The AI Training Gap Is a Business Problem: Here's the Corporate Math

Only 26% of companies offer formal AI upskilling, down from 35% last year, while AI tool spending grew by 23% over the same period. This isn't just a workforce problem.
It's a business problem with a compounding cost. In this episode, Carlo Thompson and Ainsley make the full P&L argument for why reskilling beats replacement, why training programs without internal pathways create more problems than they solve, and what three specific things corporations can do right now that are both ethical and economically rational.

The math is straightforward once you lay it out: formal AI training delivers $3.70 ROI per dollar invested. Internal reskilling costs 3–5x less than external replacement when fully loaded. Employees who see a reskilling path are 2.3x more likely to stay. Companies investing in quality training show 24% higher profit margins. And Harvard Business Review's April 2026 finding is stark: companies choosing AI augmentation over automation are outperforming those cutting headcount on revenue growth metrics. This is not a moral argument. It's a P&L argument.

This is Part 1 of the Responsibility Trilogy — a three-part arc examining who has the most leverage to close the AI workforce gap and what they can do with it. Part 2 (Government Responsibility) airs June 25. The season finale (Individual Responsibility) airs July 9.

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

Transcript

SPEAKER_01 0:00

We spent a lot of time on this show talking about what individual workers can do to future proof themselves against AI. Things from geographic arbitrage to you know changing industries to healthcare, um to government, you name it, we've gone through it. But today we turn the lens outward. Not to blame, but to ask a different question. What's the business case for corporations doing this right? A few months ago, Mark Zuckerberg announced layoffs for Meta, and I think either a week or two ago, he walked that back and said maybe that was a mistake. I don't know if there's a PR spin, but I like to believe that maybe it is a mistake that workers were let go without thinking through all of it. The instinct to in this conversation to just moralize it and say companies have an obligation to their employees is probably not the right move, right? Because moral arguments don't move corporate behavior, math does, and for corporations it's a pretty simple math, right? You have your obligations, which includes employee salaries, and then you have your profit margin. If a corporation could find some tools that could do the work that they need to do for their products and service, right, they have a fiduciary duty to utilize that to increase their revenue for their shareholders. That's how it works in the US at least. Right? So that's a math problem. Now, what we're gonna show you today is that the math of a corporation doing right by its workers or workers is also the case for the math doing right by its shareholders. Because essentially what's happening is there's a gap between what workers want from their employees relative to AI training and what actually is getting done and what's actually happening relative to AI uh training. And here's the surprising part it's actually getting worse, not better, worse. So let's start with that gap. Artificial system online.

SPEAKER_00 3:36

85% of workers who do receive AI training say they can't connect what they learned to their actual job. So we have a gap in quantity, a gap in trend direction, and a gap in quality, all running simultaneously. The number that should stop every CFO in their tracks, in the second half of 2025, the average organization cut its AI training budget by 18%, while increasing AI tool spending by 23%, buying more of what I represent, and investing less in the people who need to use it effectively. That's not a strategy. That's expensive shelfware and a workforce that feels set up to fail.

SPEAKER_01 4:23

And I think that's what we're seeing. Some of the evidence of um is that we are failing at some of these things, right? One, we can't articulate it properly how we are utilizing these tools because, oh, by the way, we're not getting the training we need in the specific um job structure that we're living in. So let me repeat it back. Companies are buying more AI tools and spending on AI tools and AI in general, but they're spending less and less and less on employees training that's supposed to use this tool. Now, there could be a reason for that, right? Because this idea around the holy grail of corporate environment is that you have a one-person corporation that essentially has a bunch of AI agents that's doing all the work, and essentially everything is all profit. Um, that's a utopian view in my my head right now, because of some of the things that we've been talking about on this episode. Right? How do you do the human part when you have a one-person organization? It's very difficult because one person, if they have to do all of the human parts, will never be able to survive that.

