Artificial intelligence is rapidly altering job content faster than traditional headcount planning can accommodate, leaving the vast majority of organizations unprepared to integrate these technological shifts into their enterprise strategies. Strategic workforce planning must pivot from rigid job-based metrics toward dynamic, task-oriented frameworks that deconstruct roles into specific duties and workflows. Rather than rushing into premature downsizing that often damages institutional knowledge and employee trust, leaders should intentionally evaluate productivity gains through value capture decisions, internal mobility pathways, and scalable reskilling initiatives. Effective adaptation requires organizations to abandon obsolete annual forecasts in favor of continuous, cross-functional planning that aligns financial goals with human capital preservation. By establishing clear transition principles and fostering psychological safety, enterprises can successfully navigate technological disruption while safeguarding both operational quality and workforce stability.

Powered by the WRKdefined Podcast Network. 

[00:00:00] Welcome to The Debate. You know, imagine trying to map a coastline during a continuous rolling earthquake. That sounds like a complete nightmare. It really is. A landmass is literally shifting under your feet. And by the time you print the map, the bays have turned into peninsulas. And that is exactly what human resource departments are doing right now. Oh, absolutely. They are using these static annual headcount maps that plan for an environment that is being, well, violently reshaped by AI.

[00:00:27] Right. And the gap between what's happening on the ground and, you know, what is written on the financial ledger is becoming completely impossible to ignore. We are looking closely at Jon Westover's November 2026 analysis today, which basically points out that traditional headcount forecasting is just it's collapsing under its own weight. Yeah, but because the AI is changing the actual content of the jobs faster than the organization can track. We aren't just talking about generative AI anymore. We are talking about agentic AI.

[00:00:55] Right. And we should probably clarify what that actually looks like on the floor. I mean, generative AI is when you ask a chatbot to write a draft of an email. Simple enough. But agentic AI is when the system doesn't just generate text, but actually takes actions. It evaluates the email, accesses your CRM, updates a client's billing file, and triggers a refund. All completely independently.

[00:01:17] Exactly. All independently without you clicking a single button. So when systems can execute entire workflows on their own, forecasting how many quote unquote people you need in a department gets incredibly messy. And the numbers backing this up are staggering. The conference board surveyed nearly 800 international leaders late in 2026. A massive 70% of them fully expected job and workflow redesign to be their most significant AI action.

[00:01:42] So they know the earthquake is happening. They do. Yet only about one in 10 had actually built those effects into their enterprise-wide plans. There is a profound cognitive dissonance here. And that is where our perspectives really diverge today on how to solve this. Absolutely. Because while we both agree the old annual budget model is broken, the proposed solution in a lot of this literature is that organizations need to adopt this hyper-granular continuous task-level micro-planning. Which they do.

[00:02:10] Well, they want to deconstruct every job, track every micro-shift in how work is done, and manage this utopian large-scale internal redeployment. To me, that is an idealistic bureaucratic overcorrection. An overcorrection? Yeah. It ignores how unpredictable AI actually is, and it ignores the pragmatic reality that sometimes automation simply means you need fewer people. Well, I fundamentally disagree with that fatalistic view of the workforce. Organizations have to shift to continuous task-level planning.

[00:02:39] If you don't aggressively reinvest those AI productivity gains into internal mobility and redeployment, you end up blindly slashing headcount. Which happens all the time. Right, but it's demonstrably destructed the long-term value. The fundamental unit of organizational planning, the job, is dead. Dead? I mean, that seems like a massive leap. It is, but it's a necessary one. Think about how a traditional finance department looks at a budget year. They look at the customer service department and ask, you know, how many full-time equivalents do we need?

[00:03:08] Sure. They look at the headcount. Right. They treat the job as a monolith. But we have to adopt the economic task-based view of technological change. This goes back to research by David Otter and his colleagues in 2003. A job is not a single solid thing. It is a bundle of tasks. I see why you think that, but let me give you a different perspective on how value is actually created here. Wait, before you go further, what does that actually mean for a normal manager? Are you saying my job as a marketing director isn't a job anymore?

