A Debate about the AI Values Test

[00:00:00] Welcome to The Debate. Right now, a staggering 57% of employees are actively hiding their use of generative AI from their managers. Which is just a massive number to wrap your head around. It really is. And you know, they aren't hiding it because they are lazy or because they are cutting corners. They are hiding it because they are terrified. Yeah, they're trying to protect themselves. Exactly. Think about it like when you sign a lease for an apartment. There is an explicit bounded agreement.

[00:00:28] The landlord promises a habitable space and you promise to pay rent. Right. There are rules. Yes. If the roof leaks, there is a protocol. If you want to paint the walls, there are rules. It is a defined mechanism for accountability. But imagine moving into that apartment and over the course of a few months, the landlord starts quietly moving the walls inward while you sleep. Just physically shrinking the space. Exactly. The physical space you occupy is shrinking. The locks on certain doors are changing. They aren't evicting you officially and there is no new lease to sign.

[00:00:58] Wow. Yeah. You just wake up one day and realize the terms of your habitation have completely altered without a single word of discussion. Which is a perfect visualization of the profound disorientation running through the global workforce right now. The rapid integration of AI is not just a technological rollout anymore. No, not at all. We are navigating what Jonathan H. Westover's October 2026 article in the Human Capital Leadership Review frames as the AI values test. Organizations are being actively judged by their workforce on how they deploy this technology.

[00:01:28] And the grading curve is incredibly, Steve. I mean, if you look at the October 2026 Gport report, 58% of employees surveyed globally said they would outright turn down a job offer if a company's AI values clash with their own. That is huge. Right. And executives know this too. That same report found 92% of executives agree their AI strategy is now actively shaping their ability to attract and retain talent. It has evolved into this highly visible signal of how an organization views human beings.

[00:01:56] That brings us directly to the core question we are tackling today. In order to pass this AI values test, what is the defining factor? What actually matters? Right. From my perspective, the foundation of trust in this transition relies entirely on establishing explicit procedural transparency, visible governance, and formalizing the unwritten expectations into a concrete deal. Well, I come at it from a different way.

[00:02:20] While transparency sounds great on paper, I argue that procedural clarity is largely just consultation theater if it isn't backed by enacted material outcomes. Okay. Consultation theater. Yeah. Passing this test doesn't depend on how well you write the memo. It depends on the material distribution of the AI generated gains and the active financial protection of human job security. Let's unpack the mechanisms behind why that written memo is so crucial though.

[00:02:48] I look at this through the lens of signaling theory and the concept of the psychological contract. The unwritten deal. Exactly. For those listening who might not spend their days in organizational psychology, the psychological contract is the unwritten set of expectations between an employer and an employee. It's the assumption that if I work hard and adapt, I have a future here. Which is fundamentally changing. Right.

[00:03:10] And signaling theory essentially means that because employees cannot see the algorithm and they aren't in the boardroom discussions, they look for signals, actions, or communications that indicate the company's true intentions. Right now, the primary driver of workforce anxiety isn't the AI itself. It is the communication vacuum. A vacuum that nature abhors. So employees just fill it with their absolute worst case scenarios. Exactly. Which is why the absence of policy is a massive failure of leadership.

[00:03:38] I mean, Gallup's 2025 data showed only 22% of U.S. employees felt their organization had a clear plan for AI integration. That is incredibly low. It is. And globally, KPMG found only 40% of workplaces even had a generative AI policy. Organizations must explain the strategy before they execute the transition. But explain it how? Well, consider the cautionary tale of Duolingo back in April 2025.

[00:04:03] The CEO, Luis von Ahn, released a memo announcing an AI-first shift, which was primarily focused on phasing out contractors. Now, I remember that. It was a mess. Total mess. Because that communication lacked adequate context and boundary setting about the job security of full-time employees, it caused a massive internal and public backlash. They spent months doing damage control just because they left the psychological contract ambiguous. Sure. They fumbled the messaging. Right. So I see why you think that material guarantees are required.

[00:04:31] But let me give you a different perspective. Bounded, written promises are the only way to manage that psychological contract at scale. By removing ambiguity and establishing clear procedural justice, meaning the rules of the game are known and fair, employers prevent the perceived breaches of trust that destroy a company's culture. The problem is that focusing purely on stated values and procedural justice misses the reality of the technology. Generative AI doesn't just automate routine physical tasks.

