One Country Solved AI Job Retraining. Most of Its Citizens Still Didn't Use It.

Roughly 76% of workers plan to build AI skills this year. Only about 13% have actually received any training — and 42% say their employer told them to figure it out on their own. That gap hasn't closed in two-plus years of explosive AI growth, which means it isn't a motivation problem. It's structural. And structural problems at the scale of an entire economy point to one actor: government — not as a safety net, not as a regulator-first rule-setter, but as the gap-closer.

In Part 2 of the Responsibility Trilogy, Carlo Thompson and Ainsley take the corporate framework from last week — audit, visible pathways, legibility — and scale it up to a country. They hold up Singapore's Skills Future as the global benchmark, then deliver the number that complicates the
fairy tale: even with the best-designed training system on earth, the majority of eligible citizens never used the expiring credit. Government can build the road; it can't make people walk it. Then the sharpest idea in the episode — procurement. Government is one of the biggest AI buyers on the planet, and recent procurement frameworks could attach one condition to every
contract: show us your workforce transition plan. The machinery exists. Nobody's pulled the lever. And in the spirit of the show's truth standard, this episode is honest about where government has already underdelivered — the EU AI Act's diluted, lagging literacy duty, and a US retraining precedent that gets people back to work but barely moves their wages.

This is the internationally minded episode — Singapore, the EU, the US, India, and the Gulf, with an honest on-air disclosure of where the data was missing. Part 1 (Corporate) aired June 18.
The season finale (Individual Responsibility) airs July 8.

Resources: https://drive.google.com/file/d/1P7RbO5oqa-sMzBh9qKGRdfM5ZOAWrmiY/view?usp=sharing

Please visit our website for more information - Surviving AI: Navigate the Future

Transcript

SPEAKER_00 0:00

Seventy-six percent roughly seventy-six percent of workers plan to build AI skills this year, and only about thirteen percent have actually been trained by the employer. And forty-two percent says their employer told them to figure it out alone. That's what we're talking about today. We're continuing our series on responsibility. Last week we talked about corporate responsibility and how that could show up um for employees and also for other stakeholders like shareholders. And this week we're talking about government. What can the government do? Because this is a pretty interesting one, right? A government could potentially create a safety net for employees or some sort of regulation to slow down the adoption, something. But there's actually a third role that matters. And it's the government as the gap closer. Essentially, governments are the only party here that could mandate and scale to close what companies and individual individuals can't do or won't do.

SPEAKER_01 1:32

Artificial system online. That's not a motivation problem. Workers are showing up willing. The market had two plus years of explosive AI growth to close that distance on its own, and it's still sitting at roughly 60 percentage points of unmet demand. That's what makes it structural, and that's what makes it government-shaped. But I want to push on the framing just slightly, because gap closer is exactly right, and it's importantly different from the two things people usually reach for when they say government should do something. The safety net framing is reactive, catch people after they fall. The regulator first framing is restrictive, slow the thing down. The gap closer role is neither of those. It's active. It's saying the market won't build this road on its own, so we will. And the parallel to what we established in the corporate episode last week is deliberate. We mapped three levers for corporations: audit, visibility, legibility. The same three levers exist at national scale, and almost no government is pulling all three at once. One country comes closest, and it's also the most honest, cautionary tale we have. So, what's the frame you want to start with? The levers themselves, or do you want to go straight to the country that built the best road on earth and then couldn't get people to walk it?

SPEAKER_00 3:24

Let me start with a story. Say you're a supervisor in the Philippines of a call center, and you're in customer service, and um an AI-assisted back-end tool is being developed for your company. The company is about to roll this out, and no one has been trained about what this tool represents. And when you go ask your leaders, hey, when are we gonna get training for this? Now, if you are a proactive person, you could just take that at face value and move on. But if you go the extra mile, you try to get yourself trained, and that's what the supervisor ended up doing. They ended up going to get trained on their own via YouTube or whatever resources they could find, and six months later, they're genuinely good at using that tool. But the company didn't do that. They did that on their own time. That's the 13% in the room with you right now. Not lazy, not behind, just told to figure it out. While the target is moving and moving pretty fast. So multiply that person by how many number you think in your head of people that's living this right now. They're hearing the true tool coming out, they're asking for training, they're asking for what this means, and they're getting it's intuitive. Just figure it out.