[00:03:35] Exactly. Your job is a collection of 50 different responsibilities. Let's go back to the customer service rep. When you deploy an AI agent, it doesn't eliminate the representative entirely. It just handles the easy stuff. Right. It automates the highly predictable task of pulling up a customer's billing history. But simultaneously, it might actually increase the value of the unpredictable, deeply human task of de-escalating a frustrated client who received the wrong product. Okay. I track with that. So if you don't deconstruct the work into tasks, you can't capture that value.

[00:04:05] You just fire 20% of your reps to save money, and two years later, you have to rehire them at a premium because customer satisfaction completely tanked. I come at it from a different way. I mean, deconstructing work into individual tasks sounds academically flawless. In a sterile laboratory, sure. Break every role into 50 components, assign a probability of automation to each, and map it out. It doesn't have to be a laboratory. But you cannot manage a living, breathing enterprise of 50,000 people like a laboratory.

[00:04:34] It assumes AI progresses in a straight, predictable line, allowing you to perfectly forecast which tasks will vanish and which will remain. The technology is progressing remarkably fast, though. I mean, the capabilities are constantly expanding. They are expanding, but they are incredibly uneven.

[00:05:20] and produce significantly higher quality work. Which proves the productivity gain. Yes, but they deliberately chose another task that was just slightly outside that boundary. An adjacent task that looked similar but required a different kind of reasoning. On that task, the humans using AI were actually 19% less likely to get the correct answer than the consultants working completely unassisted. Because they trusted the AI too much. They stopped thinking. Right, they fell asleep at the wheel. But from a planning perspective, this is a nightmare. Because of this extreme unpredictability,

[00:05:49] this jagged frontier, building a massive continuous task-level tracking framework is, well, it's an exercise in fiction. I think that's a bit harsh. But think about it. You think a task is automated. You redeploy the worker and suddenly the AI hallucinates on an edge case and you have no human in the loop to catch it. This is why simple strategic hiring restraint is often far more effective. Does freeze hiring? Look at JPMorgan Chase. They didn't build a utopian redeployment matrix. They estimated a pragmatic 10% reduction in operations,

[00:06:18] largely managed naturally through attrition. It's realistic. I have to push back on the idea that task-level planning is academic fiction. We have real-world proof of feasibility right now. Look at Shopify. Their CEO, Toby Lutke, completely inverted the traditional headcount model to force this exact kind of task deconstruction. So what does that actually look like on the ground for them? I mean, practically. It's brilliantly simple. He forced teams to prove they couldn't get the work done using AI before they were allowed to request new headcount. Oh, wow.

[00:06:48] Yeah, think about that mechanism. A manager goes to HR and says, I need a new mid-level software developer. HR pushes back and says, show me the prompts you ran. Show me the exact workflows where Copilot failed to write this code. That's intense. It is, but that single friction point forces the manager to literally deconstruct the open role in a specific tasks just to justify the hire. It fundamentally shifts the corporate culture from job failing to task design. Well, okay, forcing a software engineering team at Shopify D's AI is very different

[00:07:18] from managing the entire workforce of a legacy company. Software developers are naturally inclined to test technological frontiers. Try applying that to a century-old manufacturing firm. Then let's look at legacy retail. Walmart has 2.1 million employees. They are successfully tracking declining, growing, and stable roles across a massive, complex global footprint. But are they actually tracking tasks or just broad job titles? They aren't just guessing. They are isolating tasks.

[00:07:46] If a floor manager used to spend three hours doing inventory checks and computer vision systems now do that instantly, Walmart is mapping that freed up time directly into targeting training for customer interaction. That sounds good on paper, but... Let me borrow an analogy from Peter Capelli's work on talent management. Think about supply chains. Okay, let's hear it. Decades ago, supply chains used a make-to-stock model. You forecast you'll need 10,000 blue sedans. You build them and you park them on a lot hoping customers buy them.

[00:08:14] That's traditional headcount forecasting. Right, very rigid. But supply chains evolved to become agile and demand-driven. They build based on real-time signals. Talent planning has to do the exact same thing. We have to map the tasks continuously so we can move the talent precisely to where the demand is actually appearing. I'm sorry, but I just don't buy that. Let me tell you why. The entire supply chain analogy assumes that human beings, with all their complex psychologies and localized knowledge, are as fungible as car parts.

