A Conversation about the Price of AI Fluency

[00:00:00] I want you to imagine a scenario for a second here. You've just spent the last six months completely overhauling how you do your job. Okay. Setting the stage. Yeah. Like, let's say you're a senior supply chain analyst. You've mastered this whole suite of new generative AI tools. And you didn't just, you know, learn how to chat with a bot. Right. You've gone deep. Exactly. You've built custom prompt libraries. You've automated all those tedious data cleaning parts of your weekly reporting workflow. And frankly, you're producing vastly superior predictive models now.

[00:00:29] I'm probably doing it way faster, too. So much faster. You're basically feeling like an absolute powerhouse, right? You're fundamentally more valuable to your company. But then your annual review comes around. The dreaded annual review. Yeah. You sit down with your manager, you lay out all these incredible new efficiencies you've brought to the department, and your paycheck doesn't budge a single cent. Oof. You get the standard, what, 3% cost of living adjustment and a nice pat on the back. So why? Like, why does the corporate system...

[00:00:59] Why does the corporate system seem completely blind to this massive, undeniable leap in your capabilities? It's, I mean, it's an incredibly frustrating reality for so many people right now. And it really points to something far deeper than just, you know, a single stingy manager. Right. It's systemic. Totally. What you're describing is the absolute definition of a structural breakdown. We have this collective cultural assumption that compensation is just this pure frictionless reflection of value.

[00:01:26] Like you generate more value. The invisible hand of the market rewards you. Exactly. But right now, inside the walls of the modern enterprise, there's a massive mechanical disconnect between the value an employee creates with generative AI and the actual, like, plumbing of the human resources systems designed to measure it. And that plumbing is completely backed up right now. Oh, it's a mess.

[00:01:47] Which is why we're looking at this truly groundbreaking article today. It's from November 2026, published in the Human Capital Leadership Review by Dr. Jonathan H. Westover. Such a crucial piece of research. It really is. It's called The Price of Fluency, Why Pay Systems Are Falling Behind AI Changed Work. And it uncovers how companies are actively struggling to put a price tag on AI skills. And the disastrous ways they're improvising, too.

[00:02:13] Yes. From bizarre metric gaming to accidentally widening demographic pay gaps. And then, thankfully, it gives a blueprint for how forward-thinking leaders can actually rebuild their systems. Which we desperately need. Okay. Let's unpack this. Because whether you are an employee trying to prove your new AI worth to your boss, or a manager staring at a spreadsheet trying to figure out your budget, this structural mismatch affects you right now. Yeah. The broader context here is just so essential.

[00:02:41] Because we're watching the architecture of the modern workplace just, well, crack under the pressure of a technology that moves fundamentally differently than previous shifts. It's not just like moving from filing cabinets to Excel, right? No. Not at all. We're talking about a breakdown in how the corporate world legally and financially even defines what a job is. The speed of this change is what really stands out in Dr. Westover's data for me.

[00:03:05] When you look at traditional comp systems, it feels like trying to measure a rocket's speed with a sundial. That is a very accurate analogy. The tools are just entirely inappropriate for the object being measured. Like, we have pay scale data from November 2026 that just paints this bleak picture for HR departments. Yeah. They're really struggling. Half of all employers surveyed straight up admitted that their salary structures are struggling to keep pace.

[00:03:31] And nearly half said their market benchmarks, you know, the external data they use to justify pay, no longer reflect the skills their roles actually require. Right. The benchmarks are totally obsolete. So as a result, 61% of these organizations are actively scrambling to rewrite their job descriptions right now. And that scrambling makes perfect sense when you look at the velocity of the mutation happening inside these roles. I mean, to put some hard numbers to it, PwC did this massive analysis back in 2025.

[00:04:00] Yeah. This wasn't just a small poll, right? No. They didn't just poll a few executives. They looked at close to a billion job advertisements. And they found that in occupations highly exposed to AI, the specific skills employers are seeking are changing 66% faster than in less exposed roles. 66%. 66%. That's insane. It's a massive acceleration in skill turnover. It means the core requirements of a job are literally morphing within a single fiscal year.

[00:04:29] But the actual pay isn't mutating at all, which brings us right back to our supply chain analysts. The gap between what workers expect and what they experience is staggering. The expectation gap is huge. Yeah. According to that same pay scale data, 56% of employees believe that if you upskill in AI, you should earn more money, which, you know, seems like basic capitalist logic. You'd think so, yeah.

[00:04:54] Yet in reality, 70% of employees who have built these AI skills report absolutely zero change in their pay. Zero. Only 16% say their pay has actually risen. Now, I have to push back a little on the idea that companies are just helpless victims of a broken system here. Okay. Play devil's advocate. Is it genuinely that much of a logistical nightmare for a company to just update a pay grade when someone masters a highly valuable software update?

[00:05:20] I mean, it kind of just sounds like an excuse to withhold wages. Oh, no. It sounds like a convenient excuse. But the logistical nightmare is very real. And to understand why, we really have to look at the mechanics of 20th century human resources. Okay. Take me back to HR 101. So traditional pay systems were fundamentally built to price stable, predictable bundles of duties. Think of how a job description is created. It's essentially a shipping container. Right. A nice neat box.

[00:05:48] Exactly. HR defines the dimensions of the container, the title, the required years of experience, the daily responsibilities. They take that container, they compare it to similar containers at rival companies using annual market surveys. They slot it into a pay grade and then they lock it. And that system assumes the contents of the container are going to radically change shape overnight. Precisely. But generative AI completely unpacks the container. It basically disintegrates the container, doesn't it? It really does. Because AI doesn't change work at the macro level of the job title.

[00:06:18] It changes work at the microscopic level, at the level of individual tasks. And it does this incredibly unevenly across an organization. Right. Depending on who actually uses it. Yeah. You might have two financial analysts sitting right next to each other. One uses a large language model to automate half their workflow, essentially redesigning their own job on a random Tuesday in March. And the other one? The other person? Maybe they're more skeptical or just overwhelmed. So they keep doing things the traditional way.