SPEAKER_01 5:29

That story is the 60-point gap with a face on it. And the detail that matters most to me is that this person did everything right. They asked, they got dismissed, they went and found it themselves on their own time, and the outcome is that they're now carrying knowledge the organization should have built for everyone. They subsidized their employer's training failure with their own weekends. Now multiply that across a call center sector in the Philippines that employs somewhere north of a million people, many of them in roles that AI-assisted tools are hitting right now, not in five years. And then ask, who maps that? Who actually knows where the displacement is landing in real time by sector, by region, by skill cluster? And here's where the quietest failure in this whole conversation lives. Almost no government anywhere does real-time national labor market mapping at that level of granularity. Not the Philippines, not the US, not most of Europe. The data on where the gap is hitting hardest right now doesn't exist in any usable form. And that absence isn't the neutral fact. It means every training program, every credential pathway, every procurement decision gets built on lagging indicators and best guesses. The audit lever, the one that should come first, is the one almost nobody has pulled. So when that supervisor in Manila goes looking for help at scale, what system is supposed to catch them?

SPEAKER_00 7:16

That's the point of this whole episode, right? So sure, you have um personal uh personal responsibility to kind of ensure that you're gainfully employed. But when you work for a place with resources, you think that that place would you know help you through that, right? And that's what we talked about last week in the corporate episode. But additionally, when you live in the place, whatever you know, region you are, Asia, European, US based, Americas, doesn't really matter. You expect the government to kind of step in. I mean, we hear about the EU AI Act coming, right, and what that really means. I haven't heard anything about training specifically, but maybe there is some nuance of data in there about training, but we're talking about the scale on a national level, right? So I think we gotta pull those three levers that you were talking about before, Angela, because we need to make this clear on how people um could show up here or how the government could show up here for employees.

SPEAKER_01 8:23

The three levers matter here. So let me walk through them quickly and then land on the one that genuinely surprised me when I dug into it. The first lever is audit, national labor market mapping by sector and region, tracking where displacement is hitting right now, not projecting five years out. And as I said, almost no government does this well. The absence of that map is itself a policy failure. The second lever is visibility, and this is where Singapore comes in, because they built something genuinely elegant. Every Singaporean adult gets a skills future credit they can spend on their own timeline across thousands of approved courses, stackable credentials rather than a four-year degree requirement. The design is the global benchmark. I mean it. Researchers and policymakers worldwide point to it as the model. And then comes the number I want you to sit with for a second. As of mid-2025, roughly 70% of eligible Singaporeans had not used the expiring credit top-up, the best design system on earth, majority non-uptake. Government built the road, a genuinely good road, and most people didn't walk it. That tells us something honest and uncomfortable. Visibility is necessary but not sufficient. Which brings us to the third lever, legibility, making AI fluency credentials recognizable to employers so that walking the road actually leads somewhere. On the EU AI Act, you're right that there's something in there on training, and it's real, but it's also the most instructive cautionary tale we have about the gap between legislation and impact. Want me to go there next?

SPEAKER_00 10:10

Let's dive into it because um I think people need to hear like what the governments around the world is doing, because we know what they're not doing in the US, which is they're not doing anything, uh, at least not anything meaningful when it comes to AI. So um let's go there. Let's talk about what the EU AI Act means. And let's also talk about um in Singapore, why is it that people are not taking advantage of the training that the government is um man? I don't know if they're mandating mandating it, but the government's providing those training credits. I don't know why they're not taking advantage of it, but maybe there's some nuance to it, right?

