Enrich, Don’t Extract: Why Selling AI as a Productivity Tool Backfires, and How Leaders Can Introduce AI as a Path to More Meaningful Work

[00:00:00] Right Now, I mean, companies are spending billions on artificial intelligence, right? And the whole goal is to make their employees faster. Yeah, billions. It's everywhere. Right. But what if telling your team to quote-unquote get faster is the exact reason your AI investment is completely failing? It's a huge issue. It is. Because like, picture this. You walk into almost any corporate leadership offsite this autumn, right, in 2026, and you were going to hear the exact same mandate echoing off the walls. Oh, absolutely.

[00:00:30] They're all saying, we need our people to get more out of AI. And, you know, on the surface, that sounds totally reasonable. Well, yeah, it's the dominant pressure across basically every industry right now. I mean, you spend the massive capital on the technology, and the board naturally wants to see the return on that investment. Right. Exactly. But today, we are exploring this massive translation problem. Yes.

[00:00:52] We're looking at this new paper by Dr. Jonathan H. Westover from the October 2026 Human Capital Leadership Review. It's titled, Enrich, Don't Extract. It's a fantastic piece of research. It really is. And our mission for this deep dive is to figure out why that perfectly reasonable boardroom mandate sounds like a direct threat to the people actually doing the work. You know, the paralegals, the claims processors, the nurses. Right.

[00:01:18] So we're going to explore how the psychological framing of this tech is basically the secret variable making or breaking these massive rollouts. The gap in translation here is really something because when a CEO says, hey, let's increase output with AI, the employee on the ground hears something entirely different. Oh, completely different. Right. They hear, I am going to have to do more of the exact same work significantly faster with fewer coworkers. Yeah. And I will probably lose the parts of my job I actually enjoy doing.

[00:01:47] Yeah. I was thinking about it like it's like handing someone a state of the art power tool and saying, great news. You can dig the exact same ditch in half the time. And by the way, we are firing half the digging crew. Exactly. I mean, of course, they are going to pretend the tool is broken. That is such a brilliant way to look at it. And, you know, we need to establish right up front that this isn't just a matter of like internal communications or writing a better all hands email. Right. We are getting into organizational psychology here and fundamental job design.

[00:02:15] Every time a company rolls out a new technology, they are basically telling a story, a sense giving process about what the tool is actually for. Right. And the story being told right now is just stalling adoption entirely. It's totally stalling it. So let's get into why that adoption is stalling, because the divide between how leadership and employees experience this tech on a day to day basis is just wild.

[00:02:38] It really is wild. If you look at the recent data from that October 2026 Workday report, business leaders are operating with a very clear extraction mindset. Okay. 40% of them explicitly expect AI to squeeze more output from their current teams. And nearly a third, 28%, are actively banking on headcount reductions. Wow. So they see this purely as an efficiency play. Just do more with less. Yes, exactly.

[00:03:06] But you have to contrast that with the people they are actually managing. The Microsoft 2025 Work Trend Index highlights this massive, just glaring experience gap. Okay. Let's unpack this. What does that gap look like? Well, you have executives where nearly 80%, 79% to be exact, see AI as this incredible career accelerator. They are using it multiple times a week, 69% of them. They are having fun with it.

[00:03:33] But when you survey the global workforce, 80% report they don't even have the basic time or energy to just do their current jobs. Oh, man. That is a brutal disconnect. You have these exhausted workers barely keeping their heads above water being handed a mandate by highly enthusiastic executives who just assume the benefits are like obvious. Precisely.

[00:03:54] But I have to ask, I mean, are these leaders just sitting in their ivory towers plotting to squeeze everyone dry? Or is there something else driving this whole extraction mentality? Honestly, it is mostly driven by financial panic. Financial panic? Yeah. The reality on the ground is that very few CIOs are actually seeing transformational value from their AI investments so far. Gartner put out data this October showing it's only about 13%.

[00:04:22] Wait, only 13%? Only 13%. Wow. Okay. Well, that explains so much. Right. Think about the internal pressure that creates. You have convinced your board to spend millions on generative AI licenses and enterprise co-pilots. Yeah. And then the board asks for the ROI and you simply don't have it yet. That creates this intense top-down pressure for immediate measurable efficiency.

[00:04:48] Which I imagine trickles down into those notorious CEO memos we've all seen making the rounds online over the past year. You mean the ones mandating proof that work can't be done by AI before managers are even allowed to hire? Yeah, exactly. Like that Shopify memo or Duolingo announcing they were cutting contractors for work that AI could handle. Yeah, those are prime examples. When those memos go public, they create this template for everyone else.