[00:05:00] It touches professional, creative, and knowledge work. Okay. It encroaches on human identity. When an organization frames AI simply as an efficiency tool and writes a very polite, transparent memo about it, workers rightly infer that the productivity dividend is going to flow exclusively upward to shareholders. But isn't that inference exactly why clear, documented communication is needed to correct the narrative? But communication alone doesn't change the material reality of the board's intent.

[00:05:29] I mean, Pew research data from late 2024 showed that 32% of workers expected fewer job opportunities for themselves in the long run due to AI. Which is a valid fear. Right. That is not a communication error. That is a rational economic calculation by the workforce. Writing down an AI employment deal only builds trust if the terms of that deal reflect share gains. Even if the deal is tough.

[00:05:51] Especially then, if a company transparently communicates a bad deal, where you train the AI that eventually replaces you, that is still a failed values test. It just provides documentary evidence to the employee that their fears were 100% justified. Yet, I would argue that even communicating a difficult reality is far superior to leaving people in the dark. I have to praise organizations that explicitly define the new baseline, even when it is uncomfortable. Praise them? For what exactly? For candor.

[00:06:19] In April 2025, Shopify CEO Toby Lutke sent a memo making the reflexive use of AI a baseline expectation, tying it directly to performance reviews. And in June 2025, Amazon's Andy Jassy was entirely candid about the fact that AI would inevitably reduce their corporate workforce over the next few years. Wait, you genuinely view that broad announcement of future layoffs as a positive model of transparency? I do. And let me offer an analogy to explain why.

[00:06:48] Think of it like a legal contract in a business partnership. Okay. A well-written contract does not guarantee you infinite wealth, and it doesn't guarantee the market won't collapse. What it does is bound reality. It tells you exactly where your liabilities start and stop. But we aren't talking about business liabilities. We're talking about people's livelihoods. I know, but a bounded, candid promise about workforce reductions is vastly superior to an unconditional reassurance that the CEO knows they can't keep. I disagree. Hear me out.

[00:07:15] If leadership says nobody will ever lose their job to AI, and then market dynamics shift six months later, that becomes a textbook psychological contract breach. Candor bounds the anxiety by drawing a box around the threat. But candor only bounds the anxiety if the organization is simultaneously bounding the threat to the individual worker's livelihood. How so? Look at the Sverk meta-analysis on job insecurity and organizational psychology.

[00:07:41] It proves that diffuse anxiety, this lingering sense of, you know, will I be here next year, destroys mental health, physical health, and organizational commitment far more than actual immediate structural change. So you're saying the anticipation is worse than the event. Exactly. When a CEO broadly announces that corporate headcounts will shrink over a few years, they might be technically candid, but they are flooding the entire enterprise with existential dread. Every single morning, every employee logs on wondering, am I the one on the chopping block today?

[00:08:09] Wait, let me make sure I understand how you are framing this. If the technology is fundamentally deflationary for certain types of labor, what is the alternative to candor? It's not about hiding the truth. Are you suggesting companies should just hide the structural shifts until the day the severance packages are handed out? No, I'm saying that transparency, webinized without material support, is just a threat. Take Accenture's September 2025 restructuring as a perfect mechanism of how this fails. Okay, the retraining initiative. Yeah.

[00:08:38] They announced the massive $865 million training program, very transparent, but they explicitly stated they were exiting staff who could not be retrained fast enough to meet new business needs. Which seems entirely fair. They funded the transition, and if an employee cannot or will not adapt to the new tools, the organization must have a procedural mechanism to move forward.

[00:08:59] But look at the mechanism of human behavior that creates. When you are completely transparent about classifying your human beings as either reskillable assets or expendable costs, you trigger a survivalist environment. A competitive environment, sure. A toxic one. Do you think people in that environment share their AI problems with their colleagues? Do you think they collaborate? Of course not. They hoard knowledge because they're suddenly graded on a curve for their professional survival. I see your point, but…

[00:09:26] That level of candor doesn't build trust. It signals that workers are just spreadsheet line items to be engineered away. But that is exactly why visible governance and worker voice are so critical. It's not just about dictating terms from the top down. It's about creating formal channels for procedural justice. Formal channels usually just mean suggestion boxes that nobody reads.

