh, you know, in the organization, whether or not you built a a program or initiative around it. So it's better to do that in a very open way and an honest way where everybody is involved. One of my favorite things that I worked on at CloudFlare was piloting different um coding assistants, AI coding assistants. And it was a large group of engineers from various teams and a lot of it was qualitative. Um and this was sort of my approach coming in and I don't know if they liked it. I think they did because the the results were good in the end.

• 21:14Speaker 02

  But it's a lot of just understanding what people like and having a lot of channels of communication for like, did you try this thing out? How did it go? We're gonna give you, you know, uh a fully supported space and time to invest in trying this. And then we're gonna do it with something else and we're gonna compare them very honestly. Because we're not just gonna choose a solution that seems the best on the market right now. We're gonna choose the best solution for you, for this specific group of engineers that are part of this organization. Um and so I think. Being quite humble as a leader in that sense, too, that you don't know the best thing for everybody. You know, you're at the top, you don't have your hands in every single initiative. You shouldn't anyways, that's my point of view. So you have to trust that the people that you hired have, uh, you know, an opinion that's worth hearing and then give them space to, to share that.

• 21:59Speaker 03

  You mentioned something just now that I was really wanting to, to dive into. Um, you mentioned the word trust, uh, coming in there and, and that's complicated. And there, there's trust in multiple directions. There's, there's not only the, the trust that the leader or leaders must have, uh, in the teams that they are overseeing, but there's also the trust of those being oversaw, overseeing that, you know, that are doing the work, the engineers and trusting in the motives of their leaders. Um, and that, that raises some, some interesting things, several of which you've, you've mentioned. You know, there's, there's the, as you pointed out, there's this, uh, reality that, uh, employees are using AI, uh, in, in areas that maybe they have even been told explicitly not to,

• 22:44Speaker 03

  or, or, or at least they're, they're finding a place to kind of, to kind of bring it in, whether it's noticed or not. And you also have these top-down, you know, vouch, shout, go forward and use AI, um, with employees worried about, you know, what does this mean for my job? You know, and job security, uh, is this AI eventually going to replace me? How, there's so much involved in this and, and probably more so than I've observed in the past when we've had, you know, cloud computing was a pre, you know, before AI, you know, the AI wave, we had the cloud computing wave and before that we've had other waves. And there wasn't, there, there was a trust in the technology and privacy and stuff, but there's now a, an implicit trust within your own organization that exists, you know,

• 23:29Speaker 03

  those factors. How do you address that with, you know, with, with leaders as you're getting into? And, and get them, if they don't recognize that upfront, how do you get them to recognize and take action on, on that, that kind of new dynamic that's now in the workplace?

• 23:44Speaker 02

  Yeah, and trust is so critical here at, at every level, at the technical level, at the, the human level across the board. I think the way people notice this unfortunately is when things don't pan out. Um, or there's some sort of, you know, internal rebellion as we're seeing with this backlash that I mentioned. Or they're not seeing again the returns that they expected because people didn't adopt in the way that they anticipated. It's because there wasn't that relationship building and that trust. Because the people that they're trying to involve, you know, the workforce were not a part of those decisions. And if you have a, a vision as a leader and you want to be AI first across everything. But you're not communicating that, you didn't set any standards, you didn't publish any policies that help people understand what the.

• 24:26Speaker 02

  What's okay and what's not okay, then that doesn't uh elicit trust in the environment. Again, it just comes back to something that feels very um top down without involvement. Uh, which won't lead to any results. And then there's trust in what you're actually building and this is, you know, something that I also try to advocate a lot for because right now. Engineers are the ones really pushing this forward in a lot of organizations. There are groups that are just trying things out, you know, again, whether or not it was dictated that it should be done, that's just sort of what's happening. And so what's being built might not necessarily have trust built in by default. Because that wasn't something that was thought of at first, maybe you're just trying to build something cool.

• 25:08Speaker 02

  You got access to something, a new model came out and you just want to throw something together. And that can sometimes escalate to being now used by the company or, you know, some leader wants to see that in production, even though it wasn't really tested thoroughly. And how can you expect somebody internally, if you're building for, let's say, another internal user to trust what you've built if you don't communicate. Why it was done or how, and you can't really explain where the outputs are coming from and. There's no, you know, documentation around it, we're sort of, sort of abandoned like the product. Approach and thinking and anything around documentation. Um, or thorough testing when it comes to AI, it's sort of just like throwing things out there and. Some things go into production and are used and sometimes they work, but a lot of times they don't.