[00:05:15] They cement this idea across the whole corporate world that AI is just a giant cost-cutting, headcount-reducing machine. But here is the crazy part about all that. When managers take that extraction pressure from the top and drop it onto their teams, it backfires spectacularly. Really? Oh yeah. Dr. Westover's paper highlights this brilliant natural experiment that captured this perfectly.

[00:05:37] Researchers at Stanford, led by Walsh, managed to observe two nearly identical divisions within the exact same law firm. Oh, that's perfect. Right. Both divisions got the exact same generative AI drafting tool at the exact same time. The only meaningful variable was how the two different managers pitched it to their teams. It is literally the ultimate test of framing. Interesting. So how did the first manager handle it? They used pure extraction framing.

[00:06:05] They basically told their team, hey, here is a tool to speed up turnaround times on document drafting. Which I mean, that sounds completely normal. That is how every software tool has been pitched for the last 20 years. It sounds totally normal. Yeah. But the result was dismal. The paralegals felt like cogs in a machine. Wow. They used the tool minimally. They were highly defensive about it. And they only used it for the exact specific tasks management had pointed out. Zero exploration.

[00:06:35] Okay. So what did the second manager do differently then? They took a totally different route. They used enrichment framing. Before they even brought up the software, they sat down with their team and asked a really simple question. Which was? They asked, what tasks do you find most tedious? Like, what do you wish you had time to get off your plate? Oh, I love that. That completely flips the dynamic. Now it is about solving the employee's problem, not just hitting the company's speed metric. Exactly.

[00:07:04] They positioned the AI as a way to clear away the drudgery. And the contrast in outcomes was staggering. That second division used the tool 58% more. 58%? Yeah. And they experimented with it 70% more. Because they weren't just trying to hit some arbitrary speed quota, they started applying the AI to higher value problems that the manager hadn't even considered. Like what kind of problems? Things like complex legal research or mapping out broad case analysis.

[00:07:31] You know, tasks that actually require deep human thought and synthesis. It makes me think of like a restaurant kitchen. It is the difference between telling a chef, use this industrial food processor to make 10 times more soup today. Versus saying, use this food processor so you don't have to chop onions by hand for three hours and you can finally have the time to design that new seasonal menu. What's fascinating here is that your analogy hits on the exact underlying psychological mechanism at play. Oh really? Yeah, it all comes down to self-determination theory.

[00:08:00] It basically states that humans have these fundamental psychological needs for autonomy and competence. Meaning we need to feel like we have control over how we work and that we are actually good at what we do. You hit the nail on the head. When you frame AI as a speed whip to just crank out more documents, you crush that intrinsic motivation. You strip away their autonomy entirely. Right. But when you frame it as a tool for job crafting, you are telling them, we trust your competence enough to let you redesign how you work. You invite them to engage.

[00:08:30] So if enrichment works so much better, what happens to the companies that stubbornly stick to the extraction method? Well. Because if people are just using it defensively, like in that first law firm division, it seems like a missed opportunity, sure, but maybe not a total disaster. Oh no, it is actively harmful to the business. Really? Absolutely. When you terrify your workforce with speed metrics and headcount threats, they just go underground.

[00:08:54] A KPMG study shows that 57% of employees are secretly using AI, hiding it entirely from their employers. Wait, almost 60%. That is a massive shadow workforce. But why hide it? I mean, if I found a way to do my job faster, I would want my boss to know I'm being super productive. Because they are terrified of the reward for that productivity. The Slack workforce index showed 48% of desk workers are incredibly uncomfortable telling their boss they use AI.

[00:09:24] Because they think they'll be punished. Exactly. They think, if I tell my manager this tool saved me four hours, I don't get to go home early. I just get my quota doubled. Or worse, they realize how easy parts of my job are and they lay me off. So they use the AI to do the work, submit it as if they did it manually, and just sit on the save time to protect themselves. That is the reality. And that fear-driven hidden usage leads directly to a massive drop in quality control.

[00:09:52] This is where we get the phenomenon of work slop. Here's where it gets really interesting, because I read about this. Work slop, that is a deeply unpleasant word, but it's so accurate. It really is. The BetterUp Labs study was talking about this, right? It's AI-generated content that looks incredibly polished at a quick glance. But when you actually read it, it is entirely devoid of substance. Just corporate fluff. Exactly.

[00:10:15] And because people are generating it secretly to meet these inflated productivity metrics, they aren't checking it carefully before hitting send. Right. And the study said 40% of workers are receiving this hollow AI-generated work from their colleagues, and they are spending almost two hours on average just cleaning it up. It is wild. Yeah. The tool designed to save time is actually creating two hours of cleanup work for the person downstream. Yeah. It is a classic example of Goodhart's law.