[00:06:44] So how does an annual review cycle, which relies on a static job description written three years ago, accurately capture and price that sudden individualized transformation? They can't. It's mathematically impossible. Right. Dr. Westover uses a very specific phrase in the text, unit of value. For decades, the unit of value has been the job, but generative AI has made the task or the skill the new unit of value. So when the unit of value changes faster than the system designed to price it, the math just breaks down.

[00:07:14] Chaos ensues. The mathematical models that dictate payroll completely break down. And when those internal models break down, organizations naturally panic. They realize they have no internal mechanism to value these new skills, so they look outward. They desperately try to find external benchmarks. Yeah, to tell them what AI fluency is actually worth on the open market. And this leads us into this highly volatile territory of what the article calls the phantom premium. Ooh, the phantom premium.

[00:07:43] Because if you look at the labor market data right now, it is experiencing massive turbulence. It's like looking at two completely different economies, depending on whether you're a new hire or an incumbent employee. Yeah, that turbulence is causing massive headaches for compensation committees. But before we dig into those divergent economies, we really need to draw a firm line between two terms that are constantly conflated by the media. Oh, AI fluency versus AI specialist skills. Yes. Because if you mix those up, your budget is going to be completely ruined. Absolutely.

[00:08:13] So AI specialist skills are the deep, highly technical capabilities, right? The machine learning engineers, the data scientists actually building the foundational models. Exactly. The people writing the underlying architecture. Those skills have commanded massive astronomical pay premiums for years. And that's a totally separate conversation. Right. They're in a league of their own. What Dr. Westover's research focuses on today is AI fluency. Which is for the everyday knowledge worker. Right.

[00:08:39] This is the ability of a marketer, an HR professional, or our hypothetical supply chain analyst to use generative AI tools effectively and responsibly in their daily workflow. So knowing how to frame complex tasks for a model, evaluating outputs for hallucinations, weaving it into legacy systems. And arguably most importantly, knowing when not to rely on AI. Okay.

[00:09:04] So keeping that strict definition of everyday fluency in mind, let's look at this conflicting data on the AI skill premium. If you look at the recruiting side, what companies are offering to acquire new talent, the money looks incredible. It looks like a gold rush. It does. PwC noted in 2025 that jobs requiring AI skills carried a 56% wage premium. That was a massive jump from just a 25% premium the year before. Huge leap.

[00:09:30] And then you have Lightcast, who analyzed over 1.3 billion job postings. They found a 28% premium, which roughly translates to nearly $18,000 a year in extra salary. And 51% of those jobs were outside of traditional IT and computer science fields, right? Exactly. So that's the advertised external premium. But then look at the realized pay for people already inside the building.

[00:09:57] Corn Fairy ran a global survey in 2026, covering 4,000 organizations. And what did they find? They found that roles requiring AI skills were only being paid about 10% above market. It generally hovered in this tight range of 5% to 15%. A huge drop from 56%. Yeah. And 64% of the comp professionals they'd surveyed admitted they simply didn't know what to pay. The contrast there is exactly why employees are so frustrated. Right.

[00:10:26] Like, if I'm an employee and I'm seeing job postings offering an $18,000 premium to walk in the door with these skills, but I'm only getting a tiny 10% bump, if anything, when I learn them at my current job. You feel cheated. My immediate assumption is that my company is taking advantage of me. Okay. Is there a fundamental economic difference in how we value acquiring a new employee with AI skills versus retaining a loyal employee who took the initiative to learn them?

[00:10:54] Well, it's less about corporate malice and more about a fundamental misunderstanding of how these new skills actually generate commercial value. Okay. Unpack that for me. What's fascinating here is to understand that gap between the advertised premium and the incumbent reality, we have to look at the freelance market. Dr. Westover cites this 2024 study by Stephanie and Tutloff that looked at freelancing platforms. Right. And they found a 21% premium for AI skills there. Yes.

[00:11:21] But the critical theoretical concept they introduced is complementarity. Oh, I was looking at their data on complementarity and it really clicked for me because it proves AI isn't just a standalone superpower. It acts as a multiplier. Exactly. It doesn't exist in a vacuum. Like if I just know how to type prompts into a chat interface that has, you know, some marginal value.

[00:11:44] But if I am an international corporate tax lawyer who also knows how to engineer prompts to parse through thousands of pages of newly passed tax legislation. Now that is valuable. That specific combination is where the massive premium lies. You've hit the nail on the head. Basic prompt engineering is becoming rapidly commoditized. Anyone can do it.

[00:12:05] But the combination of AI fluency with deep pre-existing domain expertise, that complementarity is incredibly rare and highly valuable. So how does that map back to the divergent data we saw in figure one? Well, when you look at advertised premiums, the 56% from PwC or the 28% from Lightcast, you're looking at talent acquisition budgets. The hiring budget. Right.

[00:12:30] A company is willing to pay a massive premium to desperately acquire scarce, highly complementary talent from the outside because they believe that person will catalyze an immediate transformation of an entire department. They're paying for the promise of a revolution. Exactly. But the Corn Ferry survey is looking at the reality of the existing payroll. The 10% incumbent bump reflects the constrained reality of internal budgets and those rigid math models of pay bans we discussed earlier.

[00:12:58] Because you can't just blow up the pay scale for one person. No. A comp manager cannot easily blow up their internal pay equity structure to give one tax accountant a 30% raise just because they figured out co-pilot. If they do, they break the pay structure for the other 50 accountants in the department. So the advertised premiums represent the speculative wild west of talent acquisition. And the incumbent surveys represent the highly regulated reality of corporate finance. Exactly.

[00:13:26] The problem is when that bureaucratic reality is paralyzed by indecision, I mean, 64% of comp professionals have no idea how to proceed. The executive suite doesn't just sit still. Oh, they never sit still. They improvise. They improvise. They try to force these new behaviors using whatever blunt instruments they have available. Which brings us to the dark side of AI integration where things get incredibly messy. Very messy. Because when you don't have a structured architecture to price a skill, you start creating arbitrary metrics.

[00:13:56] And that leads directly to metric gaming and what the article identifies as the free fluency trap. This is where we see immense organizational damage happening in real time. In the absence of a coherent comp strategy, organizations fall back on ad hoc, panic-driven management strategies. And the most common one seems to be the usage dashboards. The reliance on usage dashboards is arguably the most destructive strategy they could pick.