[00:09:46] Not always. Look at the December 2023 partnership between Microsoft and the AFL-CIO. This was a landmark agreement explicitly designed to incorporate worker perspectives into AI development and share technology trends with labor leaders before deployment. Yes, before deployment. Or consider Salesforce establishing an office of ethical and humane use, complete with an advisory council of frontline employees. Having a formal procedural channel aligns professional values and gives workers a mechanism to shape their own upscaling.

[00:10:15] I'm sorry, but I just don't buy that. Let me tell you why. You don't think it helps? Procedural channels, advisory councils, transparency notifications, ethics boards are overwhelmingly utilized as consultation theater. Employees are invited to a committee meeting to comment on AI deployments that the C-suite has already purchased, configured and scheduled for rollout. That's a pretty cynical view. It's a reality. They are given a voice only after the fundamental financial choices are locked in.

[00:10:41] But the Microsoft partnership explicitly aims to incorporate perspectives during the development phase, not just post purchase. Even so, compare that voice to what real material protection actually looks like. Let's look at the Writers Guild of America negotiations spanning 2023 to 2026. Ah, the WGA strikes. Right. The writers didn't shut down Hollywood for months just to get a transparency report or a seat on an advisory council to discuss how they would be replaced.

[00:11:08] They recognized an existential threat to their professional identity and their craft. True. They fought hard. And they secured hard material rules. AI cannot write or rewrite literary material. Writers cannot be forced to use AI software by the studios. Training data requires explicit notice and compensation. Which are very rigid boundaries. But necessary ones. They didn't settle for procedural justice where the studios just promised to talk to them before using algorithms. They demanded and won distributive justice.

[00:11:37] They forced the material protection of their labor into reality. I hear that. But those material protections only exist because they forced a procedural framework, like a collectively bargained legal contract into existence. Yes, a contract. You cannot have those hard material rules without the governance structure to define and enforce them. And if you lack clear policies, you get the dark side of AI adoption. Which brings us back to the statistic I opened the show with. If you don't define the rules of the game, you risk operational collapse.

[00:12:05] By operational collapse, you mean the degradation of output quality. Exactly. The KPMG data highlights the mechanism beautifully. Right now, 66% of employees rely on AI output without checking its accuracy. Which is terrifying for quality control. It is. 56% report making mistakes in their work because of AI hallucinations. And 57% actively hide their AI use from their employers.

[00:12:30] You simply cannot drive high-quality AI adoption in an environment of secrecy and shadow IT. Because nobody knows what's actually happening. Right. Field evidence like the 2025 Erik Brynjolfsson study shows a 14% productivity bump when generative AI is used correctly, especially for less experienced workers. But how do you unlock that? By giving them the tools. By having structure. Generative AI requires iterative prompting, admitting when the output is flawed, and sharing those effective prompts with your team.

[00:12:58] You only get that productivity if you have clear policies and psychological safety. But let's unpack the causality there. Why are they hiding it? Employees conceal their AI use, and they blindly accept uncritical outputs because they understand the unwritten rules of the game perfectly. Meaning? If you are a mid-level analyst, and you discover a way to use AI to do your 40-hour week in 15 hours, what happens if you tell your boss? Do you get paid the same for 15 hours? Probably not. No.

[00:13:25] The company harvests those 25 hours of efficiencies to eliminate your colleagues' jobs, and then they double your workload. The concealment is a highly rational survival strategy in the absence of a material guarantee that the gains will be shared. Which circles back to my core thesis. You must explicitly write down the deal. The company has to state formally, if you use this tool to save 10 hours a week, we will not fire you. We will redirect your capacity. Writing it down doesn't matter if the enacted corporate strategy fundamentally contradicts it.

[00:13:56] Let me contrast two real-world enacted strategies to show you how this causality actually plays out on the balance sheet. Let's hear it. Let's look at Klarna versus the Inca Group, which operates IKEA. Oh, this is the perfect juxtaposition for this debate. It really is. So in February 2024, Klarna aggressively boasted to investors that their new AI assistant had replaced 700 human customer service agents. I remember that headline. They froze hiring and let attrition shrink the workforce.

[00:14:22] It was a pure, unadulterated cost reduction play. They harvested the efficiency directly to the bottom line. And it is worth noting that the market initially rewarded them for that efficiency. Initially. But let's look at the mechanism of what happened next. By May 2025, Klarna CEO Sebastian C. Matowski had to publicly reverse course. Oh, really? Yeah, they had to start scrambling to hire human agents again. Why? Because the quality of customer service had severely degraded.