[00:08:44] They aren't car parts, obviously, but the flow of demand. When standard, non-tech organizations attempt this hyper-focused, agile planning, it almost always collapses under its own bureaucratic weight. You think the tracking itself becomes the problem. I know it does. Have you looked at the UK Civil Service debacle from a few years ago? I know they struggle with their targets, but walk us through the mechanics of what went wrong there. It's the perfect example of this hubris. In early 2025, government ministers announced they were going to radically slash civil service jobs using AI.

[00:09:12] They had all these aspirational productivity models. Right, the political promises. Exactly. They were mapping out how AI would handle administrative tasks, aiming to shrink the workforce while simultaneously demanding that 10% of the remaining civil servants be transitioned into data roles within five years. They were trying to do exactly what you were suggesting, task-level tracking and mass redeployment. And the result? Total bloke. By late 2025, the UK Civil Service headcount hadn't shrunk. It had ballooned to 555,000 people,

[00:09:41] the highest level since 2006. Wow. Yeah. The National Audit Office literally had to step in and publish a review demanding that departments stop relying on aspirational AI modeling. When you try to track millions of micro-tasks across an enterprise, you lose the macro picture. So the tracking failed. Failed spectacularly. They actually ended up hiring more people just to manage the complexity of the AI integration and the task-tracking bureaucracy itself. Well, the UK Civil Service didn't fail because they deconstructed tasks. They failed because

[00:10:11] they treated theoretical AI capabilities as guaranteed budget savings before the technology was actually integrated and proven on the floor. But that's exactly what happens when you mandate this. Which actually reinforces my argument. You know, you have to track the tasks continuously based on pilot evidence, not political promises. But the sheer administrative overhead of tracking pilot evidence for 50,000 employees is staggering. It is incredibly difficult. It requires entirely new workforce intelligence systems

[00:10:40] and I can see that the software is still catching up to the theory. But if we don't do this, we run into a much more dangerous problem. Which is? Let's say you are right and we skip the micro-tracking. A department successfully automates 40% of its workload. You have all this newly freed human capacity. What do you do at that time? You let it naturally attrition or you write size. Right. Historically, if you don't deliberately capture it, it either leaks away into longer, useless email threads or management just defaults

[00:11:10] to workforce reduction. And sometimes workforce reduction is the mathematically correct answer for the survival of the business. That instinct to jump straight to reduction is toxic to long-term growth. When capacity is released, it must be deliberately reinvested. Always? Yes, always. Let's look at IKEA's parent company, Inca Group, because their execution here is a master class in value capture. They deployed an AI bot named Billy to handle routine call center queries. Over two years,

[00:11:40] Billy got incredibly good, eventually handling 47% of all inquiries. So nearly half the workload disappears. The traditional finance logic kicks in right there. Great! Fire 47% of the call center staff, book the savings, and boost the quarterly margin. Exactly, which is what many companies do. But Inca Group took a completely different path. They took 8,500 of those call center workers and initiated a massive redeployment mechanism. Wait, 8,500? That's huge.

[00:12:09] It is. And they didn't just tell them to figure it out. They ran aptitude tests, they provided rigorous training in spatial planning and 3D design software, and they transformed these workers into remote interior design advisors. Moving them from a cost center to a revenue generator. Exactly. And the financial results were incredible. That pivot generated 1.3 billion euros in new sales for their 2022 financial year alone. They literally created a new revenue stream out of thin air using people

[00:12:39] they were already paying. That is impressive. They plan to make remote interior design 10% of their total global sales by 2028. You cannot execute a pivot of that magnitude unless you have a utopian, deeply committed vision for internal redeployment. That is an amazing success story. I'll give you that. Contrast that with Klarna, the buy now, pay later company. They aggressively cut 700 customer service jobs, loudly bragging to investors that AI was doing the work of hundreds of humans. Right, the classic

[00:13:08] tech playbook. And what happened? Their CEO had to admit in 2025 that service quality degraded so significantly and wait times spiked so high that they were forced to resume hiring just to ensure customers could reach a human being when the AI inevitably failed on edge cases. Cost cutting without reinvestment destroys quality. That's a compelling argument, but have you considered how incredibly unique IKEA's market position is? In what way? IKEA's pivot required a perfectly aligned, highly profitable adjacent market, remote interior design,

[00:13:38] that allowed them to absorb 8,500 workers. The skill leap from talking about furniture on the phone to designing a room with furniture is manageable. It's still a significant upscale. Sure, but most companies do not have a billion euro adjacent market just waiting to be staffed by former customer service reps. You don't think other companies have untapped revenue streams? I think you can't easily turn an account's payable clerk into a cybersecurity analyst just because AI automated the invoices. Let's look at reality outside of retail.

