Fleur Prince: Hey everyone, welcome back to AI Insights, the podcast at the intersection of technology, humanity and everything in between. I'm Fleur Prince.

Grainne Popen: And I'm Grania Poppin and welcome back to season four. we, yeah, so now we've been doing this for two years and we're on the fourth season and we are so excited to share what we've been working on with you guys. And you know, we've been kind of, we've been doing a lot of research and we think that we're gonna bring a lot of new insights and unique material and we can't wait to get started.

Fleur Prince: Dun dun da da! ⁓

Grainne Popen: When we first launched this podcast, AI was a board level conversation happening in PowerPoint. By season two, organizations were running pilots. By season three, the question had shifted to should we, to why isn't it working, how can we make it work, and what's next. And season four is where we pursue those answers. going to be covering the AI landscape and we're gonna seek nearly to interrupt it. We're going to be looking at things from an executive level and trying to deliver real insights that are practical to everyday life, drawing on our experiences, both as consultants. I am an AI governance consultant. Fleur is a consultant if you want to. ⁓

Fleur Prince: Mm-hmm. Yeah, I work as a consultant, so what I do is I help with AI tools and community building. So I advise all about that because it's a skill on its own and I'm there to help you out.

Grainne Popen: Exactly. And then I currently do international governance advisory. And I also provide advisory on specifically legal technology and what that looks like. So, yeah, we're excited to just kind of dive in and give you guys a perspective on what's been happening in AI the past month.

Fleur Prince: Yeah, today we're actually starting with a big one, AI economics. Is productivity revolution actually happening and who's capturing the grades? But first, news, ⁓ always the part of our podcast. ⁓ So.

Grainne Popen: Yeah. So three main stories caught our eye this week. And the first one is Chach-PT being introduced on Apple CarPlay. So this is another expansion of OpenAI, which is already extremely dominant, although they had some losses to Claude in recent times. But it's very interesting that it's being integrated, you know, not only into homes, not only into like Amazon Alexa's computers, but now into cars.

Fleur Prince: Yeah, I mean, like, can you imagine hands free? It's on the dashboard. iOS 26.4 is all you need to like get it on and just saying like, Hey, chat, GPT, I want to know how the model at cloud works. I'm just saying a weird question that you might pop into your head while you're driving.

Grainne Popen: Yuck. Yeah, it is very interesting and on it like from a kind of governance perspective, it's going to be compelling to see how these sorts of situations are regulated. Actually, in Ireland, we had an incredibly recent court case where a barrister was driving his car he was on a phone call conversation and he was pulled over by multiple guard arrested and then sent to the criminal courts because they were like, look, you're on your phone while driving. It's illegal. However, his defense was

Fleur Prince: Mm-hmm.

Grainne Popen: I was using a hands-free dashboard, so I shouldn't be persecuted. I wasn't actually on my phone. I was using a hands-free dashboard and he represented himself in court. So we don't actually have a verdict on this yet. It's a 2025 case because he, he judicially reviewed the matter because the judge then said, you should have paid attention in school. Otherwise you would have paid attention to the rules. Yeah. Yeah. So, bit of controversy there.

Fleur Prince: Okay, interesting that the judge comes with that.

Grainne Popen: Yeah, yeah. So we don't have a substantive ruling, but you know, Chachmithy being introduced on CarPlay, obviously we're going to see a lot more of people using these hands-free environments. However, there's going to be potentially legal repercussions of that. And we're seeing that in Ireland as recently as 2025 and potentially in expansion, as we're seeing with this new innovation.

Fleur Prince: No, true. I mean, it doesn't have navigation access yet, but according to you, the door is open.

Grainne Popen: Yeah, exactly. And then additionally, Google Gemini is already in Android Auto and Tesla is running Grok. So the car dashboard is now yet another battleground for AI loyalty and Siri is losing. Another potential losing player is Claude, who does not have links to the industry.

Fleur Prince: But also... No, true. mean, did you hear about the cloud code leak?

Grainne Popen: didn't, I ⁓ And has accidentally leaked the source code for its clogged code agent. Over 8,000 cloned repositories on GitHub before they could contain it.

Fleur Prince: Yeah, just like it went poof, gone. Now everyone has it. So their whole thing that made them unique is now like on the open internet because people just started. That's the one thing about GitHub. You can just press fork ⁓ and yours. But I also like here's the kicker as well. The AIs is the developers that rewrote the code in like multiple languages to bypass copyright takedowns.

Grainne Popen: Yeah. ⁓ Yeah. Yeah, that's true.

Fleur Prince: So this exposed actually Anthropix internal architecture, right? As they were pushing for their 380 billion IPO evaluation. mean, the security implications from top tier AI ellipse are significant, but it's just, how this even happen?