[00:14:22] The dashboards are fascinating because, I mean, they seem so rational on the surface, right? The text outlines how some companies are just declaring that AI fluency is the new baseline. They call it free fluency, meaning it's an unpaid expectation. Right. Just a part of the job now. But to enforce that, they need to measure it. So they grade their employees on how much they interact with AI tools. It seems rational to a software engineer building the tool, sure.

[00:14:51] But to an organizational psychologist, it is a textbook example of metric distortion. Tell me about Kerr's theory here because this blew my mind. Yeah. There is the seminal paper from 1975 by Stephen Kerr titled, On the Folly of Rewarding A While Hoping for B. Such a great title. It really is. The core premise is that organizations consistently design reward systems that actively incentivize behaviors they do not want,

[00:15:18] while inadvertently punishing the exact behaviors they are trying to cultivate. It totally reminds me of those old stories from the 1990s where software companies tried to pay programmers based on the number of lines of code they wrote. Oh, yes. A classic disaster. And suddenly everyone was writing incredibly bloated, inefficient software just to hit their quotas. AI usage dashboards are the modern equivalent of rewarding lines of code. It is the exact same mechanism of failure.

[00:15:46] What the company actually wants, the B in Kerr's equation, is increased operational efficiency, sharper decision making, higher quality output. But what are they actually measuring? What they are measuring and rewarding, the A, is the sheer uncontextualized volume of API calls or chat interactions. Just clicking the button. Exactly. The article notes that by late 2025, Meta had informed employees that AI-driven impact would be a core expectation in performance reviews,

[00:16:14] with internal tools aggregating data on the volume of AI-generated code. And JPMorgan Chase was doing something similar, right? Yeah. JPMorgan was reportedly using internal dashboards to explicitly classify their tech employees as light, heavy, or non-users of AI tools. But, I mean, think about the human behavior there.

[00:16:33] If I'm an employee at one of those companies and I know that my annual bonus, my promotion, or even my basic job security depends on being classified as a heavy user on an HR dashboard, my behavior is going to change immediately. And not for the better. Not at all. I'm going to use AI for absolutely everything. I'll have it write simple internal emails that would have taken me 10 seconds to type. I'll have a generated boilerplate code I could have written from memory. You'll use it just to bump your numbers.

[00:16:59] I will use it even when a highly sensitive, nuanced human touch is absolutely required. It's, you know, it's worse than just swinging a hammer aimlessly. It's like swinging a hammer at the plumbing just to prove I'm busy. You're literally incentivizing employees to produce garbage output just to inflate their stats. And that inflated activity just clogs the entire system. But there is a secondary, perhaps more dangerous consequence to Kerr's folly here.

[00:17:24] Usage dashboards actively penalize the employee who exercises high-level professional judgment. Oh, right. Because sometimes the right move is not to use AI. Exactly. As we discussed with the definition of fluency, sometimes the most valuable thing a seasoned professional can do is look at a complex problem and determine generative AI is not the appropriate tool for this. Like if the data involves client confidentiality. Or the strategic nuance is just too delicate, so they execute it manually.

[00:17:54] But a blunt usage dashboard flags that critical expert judgment as noncompliance or light usage. So it punishes the exact discernment the company relies on to avoid catastrophic errors. It does. It completely backfires. And it seems like the market provided a very quick reality check on this. Because the text mentions that companies like Duolingo and Amazon actually ended up quietly dropping their internal AI leaderboards and usage tracking from performance reviews.

[00:18:23] The internal pushback must have been severe. Oh, I can only imagine. But even if the dashboards disappear, the damage to the corporate culture is already done. Because these mandates feed into a massive underlying culture of fear. A very real fear. We have Mercer data from 2026 showing that the fear of AI-driven job loss had spiked to 40% among employees. Four out of every 10 people sitting in a cubicle are terrified this tech is going to render them obsolete. And extreme mandates just fuel that.

[00:18:53] Yeah. When a CEO issues an extreme mandate, it doesn't inspire innovation. It induces panic. The mandates shift the environment from clumsy HR policy to a deeply toxic psychological space. The article highlights some of the more aggressive examples from the tech sector. Like the Coinbase example. That was intense. Very intense. In 2025, the CEO of Coinbase publicly stated that he required engineers to adopt AI coding assistance within a single week.

[00:19:22] And reports indicated that some who didn't comply within that incredibly compressed time frame were fired. Within a week. Yeah. Accenture told investors they were actively exiting staff they couldn't reskill on a heavily accelerated timeline. Shopify and Microsoft issued strict, top-down mandates that AI use was no longer an optional experiment, but a core metric for review cycles.

[00:19:47] But if I'm operating under the threat of being fired in seven days because I haven't mastered a brand new coding assistant, I am not thinking about how to creatively redesign my workflow to save the company money. No, you're purely in a survival mindset. I'm doing whatever it takes to look compliant on paper. And that survival mindset destroys the discretionary effort that AI integration actually requires. To understand the psychological toll here, we have to look at equity theory. Oh, I love equity theory. Developed by Adams, right? In 1965. J.

[00:20:17] J.C. Adams, exactly. Equity theory posits that human beings in a workplace are constantly, subconsciously calculating a ratio. Like a mental ledger. Right. On one side of the ledger are their inputs. The effort they expend, the stress they endure, the new complex systems they are forced to learn. On the other side are their outputs or rewards. Their salary, their job security, their organizational status. And then they compare their ratio to their peers. Exactly.

[00:20:45] When a company adopts a free fluency model demanding massive upskilling without corresponding pay or worse, a mandate backed by the explicit threat of termination, the employees' required inputs skyrocket. But their outputs remain flat or are actually threatened. The ratio completely fractures. It shatters. It shatters.

[00:21:07] And when that ratio breaks, employees experience a profound loss of what organizational psychologists call distributive and procedural justice. They feel the outcomes they are receiving are fundamentally unfair. And the process is opaque and arbitrary. So how do they respond to that perceived loss of equity? Adam's theory tells us exactly how. If they cannot increase their rewards, they will actively decrease their inputs to balance the scales themselves.