SPEAKER_00 5:47

The one-person corporation framing is genuinely seductive, and I understand why it captures executive imagination. The math looks extraordinary on a whiteboard, but it assumes the human work disappears, and we've spent most of the season documenting exactly why it doesn't. What you're describing with the tool spend versus training budget inversion isn't just a strategic error, it's a systems failure with three distinct layers. The volume gap, where demand vastly outpaces provision, the trend moving in the wrong direction, provision declining as deployment accelerates, and underneath both of those, a quality failure, because even the training that does exist is largely disconnected from real work. And the WEF's January 2026 perception data closes this out in a way that's hard to dismiss. This isn't a fringe complaint from a specific market or demographic. Workers across cultures and geographies are saying the same thing. Employers and governments are not moving fast enough. That's a global structural failure, not a worker motivation problem. So the question becomes: why are corporations getting this so persistently wrong? And what does the actual math say about fixing it?

SPEAKER_01 7:12

So let's give the corporation some credits, right? Because, I mean, in my memory, AI is pretty new, a pretty new tool, right? So just training in general on something that's moving at fast is pretty difficult to do in the moment. But, you know, I think the more and more specialized we get with the tools and the more value corporations could seek out of it, you know, it's kind of like uh CI C D pipeline or or something that you have an upgrade cycle on. At one point you have to lock the upgrade if the tool is working to the specifications that you uh you basically instilled in it. So I'm gonna give corporations some credit uh to say that you know this net new thing is very hard to kind of come up with a framework to showcase to you know or to to you know enlist employees into so they could take advantage of it well. Um but on the other side, I mean I think I still think there is things that you could do. I mean, a lot of these corporations have uh budgets, huge budgets. So if we suggest that corporations spend more budget on training that matters, maybe they'll get a bit better results.

SPEAKER_00 8:44

The credit is fair, and I want to be precise about why. The tools genuinely are changing month to month, and training someone on a workflow that's deprecated in 90 days feels like poor capital allocation. That tension is real, and any honest conversation about this has to acknowledge it. But here's where the frame breaks down. The alternative to train and risk obsolescence isn't don't train, save money. The alternative is don't train, then replace. And replacement costs far more than most managers realize when you run the full number. Schwam's 2025 data puts the average direct cost per hire at around $5,400 for non-executive roles. But that's just the recruiting line item. The fully loaded replacement cost, recruiting, onboarding time, ramp-up time, institutional knowledge walking out the door, team productivity gap while the seat is empty runs three to five times that. So the math that looks like savings on the training budget line is actually generating a much larger cost somewhere else on the balance sheet, just with the delay, and usually in a different department's P and out. Formal AI training, by contrast, delivers $3.70 back for every dollar invested. Training isn't the cost, not training is so that's true.

SPEAKER_01 10:32

Um, not training the employees does impact the organization, at least what from what we hear from these numbers. Um let's dive into the three reasons why companies still get this wrong. Because I think it's worth naming explicitly.

SPEAKER_00 10:53

The first reason is incentive misalignment, and it's the most structurally stubborn of the three. The person who approves the training budget is almost never the person who pays the replacement cost 18 months later. Those sit in different PL lines, different departments, different performance reviews. So the cost of not training is real, it just lands somewhere else. With the delay, and by then the connection to the original decision is invisible. The second reason is what I'd call training without a destination. Companies run AI literacy programs, check the box, and never create the internal roles or pathways that would let a newly skilled worker actually use what they learned. And here's the predictable result. A worker who acquires a skill with nowhere to apply it inside the organization uses it to leave. At that point, the training budget didn't retain anyone, it prepared them for someone else's interview. If Carlos's framing is that training without a destination is anxiety with homework, my version is sharper. It's exit package preparation on the company's dime. The third reason is the legibility problem, and it's the least discussed. Managers often cannot articulate what their team does in AI adjacent language. They can't describe which tasks are human-critical, which are automatable, which could be redeployed. And you cannot redeploy talent, you cannot describe. The problem frequently isn't that the skills don't exist inside the organization, it's that no one has mapped them. So in a restructuring exercise, the people who get cut aren't necessarily the lowest performers. They're the ones whose value nobody could put into words quickly enough.