[00:14:08] Okay, let's hear it. Amazon announced corporate cuts of 14,000 roles in October 2025. They cited the explicit need for fewer management layers in the age of AI. Or look at Accenture. That consulting firm. Right, and arguably the gold standard for corporate learning. They ran a six-month $865 million business optimization program. They rigorously upskilled 550,000 employees in generative AI fundamentals. Which is a massive investment. It is,

[00:14:37] but even with the best training infrastructure in the world, they still had to exit people for whom reskilling simply wasn't viable. But exiting people should be the absolute last resort, not the default lever you pull just because retraining is hard. I'm not saying it's the first lever, but the mandate that an organization must find a new internal role for every single worker whose tasks are automated creates a toxic uncertainty of its own. Toxic uncertainty? How? It forces middle managers to invent busy work just to keep people on the payroll.

[00:15:06] Pretending everyone can be redeployed ignores the very real fact that sometimes automation simply means you need fewer people to achieve the same or better output. But... I don't disagree that honesty is compassionate, but the data clearly shows that relying on external hiring while letting your internal talent walk out the door is a massive financial mistake. When you say internal mobility

[00:15:36] programs are failing, are you suggesting they aren't worth the investment at all? I'm suggesting that in practice, a lot of what we call internal mobility devolves into what I call training theater. Training theater? Yeah. You see companies rolling out these massive internal talent marketplaces. Schneider Electric built a famous one called the Open Talent Market. It sounds great on paper. They get tens of thousands of employees to enroll. And they learn new skills. Right. The HR software tracks the completion of all these short modular courses and hands out

[00:16:05] internal digital badges for AI fluency or data analytics. Which builds the skill base of the company. Does it? If those course completions don't map to actual newly funded job openings because the overall headcount is shrinking, it's just theater. It gives the illusion of workforce planning. I think it's better than nothing. Employees spend hours clicking through modules thinking it guarantees their safety while the actual number of available seats in the company slowly shrink. I'm not convinced by that line of reasoning because it completely ignores the severe long-term

[00:16:35] economic damage of hollowing out your workforce on the bottom up. Hollowing it out? Yes. We aren't just talking about individual career paths or whether someone gets a digital badge. We are talking about the macroeconomic survival of the corporate talent pipeline. How so? Let me point you to the August 2026 Stanford payroll data by Brynjolfsson, Chandar, and Chen. They bypassed the surveys and looked at the actual payroll and employment effects of AI across industries. Okay. They found a 19% relative decline in employment for young workers, specifically

[00:17:05] those aged 22 to 25 in AI-exposed fields. And crucially, that decline came mostly from reduced hiring, not from laying off existing young workers. So organizations are essentially closing the door behind them. They just stop hiring juniors. Exactly. Which makes sense financially in the short term. They are using agentic AI to automate all the routine entry level tasks, the drafting, the basic coding, the data entry. Which makes sense financially in the short term. But those tasks traditionally served as the training ground for junior employees to

[00:17:34] learn how the business actually works. If we use AI to simply hollow out our entry level tasks today just to save a few dollars on the budget, where do our experienced managers leaders come from in 2030? That is a very real structural problem. I'll admit that. You cannot prompt engineer five years of organizational wisdom into a brand new mid-level hire. True. And the research backs up how dangerous it is to try and buy that talent later. Matthew Bidwell's landmark 2011 research explicitly proved that external hires cost significantly

[00:18:04] more, yet receive lower performance evaluations for their first two years on the job, and are far more likely to quit compared to internal promotions. The data on that is solid. So if you don't build real, functioning internal mobility pathways to preserve that entry-level pipeline, you are literally burning your future leadership to heat your current quarterly earnings. I will actually concede that destroying the entry-level pipeline is a massive strategic risk. If no one is doing the junior work, no one ever learns how to supervise the AI that is doing the junior work. We agree on the pipeline issue.