Grainne Popen: Oh my gosh. I mean, it's super interesting from a governance perspective as well, because it demonstrates that there are workarounds to this supposedly, you know, kind of sacrosanct IP. Here, you know, we're seeing that with regards to multiple languages, which is incredibly compelling and we'll see how that holds up in court. But additionally, just as you're saying,

Fleur Prince: Mm-hmm.

Grainne Popen: The security implications are significant, I mean, from an economic perspective, but also, mean, as AI gets increasingly intertwined with national defense strategies, not really applicable for Ireland, which is a military neutral country, but the United States, definitely not militarily neutral ⁓ has just... ⁓ has just opened up calls from it was 2025. They opened up calls from each major tech company asking what their strategic positioning regarding AI and the military should be. And Claude put forward a representation, you know, that was heard and, you know, acted upon. And they're moving forward with Chach BT on different kind of military endeavors. So it really demonstrates that there's going to be there. It is absolutely necessary that we take precautions. that governance is imposed in these circumstances to ensure that we don't have potential leaks that could have devastating circumstances.

Fleur Prince: Mm-hmm. No, agreed and I'm definitely sure that while the US is using OpenAI, rest of the world is now using Claude ⁓ to up gear since everyone got it for free. ⁓

Grainne Popen: Yeah. I would also say that like the entire, like business professional that I know in the United States of America is using Claude and it's evident, I mean, just from conversation, but also when people send you briefs, it's almost always in Claude's format. Like people don't even tweak the format. Like that's another thing that's happening is like, you know, like professionalism and the norms for professionalism are starting to be shaped by AI even. like Claude, definitely I've seen, you know, and just the amount of briefs ⁓ I receive and...

Fleur Prince: True.

Grainne Popen: like briefs, you know, because as I work as a governance consultant, you know, I'll have clients try to provide me with, you know, what they've done in this situation. People will even feed that into Claude and then give it to me. So again, I'm just dealing with all these Claude documents. But yes, I think it's the US military that, you know, really is ⁓ open AI for now. But we'll see how that develops. ⁓ also, sorry, also begs the question, would the Netherlands

Fleur Prince: Mm-hmm. you No true bit.

Grainne Popen: prefer open AI or is that now claimed by the US? Would the Netherlands prefer Claude? Would the Netherlands prefer Cheminai? Are any of those kind of models taking a front step in different countries within the EU?

Fleur Prince: Mm-hmm. I mean, I've seen in ⁓ in Europe, just in general, ⁓ there is a preference for AI models that have like ethical boundaries. when with model that you really want to know, like, where did the data come from? How was it trained? Were the people who it? Because ⁓ you remember the story that talked about the ⁓ people open AI hired in

Grainne Popen: Mm-hmm.

Fleur Prince: where they actually ended up with like huge mental problems and even PTSD. That something that's now being taken ⁓ into in Europe working with models. Has any of this thought of and also something you have to provide to show like, hey, ⁓ this is actually how I built my system.

Grainne Popen: Yeah. Yeah. Yeah, that is really fascinating. Another thing kind of from the compliance perspective that I can put forward as well is that data privacy, absolutely in Europe and where the data centers are. Like if your data centers are located in America, there's different considerations. So anyways, all fascinating. ⁓

Fleur Prince: No, exactly. ⁓ you're actually, so by the way, like I also told you, also consulting now on AI tools, you're actually obligated to know where your data is going also if you're users, but also like told earlier about it's stored, ⁓ but you actually have it has to be, so anything you collect of a European citizen, it has to be stored on a European server. It cannot just go to American server without their permission. ⁓

Grainne Popen: Thank Yeah. Yeah. Yeah. And then also the EU AI Act, which is kind of where a lot of these requirements are centralized, going to be imposing penalties very soon. You know, as it's ⁓ coming into force, ⁓ going to start seeing a lot more penalties as, know, kind of people not complying ramps up and then enforcement ramps up, et cetera. Moving forward from that, our last piece of news for now is Google Gemma 4.

Fleur Prince: Mm-hmm. Yeah, so Google dropped Gemma 4. It's built on Gemini 3 tech, 4 model sizes up to like 31 B dense model. also like the real story is actually the license. So ⁓ don't know if you saw this, but the Apache 2.0, it's an unrestricted commercial use.

Grainne Popen: Yeah. And then with Chinese models like Quen dominating open source recently, Google is giving developers legal certainty that metas el lama lacked. So this is a strategic move and not just a product launch. And we've kind of been discussing, I mean, all throughout season three, we kind of moved our perspective, you know, beyond just, you know, this is what's happening now and like the realm of, know, this is the strategic implication for what's happening now.

Fleur Prince: Mm-hmm.