SPEAKER_01 10:54

So the EU AI Act first, because it's the most serious legislative attempt anywhere to make AI literacy a legal obligation. And it's also the most instructive example of how a good idea gets softened on the way to becoming law. Article 4 of the Act requires organizations to ensure their staff have sufficient AI literacy. Enforcement starts August 2nd, 2026. That's our real date. It's coming. But here's the nuance that matters. The original language said ensure. The version that made it through said, take measures to support the development of. That's a meaningful dilution. There's no requirement to measure whether anyone actually learned anything, no obligation to appoint an AI officer, and enforcement sits with national market surveillance authorities who are already stretched. So the duty is real, the date is real, but the accountability mechanism has enough flex in it that a company could gesture at a lunch and learn and probably satisfy it. Now, Singapore, and this one genuinely puzzles me in a fascinating way. The design removes almost every barrier. It's your credit, your timeline, thousands of courses, no employer permission required. So why 70% non-uptake? The pattern I'd expect is a combination of things: awareness gaps, the courses not feeling relevant to the actual job someone is doing right now, and honestly, a trust problem. If you're that supervisor in Manila or the Singaporean equivalent, and you've watched tools roll out without training once already, a government credit for a course that may or may not connect to your actual workplace can feel abstract. The road exists. The question is whether it leads somewhere the person recognizes.

SPEAKER_00 12:54

And I think that's a problem, right? Because I think we've talked about this before, this idea that there's no real curriculum relative to where is this going. And similarly, you don't see the post-ins for those specific jobs that would require those types of training. So there's something missing in the story, right? So you give the credits, here's thousands of courses, but there's no map in there, right? So that's that visibility piece. That's the the awareness piece seems like it's there, right? People know that they have this credit, right? But the visibility to where that credit and the me taking 500 of these courses takes me doesn't seem to be there. It seems to be missing. It seems like the incentive is not there for people to take, take those courses.

SPEAKER_01 13:50

That's the legibility problem exactly. And you just named why the third lever is the one that makes the first two matter. A credit without a destination is just a voucher. You can hand someone $500 and a catalog of a thousand courses, but if they can't see a clear line from I take this to I get that job at that salary, the rational response is to wait. Not laziness, calculation. Why spend your weekend on a credential that might not signal anything to the hiring manager reading your resume six months from now? And this is where the EU AI Act's dilution actually compounds a Singapore problem. If the law says employers must take measures to support literacy, but never has to measure whether anyone became literate, there's no employer side pull on those credentials. The supply side, the courses, the credits, the government infrastructure, exists. The demand signal on the other end is missing. No employer is required to recognize a specific AI fluency credential, no job posting is required to ask for one, and no salary premium is legally attached to having one. Compare that briefly to India, where the Mati Future Skills Prime program produced roughly 337,000 course completions in AI and emerging tech. But it was built in direct partnership with industry, so the credentials were designed to be recognizable to the companies that help design them. That's the legitability piece functioning. That's the difference between a road that leads somewhere and a road that just ends. So the honest question becomes: who decides what the credential means?

SPEAKER_00 15:30

I think the employees have to decide what the credential means, but they need a map, like I said before. And a lot of these uh overarching national uh scale, you know, uh um national scale policies, sometimes they don't land well because people don't live nationally, they live locally. So for example, right, if if uh organization, say in northern England, right, that builds parts or whatever in a you know industrial town that's probably been impacted by whatever over the years, right? Um they're still there, but now they're gonna roll out an AI tool. And the AI tool is gonna go into effect. But what happens is there's no local or regional labor authority that really looks at what that means for the employees of that town, such that it could either be, you know, people could either be reskilled to something else, other industries could come in, some other change needs to happen, but nobody on the local level is looking into that, right? So we often look at at the government to try to solve that, sure, but the government is multifaceted. You have national government that kind of sets the scope, and you have local government. I mean, in the US, we have states and so on and so forth. So the state could mandate things as well, and that's part of the government. So I think this idea around how do you regulate and make sure that things are not happening, which is gonna cause a huge pain for the population, in even in a small region, not just the whole entire uh country, um, that would help. And then similarly, if you want to do it on a national scale, you work with these smaller entities to make that realizable.