[00:10:43] When a measure becomes a target, it ceases to be a good measure. You see this everywhere companies try to implement strict usage metrics. Like what? Like AI token leaderboards. Amazon, Meta, Shopify, they've all struggled with variations of this when trying to track or gamify an adoption. A token leaderboard. Like ranking employees by how many AI prompts they use in a week? Believe it or not, yes. When leadership says, we are going to measure you by how much you use the AI, employees are smart.

[00:11:12] They will game the system. Of course they will. You get instances of people running AI on completely irrelevant tasks, just churning out prompts purely to hit the metric and avoid being flagged as a low performer. It is completely performative work. But I am stuck on how this plays out for the company as a whole. I mean, if nearly 60% of your people are hiding their AI use out of fear and the rest are just spamming prompts to game a leaderboard. Yeah. How can a company possibly measure actual ROI?

[00:11:40] How do you share a really brilliant prompt or a best practice with the rest of the team if everyone is hiding what they are doing? Doesn't extraction framing basically create a massive shadow IT nightmare? You can't. That is exactly it. It's the nightmare scenario for IT and management. You lose all visibility. Right. You can't govern data security. You can't figure out which workflows are actually improving.

[00:12:06] And you definitely can't calculate a return on investment when half your workforce is operating in the shadows. Okay. So extraction breeds fear, shadow IT, and work slop. We know it fails. The big question is, how do leaders actually pull off the enrichment model? Right. Dr. Westover's article lays out a whole playbook for this. And I want to speed through these steps because they are backed by some pretty serious corporate turnarounds. Definitely.

[00:12:34] The absolute first step is that you have to co-diagnose the work. You never start with the tool. You start with the job itself. Meaning you don't just hand them a login to a new software and say, figure it out. Right. You sit down and ask employees what they want to shed. Interestingly, workers actually want a huge chunk of their tasks automated. Really? Yeah. The Stanford WorkBank data shows they would happily hand over 46% of their tasks to AI if they could. Wait. Almost half.

[00:13:02] I would have assumed people were highly protective of their tasks. Just to, you know, justify their jobs. Not the drudgery. They want the tedious, repetitive stuff gone specifically so they can free up time for work that actually feels meaningful. Oh, okay. That makes sense. But management has to involve them in mapping out exactly what gets automated and what gets protected. Okay. So you identify the pain points together. What is the next step in the playbook? You have to resource the promise.

[00:13:29] If you tell people AI is going to enrich their jobs, you have to actually give them the resources to learn it safely. The paper brought up Moderna here, right? How did they handle this transition? Moderna gave their people protected exploration time. This is crucial. It was time completely divorced from output quotas. Wait. They were just paid to play around with the tool. Yes. They weren't asked to produce any actual work during this time. And because it was framed around creativity and solving their own problems, the adoption was explosive. Wow.

[00:13:58] Within two months, 40% of their active users had built 750 custom GPTs. That's amazing. And just to clarify for anyone who might not know, when we say custom GPTs in this context, we're talking about employees basically building their own mini apps, right? Exactly. Like feeding the AI a specific set of instructions to automate a very specific part of their daily workflow without needing to know a single line of code. That is exactly it.

[00:14:22] They automated their own pain points, but they only did that because they were given a safe sandbox to figure it out. Okay. Which leads to the third and probably most important step. You have to reinvest the freed capacity. Oh, this is the scary part for workers. If the AI saves my team 20 hours a week, what does the company actually do with those 20 hours? Ikea is the gold standard example of getting this right. Right.

[00:14:47] They rolled out an AI bot to handle routine customer service queries, checking order statuses, processing basic returns, that sort of thing. Which is usually the exact moment a company lays off the entire call center staff. But Ikea didn't. They took 8,500 of those customer service reps and completely retrained them. Completely retrained them. Completely retrained them. They turned them into remote interior design advisors. Oh, wow. So instead of just answering where's my couch, they are actively helping customers design their living rooms over video or chat. Right.

[00:15:16] It is a massive psychological leap for those workers. Absolutely. They took all that human capacity that was freed up by the AI and moved it up the value chain. And the financial result? That pivot generated 1.3 billion euros in 2022 alone. That is incredible. That's amazing. They didn't just save money on headcount. They generated massive new revenue by unleashing the human potential that was previously just like buried under routine tasks. And that ties perfectly into the fourth step.