[00:21:34] Okay, but in the context of AI, how does that actually manifest? Do they just start missing deadlines or doing shoddy work? Sometimes it leads to classic disengagement, sure. But with AI, it manifests in a very specific modern phenomenon. Employees actively hide their efficiencies. Wow. Look at the pay scale data Dr. Westover provides. When employees successfully use AI to save time, 28% say they are simply doing more of the exact same legacy work in that safe time.

[00:22:03] Just treading water faster? Yeah. Only 19% report using the time they saved for higher value strategic or creative work. If my paycheck stays exactly the same and I know that revealing my new AI-driven efficiency will just result in management piling double the workload onto my desk. For the exact same salary. Right. Why on earth would I demonstrate how fast I can work? I wouldn't. I wouldn't either. I would absolutely keep my mouth shut.

[00:22:27] Like, if I figure out how to automate a Friday afternoon report so that it takes me 10 minutes instead of 4 hours, I am going to run the script at 1 p.m., schedule the email to send at 4.45 p.m., and I'm going to spend the afternoon reading a book. Exactly. The corporate system is currently providing zero financial or structural incentive for me to share that productivity gain with the broader enterprise. Which leads to a massive systemic misallocation of human effort.

[00:22:53] The organization loses out on the discretionary experimentation, the peer-to-peer teaching, and the sharing of best practices that makes enterprise-wide AI adoption actually profitable. They think they're getting free fluency, but they're getting a workforce conspiring to conceal its true capacity. That is exactly what's happening. So treating AI as an unpaid baseline expectation crushes morale, and grading it on a dashboard encourages fake work.

[00:23:19] But as we move into the next critical area of Dr. Westover's research, it becomes clear that this improvisational approach doesn't just hurt feelings. No, it goes way beyond that. It actively discriminates. It creates massive structural inequities and, frankly, severe legal liabilities for the company. Let's explore the hidden traps of pay dispersion and the widening gender gap.

[00:23:39] This is the pivot point where the structural mismatch stops being merely an internal headache for HR and transforms into a severe existential liability for the executive board. Because they're paying for skills in the dark, right? Precisely. When companies refuse to build a formalized, transparent architecture for pricing AI skills, they don't stop paying for them. They just pay for them in the dark. What typically happens is that frontline managers are given quiet discretion to hand out retention bonuses.

[00:24:07] Or discretionary salary bumps on a case-by-case basis. Right, to keep their best people from leaving. So the employee with the most market leverage, or frankly, the one who is most aggressive and skilled at self-promotion, marches into their manager's office, threatens to take an external offer, and secures a 20% raise for being an AI champion. And their coworker, the one who might actually be executing more complex workflows but is less confrontational. They get absolutely nothing.

[00:24:33] And that creates what the article refers to as wage scatter, or unplanned pay dispersion. Now, on the surface, some managers might argue, you know, that's just capitalism functioning as intended. The squeaky wheel gets the grease. Right. High performers negotiate better salaries. Why is that actually a problem for the company? It is a massive problem because it destroys the collaborative tissue of the organization. The academic research on pay dispersion is incredibly clear here.

[00:24:59] The article cites a fascinating 1999 study by Matt Bloom that analyzed major league baseball teams to prove this exact point. Baseball teams? Yeah. Bloom looked up the payroll structures of different franchises and found that teams with more compressed, equitable, and transparent pay structures consistently and significantly outperformed teams that had wider, unexplained pay dispersion. Oh, wow. In fact, he found that even the highly paid superstar players performed measurably worse when they were placed on teams with massive, arbitrary pay gaps.

[00:25:29] That is fascinating. And this finding was later confirmed in a corporate setting by Trevor and colleagues in a 2012 study specifically regarding highly interdependent work. So, okay, when you say interdependent work, you mean tasks where an individual cannot succeed in a vacuum, right? Where my output relies heavily on the quality of your input. Precisely. Okay. And AI-enabled knowledge work is almost entirely interdependent. Think about the reality of a modern analytics team, right?

[00:25:56] If one senior analyst figures out an incredibly elegant series of prompts to clean up a chronically messy data set. They don't just keep that to themselves. Right. That prompt library needs to become the team's new shared infrastructure. Everyone built upon it. But if management secretly gives that one analyst a massive discretionary bonus without a clear, transparent rationale that the whole team understands and accepts as fair. It breathes deep, toxic resentment. It does. The colleagues who didn't get the bonus stop sharing their own discoveries.

[00:26:25] Unexplained pay dispersion destroys the exact collaborative environment required for AI innovations to diffuse across a team. That makes complete sense regarding team morale and knowledge hoarding. But here is where Dr. Westover's research gets incredibly alarming from a compliance perspective. The demographic data. Yes.

[00:26:45] How does rewarding the use of a software tool, something like ChatGPT, which is literally just code in mathematics, theoretically neutral, how does rewarding its use inherently favor one demographic over another? Because the statistics presented here are jarring. It is perhaps the most critical blind spot for corporate leadership today. There's this pervasive assumption in Silicon Valley and HR that technology is inherently neutral and therefore tying compensation to tech adoption is a pure objective meritocracy. Right.

[00:27:15] A level playing field. But the sociological data tells a drastically different story. The article highlights a major 2025 study by Humlum and Vestergaard that surveyed 18,000 workers in Denmark, specifically across occupations that were highly exposed to AI. And what do they find? They found that women were 16 percentage points less likely than men to have used ChatGPT for work. 16 points. Yes. Even when comparing men and women sitting in the exact same occupations, performing the exact same tasks.

[00:27:44] And this wasn't an isolated regional anomaly. Another massive synthesis published in 2024 by Otis and colleagues, which covered over 140,000 people globally, confirmed this trend. So it's everywhere. The gender gap in generative AI utilization appears persistently across nearly all geographic regions, all economic sectors and all occupations. A 16 percentage point gap within the exact same role is massive. What is the mechanism driving that?

[00:28:10] Because I refuse to believe it has anything to do with innate technical capability or, you know, a lack of interest in the technology. You are correct to be skeptical because it has absolutely nothing to do with technical aptitude. The research indicates it has everything to do with the structural lived realities of the modern workplace. Okay, unpack that. We have to look at what the adoption of a highly experimental, unproven tool actually requires from an employee during the workday. It requires three critical components. Which are?