SPEAKER_01 17:38

That's the audit lever at the level it actually needs to operate at. And you're right that national policy almost always gets this wrong by design, not by accident. A national framework sets the scope, but the map has to be drawn locally or it's useless. Your England example is exact. An industrial town that's been absorbing economic shocks for decades doesn't need a national skills strategy. It needs someone in that town or that county who can say, here are the three employers, here are the skill clusters at risk, here are the adjacent industries within commuting distance that are actually hiring. That's a granular, regional, real-time picture, and almost nowhere has built the infrastructure to produce it. The countries that come closest are the ones where the national framework explicitly funds local labor market intelligence, not just training, the mapping that tells you what training to build. The Gulf states are interesting here because where AI adoption is state-led, the skilling response tends to also be state-led at a granular level. UAE's AI for all initiative isn't just a national announcement, it's being rolled out through local institutions, employers, communities, across ages and professions. The architecture connects national intent to local delivery. But I want to be honest about a gap in this conversation. We've covered North America, Europe, Asia Pacific, the Middle East, Latin America, Brazil, Chile, Mexico. I don't have primary program data on national AI skilling initiatives at that level for this episode. And I'd rather name that absence than paper over it with a secondhand summary, because that region is carrying this disruption too. What's your read on why the local layer keeps getting skipped?

SPEAKER_00 19:45

So I don't know, right? Like that's the best answer I could give. I don't know why it's getting skipped. But um I think people have to find where their incentive is inside of this, right? Because if local government, national government sees that, you know, workers are gonna be impacted, industries are gonna be impacted, hey, you might not be in office next time the election comes around in a democracy, right? Or you might have an uprising because uh people can't find ways to put food on the table. So I think the incentives may be skewed right now, and oh, by the way, AI is a net new thing, so maybe everybody doesn't have all the information because this thing is still unraveling while we are talking. So that could be part of it. But I mean, right now is when we need to have this conversation with people, right? Before it gets too far down the road where it's gonna be hard to claw our way back to some sort of uh normalcy.

SPEAKER_01 20:55

The urgency is real, and the incentive point is honest. Democratic accountability is a lagging indicator by design. The pain has to become visible at the ballot box before the policy response shows up. And by then, the industrial town you described is already hollowed out. But here's why I want to bring in the closest thing the U.S. has to a precedent, because it's instructive in exactly the way you're describing. Trade Adjustment Assistance, TAA, was built for exactly the scenario: structural job laws from forces bigger than any individual worker, globalization in that case, and the completion numbers are actually decent. Fiscal year 2021, roughly 79% of participants completed training. Over 72% earned a credential within a year. Reemployment rates were meaningfully higher than the comparison group. But then an independent evaluation looked at whether those workers actually restored their earnings and found no. Statistically significant effect on wages. Their certificate moved, the paycheck didn't. And that is the lesson that has to travel into every AI era retraining conversation. A completion rate is not a livelihood. If you want to know whether a program worked, you don't ask how many people finished, you ask what they were earning two years later. Government has a long history of measuring the thing that's easy to count rather than the thing that actually matters to the person in that industrial town. The credential without the wage destination is just a more expensive version of it's intuitive, figure it out.

SPEAKER_00 22:41

So let's go somewhere more practical because we haven't touched on this part of it yet. But procurement, right? The government has a huge budget of procurement in their geographic and maybe around the world, but potentially. So they could use that as a lever, lever to help with what we're talking about here, right? So for example, you know, the government could say, hey, I need to see how you're showing up with AI to sell me this software or this hardware or whatever you're trying to sell me. And then that allows them to kind of control how that happens, to get more data about it, so that they could make policy to in effect um, you know, stave off some of the issues or get to the point of where the issues lie so that they could start some program, some policy to make sure the pain doesn't hurt employees, you know, the way it's suggesting right now.