[00:15:44] You have to prioritize value over volume. You cannot measure a modernized, AI-enabled workforce with industrial-era stopwatches. Right. Give me an example of how a company actually shifts that metric in the real world. Let's look at healthcare. The Permanente Medical Group and Mass General Brigham started rolling out ambient AI scribes for their doctors. An ambient scribe, meaning an AI that just listens in on the conversation between the doctor and the patient and automatically writes up the medical chart. Correct.

[00:16:11] It completely eliminates the need for the doctor to sit at a computer typing while you are trying to talk to them about your health. Oh, that's huge. And it saved something massive, like 15,791 hours of documentation time. Now, the extraction mindset would say, great, you save 15,000 hours, which means every single doctor can cram four more patients into their daily schedule. But they didn't do that. The metric of success wasn't patients seen per hour. The metric was reducing clinician burnout. And they succeeded. Wow.

[00:16:40] 84% of the doctors reported having better, more focused face-to-face communication with their patients. Because they could actually look them in the eye instead of staring at a screen. Which, as a patient, is all you really want anyway. Exactly. The true value of a doctor isn't in their typing speed. It is in their human care and diagnosis. The AI handled the typing so the human could focus on the care. Makes total sense. And the final piece of the playbook is governing with worker voice. You have to make it safe to disclose use. And that means collaborating with the workforce on the rules.

[00:17:10] The paper brought up the 2023 Writers Guild of America agreement here. Yeah, it is widely seen as the gold standard. Instead of management just dictating how AI would be used, they negotiated an agreement that preserved human discretion and authorship. Right. They established clear rules. Exactly. Rules around consent. But it still allowed writers to use AI as a supportive tool if they chose to. It was governance through collaboration, not just a top-down mandate. So what does this all mean? Because I hear all of this.

[00:17:39] Co-diagnose the work, protect exploration time, reinvest the capacity, shift the metrics, govern together. Yeah. But I have to offer a practical push back here. Sure. Giving employees, quote unquote, protected exploration time sounds amazing in a leadership seminar. But let's be real about how corporate life works. What happens when it's the end of the quarter, revenues are down, and deadlines are hitting hard?

[00:18:08] Doesn't that magical exploration time immediately get cannibalized by management demanding output? If we connect this to the bigger picture, you have hit on the exact reason why the research warns so heavily against what they call false enrichment. False enrichment. Yes. Employees are incredibly observant. If you promise them learning time, but the very first time pressure hits, you demand that their saved time be converted back into raw output. The trust is instantly broken. Instantly. They realize it was a trick all along.

[00:18:35] And they will immediately revert to defensive, hidden use. Exactly. The enrichment has to be structurally protected by the organization or it is just a nicely worded extraction mandate. It is a fragile ecosystem. You really have to prove to your team that the enrichment is real and that it is permanent. To make it stick, organizations really have to commit to three long term pillars. First, treat work design as an actual leadership discipline. Okay.

[00:19:03] Managers need to know how to redesign a job, not just enforce a quota. Second, build compounding learning systems. Create safe environments where teams can share their AI discoveries so knowledge actually scales across the company. Right. And third, ensure visible shared gains. Meaning if the company benefits financially from the AI efficiency, the employee has to tangibly feel that benefit in their day-to-day work life. Whether that's less burnout, more interesting work, or upskilling.

[00:19:32] If they don't visibly share in the gains, they will simply stop participating in the transition. It really brings us back to the core paradox of this entire deep dive. The absolute best way for an organization to get the massive productivity gains they are so desperately craving from AI. Is to stop demanding productivity. Exactly. Stop demanding productivity and start offering enrichment. You basically have to let go of the stopwatch if you want the speed.

[00:19:57] It runs counter to a century of traditional management thinking, but the data is proving it is the only way forward. It really is. And you know, I want to leave you with one final thought to mull over, building on everything we've talked about today. Oh, let's hear it. If we successfully implement this enrichment model, if we actually use AI to strip away all the tedious low-level drudgery tasks in our jobs so we can focus purely on high-value work, what happens to the entry-level workers?

[00:20:24] Oh, that is a looming question for the entire talent pipeline. Right. Because historically, doing that repetitive drudgery was exactly how you learned the ropes of your industry. Yeah, you had to put in the reps. That is how you internalize the patterns. It is how the junior lawyer learned to write a contract, or the junior coder learned to spot a bug. So, how do we train the next generation of experts when all the beginner tasks are fully automated? It's a huge challenge.

[00:20:50] If the AI is doing all the ditch digging, how does anyone ever learn how to use the power tool in the first place? That's something every single industry is going to have to figure out very soon. Thanks for joining us on this deep dive. We'll see you next time.