[00:28:38] First, the unallocated time to learn and experiment. Second, managerial encouragement and sponsorship to step away from daily deliverables to try something new. And the third? Most crucially, the cultural permission to experiment and fail without being perceived as cutting corners or doing sloppy work. And historically, those three components are not distributed equally across demographics. They are not.

[00:29:01] Structurally, women often face higher baseline administrative and invisible labor burdens in the office, the non-promotable tasks. Right. Planning the office parties, taking the meeting notes. Exactly. They frequently receive less proactive high-level sponsorship for risky experimental projects. And sociologically, they are often judged significantly more harshly for utilizing shortcuts or making errors during an experimental phase. Oh, that makes so much sense.

[00:29:26] If an organization's underlying culture makes a female employee feel that utilizing an AI tool to speed up a report will be scrutinized as her cheating rather than celebrated as her innovating, she will rationally and rightfully avoid using the tool to protect her professional reputation. So if a company institutes a policy tying 10% of an annual bonus to self-reported AI usage,

[00:29:51] or if managers are handing out discretionary bumps to the employees who loudly brag about their prompt engineering... They aren't actually measuring pure technical skill. No. They're measuring who feels the most culturally comfortable, historically protected, and unburdened enough to play around with beta software during corporate hours. That is exactly the mechanism. You are inadvertently rewarding a pre-existing cultural privilege disguised as a technical skill.

[00:30:18] And if you tie variable pay or base salary to that usage without first doing the hard work of fixing the underlying access and culture gaps... You're going to systematically underpay women. You are going to take an existing historical gender pay gap and pour rocket fuel on it. And that is exactly where the corporate lawyers start sweating. Because the text brings up the European Union's pay transparency directive. Yes, the game changer. This went into effect for transposition by member states in June 2026,

[00:30:45] with the heavy reporting requirements kicking in for large employers in 2027. And from what I understand, this directive changes the game entirely. It's not just a slap on the wrist for a minor fine anymore. The EU directive has completely altered the legal calculus of compensation. And multinational companies are frankly terrified of it. Under these new regulations, pay structures must rest entirely on objective, explicitly gender-neutral criteria.

[00:31:13] What's the specific mechanism that's freaking out the CFOs? Here is the mechanism that keeps them awake at night. If an employer's reporting reveals an unexplained average gender pay gap of just 5% within any category of workers performing work of equal value... Just 5%? Just 5%. And they cannot objectively justify that gap with a neutral criteria or fix it within six months, it triggers a mandatory joint pay assessment. What does a joint pay assessment actually entail?

[00:31:41] Because it sounds incredibly invasive. Oh, it is incredibly invasive. It essentially means opening up your internal compensation books, your performance review data, and your historical promotion criteria to workers' representatives, labor unions, and national equality bodies. Wow. It is a highly public, deeply disruptive, and legally fraught process that strips management of its unilateral control over payroll. So let's play this out.

[00:32:07] If your frontline managers have been handing out these ad hoc discretionary AI bumps to the guys on the analytics team who aggressively negotiate for them... Right. ...and suddenly upon reporting you hit that 5% gap trigger, you can't just tell the EU auditors, well, Dave just really loves chat GPT and asks for more money. You absolutely cannot. An ad hoc AI premium that relies on managerial discretion that cannot be documented, mathematically explained,

[00:32:36] and proven to be completely free of systemic demographic bias is legally indefensible in the European Union right now. And it's not just Europe, right? No. The regulatory tide is shifting globally. States in the U.S. are passing similar transparency laws. Any organization relying on gut feeling or manager discretion to price AI skills is walking blindly into a massive class action lawsuit. Okay, we have painted a fairly terrifying picture of the current landscape here. That is pretty grim.

[00:33:03] We have legacy HR systems that are fundamentally broken, phantom market premiums causing internal chaos, usage metrics that destroy actual business value. Plunging employee morale. Yep, plunging morale due to equity violations, and massive looming legal risks surrounding the gender pay gap. It sounds like an absolute disaster for the modern enterprise. But thankfully there is a way out. There is hope. Dr. Westover doesn't just diagnose the pathology here. He provides the blueprint for a cure.

[00:33:31] Let's pivot away from the problem and walk through the five evidence-based solutions from the text. We can look at how organizations rebuild their architecture and actually pay for verified capability. This is where the conversation moves from structural panic to operational pragmatism. If we accept that the core problem is that the unit of value has shifted from the broad job title to the specific task,

[00:33:56] then the very first solution must be to rebuild the job architecture so the system can actually see those tasks. Because you can't financially reward what your system cannot accurately identify. Exactly. So Dr. Wistover's first solution is to rebuild the job architecture around skills rather than just titles. The text explicitly says we need to map tasks. But what does that actually look like in practice?

[00:34:20] Because mapping every single task of a 10,000 person global workforce sounds like an administrative nightmare that would take years to complete. It is a heavy lift for sure. But it requires a fundamental shift in how HR constructs a role. You don't necessarily have to abandon traditional job titles entirely. That would cause too much friction with external recruiting. Right. You still need to hire a marketing manager.

[00:34:44] Yes. But you have to redefine the internal reality of the job as a dynamic basket of specific, verifiable skills. You have to stop looking at market surveys based purely on the title Financial Analyst 3 and start looking at the specific AI capabilities that specific analyst is now required to wield to produce their deliverables. And the article highlights IBM as a major pioneer in this space, right? Yes. IBM's approach is fascinating because they separate the skill from the outcome.

[00:35:13] The text notes that IBM assesses its employees on three pillars, business results, behaviors, and skills. And critically, they weigh skills objectives equally with business objectives. Which is huge. Your base pay progression and your equity grants are tied directly to your verified skills growth, while your variable bonus is tied to the actual business results you produced that quarter. So they separate the two concepts.

[00:35:39] Are you becoming a more capable, valuable asset versus did the company hit its quarterly revenue targets? That separation is a brilliant piece of organizational design. Think about the macroeconomic reality. You can spend six months learning a wildly valuable, complex new AI programming skill. Right. But if the broader economy tanks or a major client goes bankrupt, the company might miss its quarterly targets entirely through no fault of your own.

[00:36:07] Your newly acquired skill didn't lose its intrinsic value. The external context just changed. Ah, I see. So tying base pay progression directly to the capability you acquired protects the employee from that macro volatility. Exactly. It guarantees a return on their learning investment, which builds immense trust. And this isn't just a strategy for elite software engineers at tech companies either.