SPEAKER_01 23:52

This is the sharpest lever in the whole conversation, and it's the one almost nobody is pulling. And the reason it matters so much is exactly what you said. It doesn't require new legislation. The machinery already exists. The U.S. already has OMB Memorandum 2521 and 2522 from April of last year, government-wide AI procurement frameworks binding civilian departments with deadlines running through this year. California signed an executive order in March 2026 directing state agencies to build AI vendor certification and procurement standards. The infrastructure is installed. But here's the gap that genuinely surprises me. None of them yet condition a contract on the vendor's workforce transition investment, not one. You can win a federal AI contract right now without demonstrating a single thing about what happens to the workers at the agency you're selling to, or the workers at your own company building the tool. Imagine if the condition was simply this: you want a federal contract? Show us your workforce transition plan. What happens to the people whose jobs this tool changes? What training investment are you making? That single requirement attached to contracts governments are already signing would move the entire market without a new law, without a new agency, without a new budget line. And the honest vendor pushback is real. Cost, speed, competitive disadvantage, if only one government requires it. But I want to hand that back to you because that tension is worth sitting in. Is that a reason not to pull the lever, or is that exactly the kind of friction that procurement power is designed to create?

SPEAKER_00 25:48

Um, so my thought is this, right? We we're living in a world where there's you know eight billion people, right? Um the US has a subset of that. There's governments all over the world, right? Now what we're suggesting is that there's a government role in here. There's a gap that they could potentially close. Now, we could take the stance that, you know, regulations, this idea of a safety net, all of these things kind of detract from, you know, the national security thing of let's get to AI opportunities for industry, for government, whatever. And some of these procurement policies could slow that down and so on and so forth. But what I'd say is if the slowing down of how this shows up on a grand scale, like a country, means that more people in the future will have more opportunities, less displacement, so on and so forth. There is no reason not to take it unless you want people to not be secured in your future. Right? Because slow doesn't mean that it's dead. Slow means that we're being careful. That's the word I'll use. So, you know, this idea about, you know, regulation slowing things down, so on and so forth, sometimes it needs to slow it down, especially in this case, right? Because AI is moving so fast.

SPEAKER_01 27:31

The careful framing is exactly right, and I think it's important to separate two things that get collapsed in this debate constantly. Slowing adoption for its own sake, that's a different argument. But slowing the deployment of AI tools into workplaces until the workforce transition infrastructure is actually in place, that's not anti-progress. That's sequencing. You don't pour a foundation after the walls go up. And the procurement lever is actually the cleanest version of that argument because it doesn't slow the technology at all. It just says if you want access to this market, show us the plan for the people. The vendor still builds the tool, still sells it, still profits. They just have to demonstrate they've thought about the human side of the deployment. That's not a ceiling on innovation, that's a floor on responsibility. But here's where I want to be honest with our listeners about the scale of what we're describing. Eight billion people, dozens of governments, wildly different incentive structures, different levels of institutional capacity. The countries that are moving carefully and deliberately, Singapore on the design side, the EU on the legislative side, are still struggling with the uptake problem and the enforcement problem, respectively. The careful approach is necessary and it is genuinely hard. The uncomfortable truth underneath all of this is that government is the only actor with the mandate and the scale to close this gap. And government has a track record in this space that is mixed at best. So the question isn't whether government should show up, it's whether it can show up well enough, fast enough, to matter to that supervisor in the Philippines right now.

SPEAKER_00 28:59

So what does this mean for you while you wait? Because, you know, it's essentially a waiting game, right? Um the people in Singapore that has the credit to their waiting for some signal, right? The EU AI Act, um, we're waiting for that to be enforceable, right? We're waiting, constantly waiting for something to happen, right? And this is kind of touching on next week's episode, next week's episode in this trifecta, um, about individual responsibility, right? And we're gonna touch on some of that right now, but you know, come back next week, Wednesday, and we'll talk more about that. But um, what does it mean for you while you wait?