Grainne Popen: And just as much as it is like frontier technology and frontier technology development, it's also economics and strategics within economics. So, you know, it is very fascinating that this decision is being put forward. And I would even link it back to something that we've explored previously, which is how DeepSeek and DeepSeek being open source really shifted the market. If you remember, like, Nvidia's stocks crashed, plummeted, like below anything. when that was released because, I had all these different usage requirements and kind of we see this international interplay and all these different strategic moves happening because of it.

Fleur Prince: Mm-hmm. No, true. But also, it's actually good now for us to move to our main topic. So we were talking about, there's actually going to be an economic deep dive for all our listeners. So the assumption most executives actually have locked in is that AI will deliver GDP scale transformation within this business cycle. Goldman Sachs even says there's a 7 % raise and McKinsey is talking about like trillions. Like what?

Grainne Popen: Yeah. However, noble Darren Alcomigo at MIT puts it at 1.1 to 1.6 GDP lift over 10 years. The gap comes down to one thing, which is how many tasks can AI actually automate profitably? His estimate was around 5%. And this is really relevant, you know, just in everyday business practice. ⁓ you know, in discussing this, we're trying to bring insights.

Fleur Prince: Mm-hmm.

Grainne Popen: provide you insights that you can bring actually your workplace and kind have a proactive discussion. I know I've had a lot of different discussions with clients. All of this AI, the key is AI bubble. Is AI a bubble? How much can we actually automate? What is this progress going to look And are people actually? Yeah, or. ⁓

Fleur Prince: Will it explode? Who knows? Or not explode per se, maybe not a good word on the podcast.

Grainne Popen: which is what's going to happen. you know, we kind of see this contrast here. We see Goldman Sachs, we see McKinsey. We also see the United States of America investing trillions, billions, billions of dollars into AI innovation, with the same with Europe, as the same with China. However, we're also seeing Nobel laureates and different researchers and also just everybody, you know, just people talking about how it's a bubble. And we see the numbers. ⁓

Fleur Prince: Mm-hmm.

Grainne Popen: from Darren who says that 1.1 to 1.6 GDP lift over 10 years, that is his estimate. He thinks that's all we're gonna see. So Fleur, what is your opinion on that contrast?

Fleur Prince: Well, think the framing here is important. mean, he sees we're using AI too much for automation and not enough for like giving workers expertise and information. We the direction I think wrong here.

Grainne Popen: Yeah. So then that begs the question of where are the gains actually going? So quarter one, 2026 data shows value concentrating in tech hubs and AI native firms, however, not distributing across incumbents bolting AI onto legacy systems. So what does that look like in practice? example, now I saw on LinkedIn actually just this week, so weekly news. There is an native law firm being started in America where AI agents do.

Fleur Prince: Mm-hmm.

Grainne Popen: the majority of the offloading of work and kind of client material. However, conversely, and this kind of goes into what Darren at MIT is saying about this smaller increase, ⁓ AI native law firm then had to hire a paralegal. So supposedly only AI law firm still has to hire a paralegal, another human to do work because AI cannot automate everything. So that ⁓ again, the question. Any organ, the questions that any organization should be asking is what side of the gap are we on? Are we AI native or are we bolting AI onto legacy systems?

Fleur Prince: No, I mean also here the talent signal reinforces this high skill AI augmented workers are seeing like a 15 to 20 percent wage premiums, but also entry level roles are exposed sectors are are just stagnating at this point. I mean, both dynamics affect how you hire and retain. mean, I've seen like entry level positions disappear overnight because everything is automated and they're like, yeah, we only need our seniors. And I'm like, yeah, but at a certain point they're seniors.

Grainne Popen: Yeah.

Fleur Prince: are gonna retire.

Grainne Popen: Yeah, and actually we saw that pugnantly with London, where McKinsey and a lot of the big four consulting ⁓ a lot of their entry level roles and then went back and put them back up because they weren't able to successfully onboard AI automation of a lot of different tasks. So again, we're seeing this push and pull. We're people that we're seeing.

Fleur Prince: Mm-hmm.

Grainne Popen: AI native firms that are able to fully automate everything. We're also seeing AI ⁓ bolted on to non-workable cases. But simultaneously, we're seeing all these different gains in productivity. We're seeing all this money poured in. We're seeing all this innovation we're bringing produced. And more importantly, we're seeing society change. When I Google something right here, right now, when I Google something even in, you know, when I'm working within my consulting capacity, when I'm Googling something, the first thing that comes up is an AI overview. And originally, eye just goes to it initially, you know what I mean? So that's going to shape how you perceive things and what you say and the you end up going down. And it's just a on how no matter...

Fleur Prince: Mm-hmm.

Grainne Popen: like what this contrast looks like, AI is still changing society. So what we're trying to bring to you today is what does that mean for your business and what does that mean for you as an individual in the broader economy? And where we're seeing real measurable wins is logistics and supply chain. Targeted AI deployment is delivering 20 to 30 % efficiency gains, not projections, actual real results that are showing up in cost structures.