• 25:53Speaker 02

  And so it's it's hard to build trust when you're moving that way uh without a lot of intention and without a lot of clarity that you can express to other people. Um something that we see failing a lot, you know, even when something works really well technically and it's sort of perfectly executed. If you can't explain that to maybe somebody in risk or compliance, it's not going to get very far. And they won't be able to to roll that out and it won't be, you know, it won't be trusted, even if you as an individual that built it feels like it's good. Um so again, it's it's about this transparency and open communication as you're going. And why you can't really abandon documentation and you can't abandon, you know, the reasons that you built things um or having observability or logging.

• 26:35Speaker 02

  It's it's not enough to just make something that seems really cool. You have to actually back it up and to be able to explain it to to the others around you.

• 26:43Speaker 01

  Yeah, I some of what you said there is definitely applicable like internally and externally, but certainly a lot internally in terms of the documentation, how reliable something is, the testing, all of that sort of thing. Part of what I was thinking in my mind while you were talking is uh in internally uh here, we we like to uh talk about certain ways in which we would like to build things. And one of those things that we talk about is we would like to build things that kind of restore trust in human institutions rather than further erode that via like AI and and automation. And I'm wondering from an external standpoint, I mean, it's one one side of this is internal,

• 27:25Speaker 01

  how you kind of integrate AI features, test them, deploy them, et cetera. It kind of gets to another level when, let's say, you're releasing your voice assistant, you know, publicly to, you know, the world or you're rolling this out, you know, to your external customers. And you say, hey, this is our new AI feature. And that could go a lot of different ways, some of which, like I say, could erode trust further with your customers. Hopefully, it's not already low, but maybe it could erode some of that trust that you built up over time. But maybe doesn't have to be that way.

• 28:07Speaker 01

  What what have you found to be some of those kind of key principles that leaders could keep in mind as especially as they're releasing things to their users or their customers or to the public that can help? The public understand, or their customers understand that this has kind of trust built in, I think is what is how you phrased it.

• 28:29Speaker 02

  Yeah, what's interesting is that the gap and sort of disparity between what's going on internally with AI and what's going on externally is so wide. Like the experiences are so different from what people are building for their own teams and then what they end up putting in production. Maybe we can come back to that, but it's just something I see very obviously in the space. But I think one thing that's super important here is the user research. Right now, everybody is putting AI into their products everywhere. You know, if you have a bunch of vendors that you work with as an enterprise, you'll see now that all of them are offering AI and they're all sort of offering the same AI features. And maybe you asked for them and maybe you didn't. And so I think understanding still your user base, they might not need something to change

• 29:11Speaker 02

  or the thing that you that they do want to change, you might not need to use AI for it in the way that you think. So really asking and understanding your users before you start deploying that kind of experience. Because then if they didn't ask for it and they didn't actually need it, then why would they trust it? And why would they start to be happy that it's out there? Unless, you know, it makes the experience so much better. But a lot of times it doesn't because this is still quite nascent for a lot of organizations. So that's one thing. And then the other again is the transparency. So as a user, for example, and we can go into financial services. I think it's a really important industry and we think about the financial services a lot in terms of of risk.

• 29:53Speaker 02

  But let's say that I am now a user of some sort of fintech app or something around my finances and you've put AI in there and I see something in front of me that I don't understand. And I ask like, how did you get to that decision? Or even if I'm very technical, like what model did you use to get here? And you don't have an answer for me, like that will erode trust. And that's something you need to think about, especially as people are becoming. More literate, but also sometimes at a shallow level. Like, they can ask a question. They might not fully understand what they're asking, but they might ask you something, and you need to be able to respond to that. And so, again, the documentation, do you have system cards in place? Do you understand what your guardrails are? Are they documented somewhere?

• 30:38Speaker 02

  Do you have, you know, tracking? Do you have system versioning? Like, can you actually back up what you've done? So that when somebody does ask you a question, and they're looking for that confidence in you, they're looking for you to bring back the trust, right? And it's an opportunity for you. And if you don't have an answer in that moment, you will erode trust with your external users. So I would say for leaders thinking about that, to be able to ask those questions first, and make sure that they have everything in order before they start deploying. And then, when you do have something that is driven by AI, being explicit about it, and if people are uncomfortable, maybe it's because they don't know enough. And it's your job, your responsibility as a leader in that space,

• 31:20Speaker 02

  to maybe educate them within your product and the experience around what you're offering. So if they're off put by it, understanding why, understanding your users, like, that has not changed. But again, somehow, it's gotten lost, where suddenly we don't really care what users ask for or what their feedback is. And I think we really need to go back to that.