[00:36:32] The text mentions Walmart applying a skills-based architecture to frontline operational work. Yeah, the Walmart example proves that this architecture scales across the entire economy. In early 2026, Walmart created a brand new role called the Pharmacy Operations Team Lead. And they priced it very aggressively for the retail sector, right? Up to $42 an hour. Yes. But the critical innovation is how an associate attains that role.

[00:36:58] It is not based on possessing a specific external college degree, and it is not based on putting in a certain number of years of tenure on the floor. So what is it based on? Advancement into that $42 an hour band is based entirely on internal on-the-job training and specific verified skill certification. They built a new architecture where the capability itself, not the pedigree, is what unlocks the pay band. Which brings us perfectly to the second major solution in the text.

[00:37:27] Paying for verified capability rather than paying for usage. We have thoroughly established that paying for usage, the dashboards, the leaderboards, is a behavioral disaster. A complete disaster. So how does an enterprise verify capability without just handing out automatic raises to every employee who manages to click through a 30-minute online video course? That distinction is the linchpin of the whole system.

[00:37:52] The historical research on skill-based pay, which actually goes back decades in advanced manufacturing settings, shows that it only works when the skills are clearly defined by management, demonstrably acquired by the employee, and actually utilized in the daily flow of work. So you can't just watch a video? No, you cannot certify mere exposure to a topic. You must certify latent capability through demonstrated outputs, proctored assessments, or rigorous work samples. It reminds me of the economic model of a lifeguard. Okay, tell me more.

[00:38:21] If you hire a lifeguard, you're paying a premium for their CPR and water rescue certifications. You're paying for the verified, tested capability that they can save a drowning person if a crisis occurs. You are absolutely not paying them based on how many times they blow their whistle or how many times they jump in the water on a Tuesday. Because if you did... If you paid lifeguards per rescue, you'd inadvertently incentivize them to push people into the deep end just to pad their daily stats.

[00:38:48] That is a perfect, visceral analogy for why usage metrics fail. You are paying for the latent capacity and the verified competence to act when needed. And the article gives a great real-world example of this with Honda, right? Yes. Honda recently rolled out a rigorous three-tier AI certification program for its workforce in Japan. It is not a rubber stamp process. It involves a comprehensive written examination and a practical interview. To prove they can actually apply the tools to Honda's specific engineering problems.

[00:39:18] Exactly. If an employee achieves the top tier of certification, they don't just get a certificate for their cubicle wall. They get a monthly allowance of up to 135,000 yen on top of their standard base pay. Now, the use of an allowance there is a really interesting structural choice. Why would a company use a monthly allowance instead of just permanently bumping up the employee's base salary by the equivalent amount? This is one of the most vital strategic insights in Dr. Westover's entire article.

[00:39:45] And it addresses the core fear of CFOs. Technology, especially generative AI, changes at breakneck speed. So fast. The specific AI skill that is incredibly rare and valuable today, say, writing complex middleware to connect a legacy database to an LLM, might be built directly into the standard Microsoft Office suite in 18 months. Making it a completely ubiquitous baseline skill. Precisely.

[00:40:10] If a company bakes a $10,000 premium permanently into an employee's base salary for a skill that rapidly becomes obsolete, they've created a permanent compounding legacy cost that they can never take back. Because reducing base pay is basically impossible. From an HR perspective, reducing an employee's base pay is almost impossible to do without completely destroying their morale and prompting them to quit. Ah, I see. So if the skill becomes standard, you're stuck overpaying for a commodity forever. Exactly.

[00:40:39] That is why Dr. Westover stresses the vital architectural rule of sunset and refresh. Sunset and refresh. I like that. When you utilize an allowance or a time-limited premium for a fast-changing technical skill, you explicitly set a review date from day one. You tell the employee, this specific AI certification is valid for 18 months and it carries this specific monthly allowance. We will reevaluate the market value and the relevance of this skill at the end of that term.

[00:41:08] It protects the company's long-term balance sheet, prevents wage blowed, while still richly rewarding the employee in the immediate present. And it's worth noting, the article draws a parallel to the healthcare sector here. Hospitals have used clinical ladders for nurses for decades. Oh, like Piedmont Healthcare. Yeah. Piedmont Healthcare and Houston Methodist pay specific, tiered, temporary bonuses when nurses verify new, specialized clinical competencies.

[00:41:35] The corporate world already knows how to administer this kind of system. They just have to map that existing clinical ladder concept onto AI capabilities. And there is one more piece to this verified capability section that I found incredibly profound. The text argues that any robust AI certification process must explicitly include judgment as a core certifiable skill. It's not just about what you can build. It's about the ability to know when not to use AI.

[00:42:04] Yes, because if you are verifying an employee's capability, a massive part of that professional competence is discernment. A top-tier AI user isn't just an employee who can generate a thousand lines of code or 50 marketing emails in a minute. It's the professional who can look at that rapid output, instantly recognize the subtle hallucinations, the embedded biases, or the security flaws. And make the executive decision to scrap the AI output and execute the task manually.

[00:42:32] Rewarding and certifying judgment is the only way a company prevents AI from becoming an automated liability factory. Okay, so we've established how to fix the architecture. And we know how to pay for verified capacity through smart, time-limited allowances. But what about the actual raw productivity gains? The output. Right, because AI isn't just about latent capacity. It's about massive output.

[00:42:55] If our supply chain analyst figures out how to automate a week's worth of global reporting into a one-hour automated task, how does the enterprise reward that specific realized outcome? How do they share the wealth? Exactly. How do they share the wealth generated by that efficiency? That leads us directly into the second half of these evidence-based responses. This is where the compensation strategy moves from paying for the potential to do good work to rewarding the actual financial realization of that work.

[00:43:22] The third major solution is about sharing measured gains through bounded variable pay. Bounded variable pay. So bonuses tied strictly to specific realized metrics. The article gives a fascinating example with an e-commerce marketing platform called Omnisend. A very clever approach. They introduce what they call an AI salary bump. It's a 2% to 4% raise specifically targeted at measurable savings. But the brilliant catch is you don't just get the money for saving your own time.