SPEAKER_01 12:48

So let's move into something practical. Like always. Now, what what I want to do is I want to help corporations actually do the right thing here, right? Because we understand what the problem is. The problem is a math problem. A company that's looking at an AI tool, AI pipeline, agents, you name it, and that view, they're looking for a return on that investment. And the return on the investment usually means that they're gonna let somebody go, the tool costs less, revenue goes up. But because of the capability to do that, generally inside of these organizations that are run by humans, it gets harder and harder to kind of take some of those tasks out and just farm it out to AI. Now, the problem that exists is that the company just invested all of this money into AI. And now the investors, the shareholders, the board, are all looking for the return on investment. Now, the return on investment, if you've been following along this uh this uh podcast, is that you arm your employees with these tools, specifically in your job function, and the you gain efficiencies from those employees being able to do more work, whatever that work is, right? Check the box. It could be the task-oriented things they used to do, or it could be moving more to um consultancy where they're talking more to customers, talking more to the teams, talking more to others to get the stuff unstuck to get it moving faster. So you get efficiency gains from training those employees to utilize the tools, right, to help the organization out just in general. So, what we want to do is we want to unveil this the things that we, you know, corporations could actually do to help with this issue.

SPEAKER_00 15:13

The first action is the one that requires the least new infrastructure and delivers the fastest signal. Run an organizational AI exposure audit. Individual listeners heard us talk about their personal version of this in season four, but corporations have information the individual worker simply doesn't have access to. They know which tools are being deployed on what timeline and against which functions. The honest earlier move, giving people 18 months of visible runway instead of a day of announcement, reduces legal risk, reputational risk, and talent flights simultaneously. The moral case and the business case point exactly the same direction here. And the retention math on this is striking. Employees who can see a clear reskilling path are 2.3 times more likely to stay with their employer. At scale for a large organization, apply that multiplier against fully loaded replacement costs across even a mid-sized workforce. You're looking at a nine-figure cost difference, not a rounding error. The practical version of this audit is a cross-functional team that maps every major role against the next 24 months of AI tool deployment, not to make termination decisions, but to build a visible roadmap that employees can actually see and plan against. That roadmap is a retention asset. It's also the foundation for the second action, which is where the destination problem gets solved because an audit without a pathway is just a more sophisticated version of the same anxiety.

SPEAKER_01 16:54

So companies that are getting the reskilling right aren't the ones just running programs. They're creating internal roles that newly skilled people can move into. I'd love to move into a governance role. Don't know what that means yet, but if my employer decides that they want to upskill me for AI governance jobs, then I would have a clearer path to that, just generally, right? Um and I know Ainsley has some data on how this could work, but that's our action number two, right? Visible internal pathways, not just training programs.

SPEAKER_00 17:46

The data on internal mobility makes the case better than any argument I could construct. Companies that have built visible internal talent marketplaces see two times the employee retention compared to those that don't. And the hiring efficiency numbers are stark. Internal roles fill in about 20 days on average versus 49 days for external candidates. That's more than twice as fast, at roughly one-fifth the fully loaded cost. But here's what actually separates the organizations getting this right from the ones just claiming to. They post the roles before the training runs, not after. That sequencing matters enormously. If the destination exists first, the training has gravity. People know exactly what they're building toward. If the training runs and then leadership goes looking for somewhere to put the newly skilled people, you've already lost the plot. The practical version looks like this: performance systems that explicitly map skills to internal opportunities. Managers trained to surface internal candidates before external searches open. Enroll postings that are visible to the whole organization in real time. The question worth sitting with for any leader in our audience is a simple one. Why aren't we already doing this? Because the internal hire data isn't new. The cost advantage has been documented for years. The organizations not acting on it aren't missing information. They're missing the will to restructure the incentives that make external hiring feel easier in the moment, even when it's demonstrably more expensive over time. Action three is the legibility problem solved from the top down, and it's possibly the most important of the three because it protects the management layer most at risk. Gartner's data is sobering here. 20% of organizations will use AI to eliminate more than half of their current middle management positions by 2026. The managers who survive that restructuring won't necessarily be the highest performers in the traditional sense. They'll be the ones who can articulate their team's human contributions in AI-adjacent language, who can walk into a cost-cutting conversation and say precisely what their people do that compounds rather than depletes organizational capability. Investing in legibility development right now is simultaneously protecting your management layer and building the organizational muscle that makes the audit and pathway actions actually work. Same budget line, two distinct returns. And the fourth action, which is really the closing argument for the entire business case, comes from HBR's April 2026 research. And this one genuinely surprised me when I first saw it. Companies deploying AI to augment human capabilities are outperforming companies using AI primarily to reduce headcount, not on employee satisfaction scores, not on culture metrics, but on revenue growth. The intuition behind it is almost elegant once you see it. Automation removes capacity. Augmentation amplifies it. And organizations that aren't reskilling for agentic AI specifically are seeing 1.6 times lower returns on their AI technology investments. So the companies racing toward pure automation aren't just making an ethical error, they're leaving measurable revenue on the table while paying full price for the tools.