• 31:39Speaker 03

  So I'm wondering, as we've been talking a lot about kind of adoption and the trust issues that go with that, one of the things that I've been thinking about and kind of curious on how you're approaching it is kind of how you position the notion of responsible AI to different organizations that you're working with as you're going through this educational process. And it's like, because there's not a single golden path, you know, down that road. There's a number of organizations that have weighed in with different types of policy guidance and such on that. Everything from government, multiple government organizations,

• 32:22Speaker 03

  to non-government organizations and nonprofits. And so, as a company is looking to try to make their AI strategy work, and they're starting to address these various things we've been talking about, how do you guide them into framing that whole effort? So that they, you know, because it's a little bit different. There's not just a go do this, and they're done. How do you approach that?

• 32:49Speaker 02

  Yeah, I mean, first, there's no standardized definition of responsible AI. You know, there are a set of principles that a lot of people agree on and they're sort of overlapped when you see some of these policies that have been put out, but there isn't something single across the industry that everybody has aligned on, so that's one thing. The second is that a lot of leaders don't care about being responsible. That's just the honest truth. They care about their bottom line and they care about financials and they want to see numbers move and that's it. And so a lot of the framing does come back to that. And luckily, being more transparent, being more accountable, having robust systems, all of these lead to better results and more money at the end and better trust with your customers. So luckily, it lines up that way.

• 33:31Speaker 02

  But it is a different story to tell and I think leaders need to start understanding that having this, you know, these answers to your customers when they come asking or being compliant and not facing major fees when you are not compliant, especially here in the EU, for example, and being able to lay it out in a very financial way that makes sense is sort of where to go in here. And then, again, we're, of course, trying to shift the mindset of leaders to understand what their own principles are and what they actually care about. And a lot of times, organizations already have these. They might already be security first, they might already care about privacy, and so you can use that lens. So if you already care about privacy, are you thinking about access management and secured access when you're building your AI systems?

• 34:15Speaker 02

  A lot of people are not right now and that's sort of like a gap that we're seeing where they have all of these governance structures in place, but then they built an AI system that like completely erodes all of that and sort of like finds its way around and they didn't think about that before. And so you can use their, their own values and their own framing of how they're running their business and tie it back to how to be more responsible in practice. And we are seeing that this is changing a bit in terms of how companies are presenting themselves, sort of going back to. Financial services, for example, I was looking through, you know, the, the top banks that are leading in the AI space right now.

• 34:54Speaker 02

  And what we've seen mostly in the last couple of years is a shift in what they're presenting externally in terms of explainability and their responsible practices and their leadership. They have a lot more people talking externally about how they're handling AI, uh, sort of giving more insights into how they're approaching. And so I do think that the tides are shifting because people are realizing that you do need to come, at least come off as if you care about it a bit, and maybe along the way you will actually start to care and make some, some differences. So that's, that's how I think about that. But for myself, I also approach this from the engineering side. So because of my background and because I've built my entire career alongside engineers,

• 35:35Speaker 02

  um, a lot of what you want to do as a good engineer to build good products aligns with these responsible practices as well. So when you have testing in place, when you do have security in place and observability and you understand what you've built and you do, you know, again, have these like, uh, options for versioning and you have your ML ops figured out, like you will have a better outcome when you're building an AI system. And all of those things also benefit on the responsible side that then you can have other teams looking into to understand how you got to that point. And again, you're, you're bringing in the, the multi-discipline, multi-disciplinary trust that is so necessary for this.

• 36:14Speaker 01

  As you're discussing things with leaders around kind of responsible practices, you know, how they should lead out with AI strategy, that sort of thing. One, one of the questions that's come up in my mind as, as you've been talking is, is the appropriate level of literacy around these subjects on the, on the technical side that a leader does want to have. Cause I've, I get into so many discussions, like because we're kind of intersecting, you know, my company intersecting with the world of security, talking to like CISOs or CIOs or whatever. And they're, you know, they'll say things like, yeah, we have our own model. It's running internally, none of our data leaks.