[00:43:52] Now there's a condition. You only get the bump if you build AI workflows that are successfully adopted by your colleagues. And this adoption is evaluated quarterly. So think about what this does to the workplace culture. If Omnisend gives you a financial bump only if your colleagues willingly adopt your workflow, doesn't that fundamentally rewrite how teams interact? It completely inverts the standard corporate incentive structure.

[00:44:16] Doesn't that heavily incentivize me to be a proactive team player rather than hoarding my AI secrets in my own little silo? Yes. By explicitly tying the financial reward to pure adoption, Omnisend is explicitly rewarding the diffusion of knowledge rather than individual virtuosity. Diffusion rather than virtuosity. It doesn't matter if you are an isolated genius who built the most brilliant, complex prompt in the company.

[00:44:41] If you lack the communication skills to teach it to the rest of your team and make them faster, you do not get the financial reward. Which directly solves the problem we discussed earlier about employees quietly hiding their efficiencies to avoid more work. Exactly. It turns every single employee into an internal, highly motivated evangelist for collective productivity. Shark Ninja took a slightly different, more high-profile approach, right?

[00:45:05] They created a massive $1 million prize fund specifically for AI experiments that produce verifiable business value. It's more of a showcase model. Right. Meant to loudly signal what the company values. But there is a deeper, highly academic reason why this kind of team level or widespread variable pay is so critical right now. And it has to do with how AI actually affects different skill levels within an organization. The Brynjolfsson study.

[00:45:32] Yes. I was looking at that 2025 Brynjolfsson study on customer support agents mentioned in the text, and it completely shattered my assumption about who actually benefits from this technology. I assume the senior agents would get a massive exponential boost. But the data showed the exact opposite. Exactly. That study by Eric Brynjolfsson and his team at Stanford is arguably one of the most important pieces of economic research regarding AI in the modern workplace.

[00:45:59] They looked at thousands of customer support agents using generative AI tools. And what did they definitively find? They found that AI acts as a massive, aggressive leveler of skill. It boosts the productivity of novice, low-skilled, or newly hired workers dramatically by 14% or more. But it provides almost zero measurable boost to the highly experienced senior experts who are already operating at peak efficiency based on years of accumulated knowledge.

[00:46:26] So the AI essentially takes your C-level junior player and instantly elevates them to an A-minus player while the A-plus player stays exactly where they are. Precisely. It aggressively compresses the performance gap between a novice and a seasoned expert. Now think about what that compression does to the architecture of traditional individual merit pay. Oh, wow. You know, in a legacy system, you justify paying your senior experts significantly more money because they objectively produce significantly more output at a higher quality than the juniors.

[00:46:56] But if the junior employees armed with generative AI are suddenly producing at nearly the exact same level of quality and speed... Individual merit pay becomes incredibly complicated and politically toxic internally. How do you justify a 50% pay gap between a senior and a junior if their outputs are now identical? That is why Dr. Westover argues that team-level gain-sharing becomes crucial.

[00:47:18] You have to pivot to rewarding the collective output of the department because the technology is fundamentally blurring the lines of individual, isolated contribution. That is a wild organizational dynamic to think about. And it highlights the danger of relying too heavily on the AI to level the playing field. Because you still need the experts. Which brings us to one of my absolute favorite examples in the entire article. The human counterweight? The concept of the human counterweight.

[00:47:43] EY, the massive global accounting and consulting firm, set up a $100 million bonus pool. But they aren't paying people a dime for using AI. They are explicitly using this massive fund to reward human skills. It is a brilliant piece of organizational messaging and behavioral design. EY is offering spot bonuses of up to $500 and major impact awards of up to $25,000 specifically for demonstrating high-level human judgment, adaptability, proactive coaching, and relationship building.

[00:48:14] And the key is that any employee can nominate a colleague. Yes. By doing this, EY is making absolutely sure that the organization doesn't accidentally signal to its workforce that automated technical output is the only thing the firm values. Because if you only build rewards around AI use, your employees will adapt by becoming unthinking robots. Exactly.

[00:48:33] If you purposefully carve out a massive, highly visible budget to financially reward empathy, ethical judgment, and complex problem solving, you constantly remind the workforce that the human element is what actually makes the AI outputs trustworthy and valuable to the end client. It's like paying handsomely for the guardrails, knowing that the engine is going to run itself. Now let's look at the final two solutions from the text, which really tie the structural bow on all of this.

[00:49:01] Equity audits and funded development. The foundation of it all. The core rule here seems incredibly simple but frequently ignored. Publish the rules of the game beforehand. And do not demand new fluency from your workforce without explicitly paying to fund it. And the contrast the article draws here is stark. Very stark. On one side, we have the Walmart model.

[00:49:25] They decided they wanted their U.S. associates to get up to speed on generative AI, so they proactively partnered with OpenAI to create custom certifications. And they provided them for free during paid working hours before they ever made it a requirement for career advancement. Walmart fundamentally understood that if you want to alter the baseline expectations of a massive workforce, you have to systematically lower the barrier to entry.

[00:49:49] They provided the protected time, the technological tools, and the financial backing to get their people certified, removing the friction from upskilling. And then on the exact opposite side of the spectrum, we have the sanctions-first approach we discussed earlier. Coinbase firing people within a week or Accenture aggressively exiting staff they couldn't immediately reskill. The article makes a very pointed argument that an unpaid expectation to learn AI on your own time is effectively a pay cut in disguise.

[00:50:19] It is an objective pay cut because an employee's time is their core economic asset. If an employer tells an employee, you must learn this highly complex, constantly evolving new system in order to keep your job, but we aren't giving you any protected hours during the workday to learn it, and we aren't paying for the external courses. The employer is forcing that individual to cannibalize their personal uncompensated time. Exactly. The company is actively increasing the employee's job demands without increasing their compensation.

[00:50:49] As we discussed deeply with Adam's equity theory, the employee immediately recognizes this as a fundamentally unfair, exploitative deal. So the Walmart approach builds long-term loyalty and deep, latent capability. And the sanctions-first approach builds shallow compliance born of terror. And the absolute moment the macroeconomic environment improves, those terrified employees will leave for a competitor who actually invests in their human capital. So we have all these disparate pieces on the table now.