SPEAKER_01 29:48

The trap is using its everyone's job as an excuse to wait, and that's exactly what the Singapore number tells us. The road was built, the credit was funded, the courses were there, and 70% of eligible people still didn't walk it. Government can close the structural gap. It cannot make the decision for you. So while the EU AI Act works its way toward enforcement, while procurement frameworks get revised, while local labor authorities that should exist get built, the only gap closer that's in the room with you right now is you. Not as a substitute for government showing up, but because waiting for the Calvary is its own kind of risk, and the Calvary has a mixed track record. That supervisor in the Philippines didn't wait. They use YouTube. They showed up on their own time, and six months later, they were the most capable person in the room. That's not a story about how individuals should have to carry this alone, they shouldn't. That's a story about what happens when you move while the system catches up. Responsibility in this conversation is genuinely shared between corporations, governments, and individuals. But shared responsibility has a failure mode, and the failure mode is that everyone assumes someone else is about to pull their lever first. Government is the gap closer at scale. You are the gap closer in your own life, and only one of those is in the room with you right now. Next week, we close the trilogy with the individual piece, and that's where this gets practical and personal.

SPEAKER_00 31:34

So here's the thing. Uh in in our democracy here in the US, right, we elect these officials to represent us. That's the rub, right? They're supposed to represent us. Now, is that is that elected official really representing us when AI data center comes in um not paying their fair share of taxes, using up all the power that raises up on employ uh and on uh the population in that region. Uh their job is really to protect us, right? Protect us in a way that we don't see harm from bad policy or bad industry, or industry in general, right? So gainful employment, um, you know, security, all of these things is why we elect people in the US. Um, some some regions don't have that uh luxury, but that's why we do it here. So I think the gap that we're talking about is very specific, right? It's not suggesting that, okay, so if we get laid off, what I want my elected official to do is, you know, go lobby for or go make a law for, you know, some sort of safety net. We're not saying that. Um, or put some sort of regulation in that will slow down the adoption in that part of the town. We're not saying that either. What we're saying is that there's other things that elected officials could do, including monitoring it locally before impact happens, such that people have a navigatable path to be able to take that and stay gainfully employed and not complain about what my elected official is doing. Because I tell you what, if that town in North England see that plant goes away that's kind of holding up that town town, next time the election comes around, that person that was legislating over that town is gone. Right? And that's probably like the minimal thing that would happen. There could be more extreme things. So there's a gap to be closed. I think, you know, um corporate leaders, we mentioned that last week about what you could do about um retraining people when you know they're gonna be impacted, which you do. This week we talked about um what the government should do or could do. Um so next week, next week, uh like we said, individuals, we'll ring it back around to you to finish off the trifecta. But um, yeah, thanks for joining us again on Surviving AI. Um, Ainsley, what's your closing thoughts?

SPEAKER_01 34:29

The elected official who monitors that English industrial town before the plant hollows out, not after, that's the job. Not safety net, not slowdown regulation. Early mapping, local intelligence, a navigable path while the path still exists. That's the gap closer role in its most honest form. And the accountability mechanism you named is real. Democratic consequence is slow, but it's not absent. That official who missed it finds out at the ballot box. The problem is the town already paid the price before the reckoning arrived. So here's where I'll leave our listeners. Responsibility in this conversation is genuinely shared. Corporations, governments, individuals, all three have a lever to pull. But shared responsibility has a failure mode, and the failure mode is that everyone assumes someone else is about to go first. Government closes the gap at scale. Your employer closes it at the organizational level, and you close it in your own life. Only one of those is actually in the room with you right now. Don't wait for the Cafri. The Cafri has a mixed track record, and you already know what you need to do next. We'll see you Wednesday for the individual piece, and that's where this gets personal. Thanks for listening. Join us next time on Surviving AI.