• 36:59Speaker 01

  And you sort of probe into that a little bit and you're like, no, actually what you just have is like an API key to a model endpoint that's not running in your infrastructure and all of your data is living at rest in someone else's infrastructure. There's just like such a wide gap between what they apparently think they have and what they actually have. And I understand like the AI, us as AI people have probably not helped that because we've sort of obfuscated some of that terminology and made things maybe seem like they kind of are what they aren't. And so I kind of feel sympathy for a lot of people that we have made it extra hard for people to gain this literacy, maybe.

• 37:45Speaker 01

  But around things like model hosting, open, closed, fine tuning, like all of these things are like very confusing for people. What is your recommendation as you're going through this material with leaders around the appropriate? Like if there's leaders listening in our audience, you know, what should they be expected? Where should they be expected to get to technical literacy wise to be an effective leader around AI things?

• 38:12Speaker 02

  Yeah, this is definitely a challenge even for me because I'm in this like every single day and I listen to these terms all day long and I enjoy hearing about them. And you know, I'm listening to these kinds of podcasts. And so to distill down what is actually critical for people to understand is even hard for me because I'm like, oh, I I want you to know everything that I know and that's obviously possible. And so I think something, an easy way to approach this is sort of what you just called out as well as like what's happening within the walls of your organization, like that's a really good place to start. If you do have a system that is, you know, you have an API call to some model, open or closed or whatever, like who built that? Do you understand what's going on in your own infrastructure?

• 38:56Speaker 02

  Who is designing this? Do you have an architect? Do you have a technical leader? Do you have a lead engineer? Do you have a CTO? Like somebody should be responsible for understanding what has been built and why? And if you don't understand that and you don't have a person and you don't have a relationship with that person, that's your first problem. So I think starting there and knowing like, okay, what do we actually have going on here that's in production live right now? Like, let's walk through what those terms are, so I can understand now how to navigate this space, like, right now in front of me, rather than every single potential capability out there that other organizations are using, because that might not be super helpful for you right now. So I would start with that. Another one is sort of abstracting it and

• 39:37Speaker 02

  focusing more on what you're trying to solve. Again, like what we talked about, so it doesn't really matter if a hundred new things come out, as you mentioned, like, you probably already have the capabilities out there to solve what you want to. So asking good questions and understanding, like, okay, I care about security, that means I don't want this to happen. Is that happening? Like, when I make this kind of call, is the data leaving? Like, you don't need to understand everything in that moment, but you do need to understand what kinds of questions to ask, which is also why the principles are so helpful. Because if you care about things like transparency, or you care about, you know, supporting open source, we can go into that as well, like you mentioned, um, then you can focus on a specific set of terms or concepts to really understand. But, like, even before that, you

• 40:21Speaker 02

  know, people don't understand what AI is, they don't understand what traditional ML is. A lot of organizations are still running, like, legacy ML systems and modern AI, and then they started adding GenAI, um, and they don't have a, an understanding of what those concepts even are. Or, you know, Agentic is thrown around a lot, and that definition varies a lot, too. So I think there are some that it's like, if you are hearing it out there, try to understand it a bit. But, but starting with what's actually in front of you and impacting your business, I would say is the most critical.

• 40:50Speaker 03

  I, I'd like to kind of follow up on that a little bit. I think that's great guidance. And, um, and I've been thinking about it kind of in my own space as you've been talking through it. And, you know, one of the, the challenges, even if they're looking in, even if the, uh, the person who's receiving the guidance is kind of looking in their own walls, is there's, there's still, we're moving so fast, and you like, like, Agentic is the word of 2025, you know, I mean, it's, it, you know, it was building up in the latter part of last year, and it's full on now, and, and as people, but that's moving so fast right now. How do you get people to focus on that kind of useful thing, even if they're gonna stay within their own walls, and therefore they're, they're kind of limiting

• 41:35Speaker 03

  the scope of what they're addressing, to your point earlier? How do you get them to take it from that point of limiting scope to the point of, of, like, Finding points of productivity that are realistic and achievable, you know, within reasonable levels of time and resource on that. Because I see all the time people struggling with that, you know, even within limited scopes, on kind of figuring out how to make those choices and make it real. And that's, you know, I've, I think like for me, I've looked at Daniel is really, really good at that, at coming down. And, and I think as you're out there educating the world, like, how do you get people when you don't have a Daniel at a company?