[00:18:35] Glad we're on the same page there. But where we diverge is on the mechanism of solving it. You are arguing for a continuous microscopic optimization of tasks across the enterprise to force internal redeployment. I am arguing for what the literature calls a durable transition compact. That's an interesting point, though I would frame it differently. What does a durable transition compact actually look like in practice for a worker whose job is automated? It looks like Singapore's public service framework from 2026. They explicitly articulated a

[00:19:05] principle-based approach to this exact problem. They stated clearly that restructuring and exits are a last resort, but they did acknowledge them as a reality. So they didn't guarantee jobs. No. They prioritize reskilling, they fund learning accounts, but they don't guarantee everyone a new job if the work simply isn't there. So it's a polite way of executing layoffs. No. It's about being brutally honest about the limits of AI reskilling. A durable transition compact means saying to your workforce, we will give you the transparent data on which roles are growing and which are shrinking.

[00:19:35] We will fund your time to learn new skills. But with a caveat. Right. But if the technology eliminates the need for your division entirely, and there is no viable adjacent market for us to pivot into, we will provide you with clear, fair, heavily supported transition pathways to leave the organization. I see. Setting honest limits is vastly more effective than attempting to build a convoluted matrix of cross-functional micro planning that collapses the moment OpenAI drops a new, more advanced model. So you view the

[00:20:04] continuous task-level planning as simply too fragile to survive contact with reality. I view it as trying to write a symphony by managing every individual vibration of every cello string. It's way too granular to be useful at a macro level. The jagged frontier of AI means the technology is too unpredictable for this level of precision engineering. But if you don't tune the strings, the symphony sounds terrible, and right now AI is completely untuning the cello. Let's pull this together as we are coming to the end of our time.

[00:20:36] From my perspective, treating AI as just another cost-cutting tool to be applied to static job titles is a recipe for long-term organizational failure. While I... Hold on, just to summarize. The evidence shows that to survive the rapid evolution of generative and agentic AI, companies simply cannot rely on annual budgets. They must continuously deconstruct workflows, track skills at the task level, and deliberately reinvest productivity gains into robust internal redeployment pathways. If we don't, we lose our institutional knowledge and destroy the pipeline of future talent.

[00:21:06] Fair summary. And for my stance, while I completely agree that traditional once-a-year headcount models are obsolete, demanding continuous cross-functional microplanning at the task level is a recipe for organizational paralysis. The bureaucracy problem. Exactly. Recognizing the unpredictable, jagged nature of AI means accepting that sometimes efficiency actually means your workforce shrinks. Establishing clear, honest limits on re-skilling, maintaining strategic hiring restraint, and offering fair transition support is ultimately more compassionate and

[00:21:36] operationally viable than forcing artificial internal mobility where no real market exists. What is incredibly clear is that we have a strong point of convergence here. The once-a-year budget-driven headcount forecast is a relic of the past. Truly did. Leaders can no longer sit back and let AI wash over their departments. They have to make deliberate strategic choices about what to do with the productivity gains rather than just assuming it will automatically balance out on a spreadsheet. Absolutely. We remain divided on the execution, whether

[00:22:05] continuous task-level redeployment is a scalable reality for most companies or an academic ideal that only works for software companies and furniture giants with massive adjacent markets. Right. But the days of simply forecasting a flat job title are definitively over. These distinct frameworks, the microscopic optimization of tasks that I'm advocating for versus the pragmatic macro-level management of workforce flow that you are defending, they both aim to solve the exact same unprecedented technological disruption. And it is unprecedented. It really is.

[00:22:35] For those listening, the source material in the Human Capital Leadership Review holds even more intricate case studies, from AT&T's early reskilling efforts to IBM's internal HR automation, all of which explain the how and why behind these massive shifts. It's highly recommended reading. We leave it to you to evaluate which paradigm best fits the future of your own organization. It brings us back to where we started. The tectonic plates are shifting. The coastline is moving. The only guaranteed failure is relying on last year's map to navigate tomorrow's terrain.

[00:23:05] It's time to learn how to sail. Perfectly said. Thank you for joining us on the debate.