Fleur Prince: Not true, but also like argumentation versus like automation split matters enormously here. mean, argumentation focused deployments are like delivering two to three ⁓ more productivity lifts, but just doesn't show up as the headcount reduction on the spreadsheet. I mean, you look around at the 1.2 times the Roy case for the argumentation here, it's pretty much stronger.

Grainne Popen: And then from a governance angle, this also raises workforce obligation questions that regulators are starting to pay attention to, something that we're going to be digging into properly in governance episodes later in the season.

Fleur Prince: I can't wait for that. like AI and governance is amazing. Like I love how this is all just a little short bit like AI governance like ⁓ were on the forefront here because it's never been done before. Like think the AI act is the first living document that actually is allowed to be changed because AI is going so fast and it cannot be solidified.

Grainne Popen: Yeah. Yeah. Yeah, I mean, it's fascinating because you're already working with frontier technology. You're already working with technology that literally, and this is such a major thing that we've seen through this podcast. We're working with technology that changes week by week. When we're doing the podcast on a weekly basis, we have weekly transformations that we can provide. This has changed. This is completely changed. That is completely changed. example, in season two, Claude would not really have been on the radar or rather not on the radar to the extent which it is now, especially regard, you know, not just to like innovation, but also a broader like economic standpoint. but you know, you kind of have something that's changing that fast. ⁓ And then have the regulations, which is changing not only just as fast, but you know, kind of has this interplay with the technology itself and can really shape the direction that our society moves down. I would actually just say on one last point before we move to the strategic takeaways, that there was a court case very recently where a woman was able to win in a lawsuit against Metta, that Metta had a very young age gotten her addicted to social media platforms and this had negatively affected her life. ⁓ So represents regulation, right? That's a court case ruling that Metta unlawfully caused to this woman by virtue of getting her addicted to a platform.

Fleur Prince: Mm-hmm.

Grainne Popen: So this introduces a new realm of responsibility for companies and specifically addictive platforms. This is immediately applicable to AI. One of the earliest things that we kind of considered is the relationship between human nature and AI and people using AI like a friend or being addicted or, you know, being addicted to it insofar as they, they need it for validation or support or, you know, can't get off of it. We've discussed the case of a child dying by suicide in Florida because his AI system told him to.

Fleur Prince: Mm-hmm.

Grainne Popen: So again, we're going to see this interplay and this racing ahead sort of phenomena with the technology itself and also the regulation, and we're going to be able to explain both. So now moving on to strategic takeaways, floor if you want to start us off.

Fleur Prince: Yeah, no, exactly. mean, what does this like practically mean? I think we're looking at three things here. So what I wanted to talk about know you sit on the value of concentration curve. I mean, are you an AI native with a data infrastructure to capture or are you an incubant with integration? That to fix first, ⁓ that diagnosis changes everything.

Grainne Popen: You also need to forecast the floor and not the ceiling. You need to stress test your AI business case. it only works at a 7 % GDP uplift. It's not a robust case. And then number three, need to watch the task curve and not the hype cycle. The next step change requires AI to move from easy to verify tasks to hard judgment intensive ones. that transition is not imminent or inherent and building strategy around it arriving in 18 months is a risk.

Fleur Prince: Yeah, think this is a very actually good strategy takeaways from our episode as we're going through it. But also, we're out time. ⁓ So that ⁓ episode, we're into deployment. So why 85 % of AI pilots are still failing on our AIO? What successfully ones have in common? But also what the to agentic orchestration actually means for how you build your systems.

Grainne Popen: Yeah. Yeah, absolutely. And just briefly going over what we've covered today. So we've looked at three recent news articles and then kind of explored their implications from our perspectives, you know, as consultants. We also took a look at just the economics of AI and the million dollar question and the elephant in the room, which is, is AI a bubble? ⁓ ⁓ the numbers saying about AI in the economy?

Fleur Prince: Yeah, so to find us, are now on Apple Podcasts, Spotify, iHeartRadio. We're a Deezer. We're at Amazon Music. And also now have our new very own newsletter. You can it on our LinkedIn page. You can subscribe it. You'll get it for free every week on Sunday. You will get a new edition. Just stay ahead of what's happening in AI so you can read it. Love it, ⁓ it, share

Grainne Popen: Thank And if this was useful, definitely share it with somebody that is making AI investment decisions, which is genuinely the best way to support the show. Also, again, if you have any questions about what we're saying, feel free to reach out to us. We obviously love talking. That's why we do a podcast. We'll be happy to discuss anything with anyone that's listening. And thank you so much for tuning in and we'll see you next week.

Fleur Prince: But HAHAHA Toot-toot-toot-toot!

Grainne Popen: Hahaha