• 42:20Speaker 03

  How do you get them to be able to focus on those different things to focus their resources on?

• 42:27Speaker 02

  Yeah, so again, always back to business challenges. Like you should be putting your resources where you actually have, you know, areas that you want to make a difference in, or and or investing in research and exploration. And in that sense, you don't need as many barriers and you can have a team. And the ones that are leading right now have already had, you know, best talent in research for the last years. They're not just starting now and trying to build things. They've invested time in exploration and in failure and in learning. And I think that's really critical. So you have that space. So again, it's not a rushed decision of like, suddenly I need to understand this thing fully right now because, you know, we've deployed it already. It's like we've made space and time to explore a concept fully and what it looks like actually when you build it out in practice,

• 43:12Speaker 02

  not just theoretically. So I think that's one thing. Another is the user experience. I was just walking through a wireframe with somebody who doesn't have a technical background at all. And they were asked to build out an agentic workflow, like multi-agent experience, you know, in the, in the financial services space. And they were having a really hard time with this. So they called me and were asking about it. And as we walked through the experience, I was like, you know, you can actually see very clearly what you do and you don't want. And as you reach each point of like questioning or experience, you can address the concept that's being used in that moment. And it becomes a lot easier when you break it up that way and you can relate it very tangibly to what it's doing like in, in practice. You know, so and what we're seeing actually with agents, when you build out the experience,

• 43:57Speaker 02

  people are like, oh, no, I don't, I don't want to do that. When it, when it has this, then I want to do this specific thing. It's like, okay, what you're describing is like automation and, you know, deterministic outcomes. That is not a fully autonomous multi-agent experience that you had in mind. And so you can actually come to that quite quickly and people can understand it a lot more simply when they see it in front of them in action. When you just have like a million words in front of you and you have no way to know what that actually looks like when it's deployed and in a product, then it's not going to make sense to you. And I don't know if that's an effective way to approach it. So I think having this user experience thinking, again, like the product mindset, when you're going through this will help a lot in grasping the technical terms.

• 44:38Speaker 01

  Yeah, I, I think every time I've actually asked people about what they want with their AI agent, it turns out what they want is either a rag chat bot or an automated workflow.

• 44:50Speaker 02

  Right.

• 44:51Speaker 01

  That's basically, basically every time. I'm sure there's other cases out there. But yeah, Allegra, I, I feel like we will definitely need to have a follow-up to the show to get more, more of your insights and continued insights from Lumiera. As we kind of draw to, to a close here, I do want to give you the chance to kind of look forward a little bit towards the future. We've obviously talked about a lot of challenges in terms of leadership and trust and implementing responsible AI practically. What, from your perspective, gets you excited about the, the future of how companies are adopting this technology or the possibilities of how they might adopt this technology?

• 45:36Speaker 02

  Yeah, I think a really good one just happens to be our company vision, which is a future equipped for humanity. We had started with humanity equipped for the future, but we want to be human centered here and actually shape technology for what we care about and what we actually want to maintain about the human experience, rather than having technology shape the, the ecosystem and our surroundings without us involved. So that's really where I see the future going is all of us being a lot more actively involved in the decisions we're making around AI, whether you're a leader or not. And then the other is that responsible AI just becomes what AI is. It's the standard. You're not thinking about it as an add-on at the end or something that feels like a hindrance or a barrier.

• 46:21Speaker 02

  It just is the standard when you're building. That's the future that I I hope for.

• 46:26Speaker 01

  That's awesome. It's a great way, great way to to end. And yeah, like I say, we'll definitely have to have you back because I feel like we could have talked for for a few more hours. But thank you for the work that you're doing. Thank you for the way that you're helping leaders in this space. And we'll, we'll definitely provide links in the show notes to what what Allegra's working on and some, some other talks. So make sure you check that out. And yeah, talk to you again soon, Allegra. It was great.

• 46:55Speaker 02

  Thank you so much.

• 47:03Speaker 00

  All right. That's our show for this week. If you haven't checked out our website, head to practicalai.fm and be sure to connect with us on LinkedIn, X, or Blue Sky. You'll see us posting insights related to the latest AI developments, and we would love for you to join the conversation. Thanks to our partner Prediction Guard for providing operational support for the show. Check them out at predictionguard.com. Also, thanks to Breakmaster Cylinder for the beats and to you for listening. That's all for now, but you'll hear from us again next week.