[00:14:49] AI hallucinations and a lack of nuanced human empathy and complex edge cases ruined consumer trust. So they pushed too hard, too fast. They harvested the short-term productivity, but destroyed the long-term quality. Because they viewed human workers purely as a depreciating cost. So how did Inca Group handle the exact same technological capability differently? It is a master class in enacted values. Inca Group deployed an AI bot named Billy.

[00:15:16] Billy was incredibly effective and quickly handled nearly half of all routine customer queries. Things like, you know, where's my package or what are your store hours? Very standard automation. Right. But instead of firing the call center staff who used to answer those routine questions, they took those 8,500 workers and put them through a massive logistical retraining pipeline to become remote interior design advisors. Wow, 8,500 people. Imagine being a call center rep, accustomed to answering rote questions about lost Allen wrenches,

[00:15:46] and suddenly your company says, the bot takes that now. We are investing in you to do high-end remote room design for our customers. That is a massive shift in human capital allocation. And look at the material result. They reinvested the time saved by the AI into a higher value, distinctly human service. That remote design service went on to generate 1.3 billion euros. That's incredible. It made up over 3% of their total global sales.

[00:16:13] If you want quality AI adoption, you must redeploy before you reduce. The AI values test is passed through structural reinvestment, not just a well-written memo outlining your obsolescence. I will readily concede that Inca Group's redeployment model is the absolute gold standard for how to handle the capacity freed up by automation. It really is. But I would argue that INCA's success was fundamentally driven by procedural clarity. They didn't just throw money at the wall. They mapped the workflow meticulously.

[00:16:42] They established a clear procedural reskilling pathway, and they communicated the transition timeline transparently to those 8,500 workers. Of course, they had to have a plan. Without that rigorous operational blueprint and written commitment, a desire to reinvest is just empty sentiment. It requires both, certainly. But a beautiful procedural blueprint without the heavy financial commitment to actually retain and pay those 8,500 workers during their training period, well, that would have just resulted in 8,500 layoffs. Fair point.

[00:17:11] The material outcome, keeping people securely employed in higher value, more meaningful roles, is what proved their AI values to the workforce. As we synthesize these two perspectives, looking at the landscape Westover describes, my position holds that establishing clear terms, robust governance structures, and systematically removing ambiguity is the absolute prerequisite for navigating this era. It's a starting point, sure. If you do not explain the strategy before the memo, if you do not establish feedback loops,

[00:17:40] and if you do not define the boundaries of the new psychological contract in writing, employees will inevitably fill the silence with their worst fears. A fair, transparent process is the foundation upon which organizational trust is built. And my position remains that procedural clarity is completely hollow without distributive justice. An impeccably written policy that efficiently and politely manages a worker's obsolescence is not going to win their loyalty. No, probably not.

[00:18:05] The AI values test is ultimately graded on whether an organization enacts a strategy of reinvestment. Are they sharing the productivity gains with the people generating them? Right. Are they protecting human quality and meaning like the WGA fought for? Or are they just harvesting productivity for short-term cost reductions like Klarna attempted? Yet, despite our differing emphasis on what drives trust, we are converging on a crucial undeniable reality.

[00:18:31] AI adoption has decisively evolved from a technical implementation puzzle into a profound test of employer character. It really has. Leaders who view this purely as an efficiency exercise, who obsess over API integration and deployment speed without mapping the psychological and material impact on the human beings doing the work, well, they are walking blindly into a massive talent attrition crisis. Absolutely. When nearly 60% of highly skilled people will reject an employer over a misalignment in AI values,

[00:18:58] the stakes are nothing less than existential for the business. It forces us to reconsider the metaphor we started with. We are in the middle of a massive structural renovation of the modern workplace. The landlord is absolutely moving the walls. Without a doubt. The question is whether organizations will try to sneak the construction crews in during the night and hope nobody notices the shrinking floor plan, or whether they will sit down with the workforce, lay out the blueprints, and negotiate a new lease that fundamentally respects the people living inside. That's the real test.

[00:19:28] If you are listening to this, whether you are a manager trying to implement these tools or an employee trying to navigate them, we leave the ultimate conclusion of what constitutes a fair deal to you. Pay attention to the signals your own organizations are sending. Thank you for joining us on the debate. Thank you for joining us on the debate.