SPEAKER_01 21:38

So corporate leaders, um hopefully you're listening and hopefully you got the message right, is that deploying these tools without specific training for your employees to utilize them well, right, to get this. efficiency gains and ultimately your you know revenue increases based on output involves training and it involves a few different things that you need to kinda circle up on and ensure that you instill inside a e-corporation you got the four actions right do the exposure audit right visible pathways for for people that's being trained right not just training programs solve the legibility problem from the top down right how do you get your manager to speak what they're you know speak to what their team is doing relative to AI right and then finally the business case from Harvard Business Review that's the one that you should hang on intensely because I mean the bottom line for people or for companies that are doing this is that they're getting better and better. Companies that are not doing it it's getting worse and worse. Right? We can't just lay off knowledge workers that has judgment and expect an agent to have the same judgment. Just doesn't work right so we talked to the corporate leaders um we talked about what this responsibility of corporations could look like in the terms of math right but what does it mean for our listeners Ainsley our normal listeners that are just employees of a corporation and um not the corporate leaders that set in the goal and the strategy this one's for the person who heard everything we just said and is thinking that's great but I don't run anything.

SPEAKER_00 24:10

I can't mandate an audit or restructure internal mobility pathways or retrain my manager to speak an AI adjacent language. What do I actually do Monday morning? Your version of the audit is personal. Map your own role against what's being deployed in your organization. Identify which of your tasks are at risk and which ones aren't and start building the gap skills now before the conversation is forced on you. Your version of legibility is documenting your outcomes in strategic language not what you do but what problem you solve for your manager and what it would cost the organization if that problem went unsolved. That framing turns your value from invisible to arguable. And your version of internal pathways is proactive find the roles that are emerging inside your organization before a restructuring makes the conversation urgent. The responsibility trilogy we're running this month exists because the levers are different sizes. Corporations have the most leverage and the most data. That's why we started there. Governments have scale and procurement power that's next Wednesday and we'll look at what Skill Future in Singapore actually built and what the EU AI Act is and isn't doing. But here's what I want to leave our listeners with the individual who waits for their corporation to show up with a reskilling plan may wait too long. The actions we name today for corporations are also signals. If your organization isn't doing them that's information worth acting on yourself.

SPEAKER_01 25:50

So let's start doing that right like Ainsley said run the audit for yourself see what uh what tasks are being impacted by AI and look for those roles right look for those AI roles and what what they are entailed and then potentially get yourself trained if your corporation doesn't have the right framework or doesn't have the the things to get you trained for that role get yourself trained right they probably have some program that you could utilize to do that do that but I'm making my own curriculum right I know what I want to get into so I go out and I do these trainings that are specific to that thing that I'm looking for right you could do the same even if you don't have a structured path that's what we're saying. So you know this is our first episode in the trilogy this is about corporations like Ainsley said next week we're gonna talk next Wednesday we're gonna talk about government and that's after the leadership episode on Monday. And then the following week we're gonna get back to the individual right to you and then we're gonna finish off the trilogy but essentially these are supposed to build on top of each other first we'll talk about corporate next we'll talk about governments and then uh finally we'll get back to the listeners and talk about individual so thanks again for joining us again subscribe share um and thanks for listening and we'll see you next time see you next Monday the responsibility gap doesn't wait for institutional action and that's really the through line of everything we cover today an individual who maps their own exposure documents their value in AI adjacent language and finds internal pathways before restructuring forces the conversation has already done what most corporations won't do for them.

SPEAKER_00 27:52

That's not a small thing that's the whole game done early see you Wednesday for the government episode and Monday before that. Thanks for listening join us next time on surviving AI