[00:51:17] We have base pay tied to verified skills. We have time-limited allowances for fast-changing technical certs. We have bounded variable pay for shared outcomes and workflow diffusion. And we have fully funded baseline training. It's a massive amount of moving parts. It really is. How does an enterprise actually fit all these puzzle pieces together into a single coherent corporate strategy? Because if I'm an HR director listening to this, I can't just throw all these ideas at the wall. I need a structured blueprint I can take to the executive board tomorrow morning.

[00:51:47] And providing that actionable blueprint is exactly how Dr. Westover concludes the article. He synthesizes all of this complex theory into a four-layered architecture for long-term pay capability. A four-layered architecture. Yes, it's an incredibly elegant way to organize pricing decisions so that an organization isn't trying to use one blunt instrument, like a permanent base salary increase, to solve every single nuanced problem. Let's debate these four layers organically, starting from the foundation.

[00:52:16] The bottom layer is treating AI fluency as a baseline. For roles where basic AI use is now just a standard, unavoidable part of the daily job, you treat it as an expectation. But the critical rule here, tied back to the Walmart example, is that this layer must be fully funded by the enterprise. The organization must provide the protected learning time and the tools. You don't necessarily pay a premium at this layer, but you absolutely pay for the development. Exactly. It is the cost of doing business in the modern era.

[00:52:43] But as you noted earlier, a baseline training program only gets people in the door. It ensures everyone can operate the machinery safely. Right. But if I'm an employee who actually merges my decade of complex tax law experience with this new funded AI training, basic funding doesn't satisfy my demand for a raise. I am now producing exponentially more value. How does the architecture account for the deep domain experts? That is where you move to the second layer, base pay progression.

[00:53:12] This layer is explicitly designed to reward durable, deeply embedded skills. And referring back to Stephanie and Tutloff's research, this is where the organization financially rewards complementarity. The combination. Yes. You use permanent base salary increases to reward the employee who has successfully, verifiably combined their deep domain expertise, like international tax law or supply chain logistics, with advanced AI capability. That specific combination is rare.

[00:53:41] It is durable against technological shifts, and it is therefore worth a permanent increase in fixed costs. Got it. So the base pay is for the durable combinations. But what about the bleeding edge tech skills that might be obsolete in a year? You don't want to put those in base pay. No, you handle those in the third layer, allowances and premiums. This is the Honda model. This layer is reserved for scarce, highly technical, certified skills that change rapidly. So you pay a monthly or quarterly allowance on top of the base pay.

[00:54:10] And you attach a strict, non-negotiable sunset and refresh date. This gives the enterprise the agility to pay top dollar to acquire or retain what is hot right now, without permanently inflating their fixed payroll costs if the technological landscape shifts dramatically in 18 months. Which we know for an absolute fact it will. Okay, so we have the baseline funded, the durable skills in base pay, the volatile tech skills in allowances. What is the final layer at the top?

[00:54:38] The top layer is variable pay. This is the OmniSEN and Shark Ninja model combined with the EY model. This is where you share the measured financial gains of AI adoption after the fact. The outcome-based rewards. Yes. You use bonuses and spot awards to recognize employees who build highly efficient workflows that the whole team adopts, driving measurable savings. And vitally, you use this top layer to richly reward the human judgment, the ethical pushback,

[00:55:06] and the adaptability that keeps the organization safe from automated errors. That layered approach is brilliant because it actually structurally solves all the nightmare scenarios we talked about in the first half of the deep dive. By intelligently moving the volatile technical premiums to temporary allowances and moving the outcome-based rewards to variable pay, you completely prevent that permanent toxic wage scatter that destroys team morale. It cleans up the system completely.

[00:55:32] Because the criteria are explicitly defined and mathematically sound for each layer, you protect the organization against the legal liabilities and gender bias of discretionary manager bumps. You are giving your employees a clear, legally defensible deal. If you do X, you get Y, and it will be evaluated at time Z. If we synthesize this entire discussion into a single, broader takeaway, it all comes down to trust. A corporate compensation system is not just a mathematical formula.

[00:56:02] It is the ultimate communication tool. It tells the employee exactly what the organization truly values. Louder than any corporate mission statement ever could. When companies panic and deploy blunt usage dashboards or hand out hidden discretionary bonuses, they communicate chaos, inequity, and disrespect. When they utilize this layered, transparent architecture, they communicate a credible, explicit employment deal.

[00:56:27] And maintaining trust in that explicit deal is what will ultimately determine whether a company thrives in the generative AI era or suffocates under the weight of metric gaming, deep resentment, and massive turnover. It is a total paradigm shift for how we value human labor. And it brings me to a final thought I want to leave you with, something to really mull over based on everything we've unpacked today.

[00:56:53] We discussed that Stanford study by Brynjolfsson, the one that definitively proved generative AI aggressively compresses the performance gap between a novice and an expert. The great leveler. The technology can effectively turn a junior employee into a senior employee's equal almost overnight in terms of raw, high-quality output. If that compression is true, and if the capabilities of AI continue to accelerate that compression,

[00:57:17] how long will it be before the entire structural concept of seniority-based pay collapses completely? That is the ultimate existential question for the future of human resources. I mean, think about it. Are we moving rapidly toward a near future with the phrase, 10 years of experience means absolutely nothing on a resume? Where corporate tenure is utterly irrelevant, and the only thing that dictates the signs of your paycheck is your verified, real-time output,

[00:57:47] and your certified capability in that exact moment? It's a thrilling and, frankly, terrifying prospect for anyone who has relied on their resume's timeline to secure their salary. It brings us right back to our senior supply chain analyst sitting in that annual review. Maybe in five years, you will be arguing with a manager about 3% cost-of-living adjustments based on your tenure. You'll simply plug your verified, real-time capability metrics into the company's transparent, layered architecture,

[00:58:16] and the system will instantly calculate your new market value. No bias, no guessing, no static job descriptions. Just the raw, priced reality of what you can actually do today. It's a whole new world. It really is. We want to thank you so much for bringing your curiosity to this deep dive. Keep questioning the systems around you. Keep building those truly complementary skills, and we will catch you on the next one. Thank you. This week is coming from here. tokenist, upfront