When AI Replaces Managers, What Gets Lost?

[warmly] Welcome to the show. This is The Human Equation: innovation, risk, and the cost of forgetting people. Before we jump in, if you like conversations about work, leadership, and AI that actually wrestle with the messy human bit, like, share, and subscribe. I'm Simon Carver, I'm here with Lachlan Reed, and today we're also welcoming a new voice to the podcast family: Jacques San Dimas. [warmly] Jacques, mate, genuinely glad you're here. We've been circling this topic for ages -- companies flattening org charts, trimming managers, plugging in AI tools, and calling it transformation. And sometimes it does feel a bit like watching someone strip parts off a trail bike to make it lighter, then act shocked when the brakes are gone. [calm] Thank you, my friends. It is a pleasure to be here. And I think that image is more accurate than many executives would like to admit. We are in a moment where organizations are pursuing speed, compression, and automation with extraordinary confidence. But they are often removing layers they do not fully understand. Not because those layers were glamorous... but because they carried memory, judgment, interpretation, and emotional stability. [curious] That phrase -- emotional stability -- is the one I want to grab. Because when people say "middle management," they usually mean meetings, approvals, bureaucracy, PowerPoint decks nobody wanted. They don't usually mean the human shock absorbers. Exactly. Historically, middle management was never just administrative drag. It translated strategy from executives into reality for teams. It escalated risk upward before it became catastrophe. It coached people through ambiguity. It absorbed pressure from above and prevented that pressure from becoming panic below. If you remove that layer and replace it with dashboards, automated workflows, and agentic systems, you may gain speed in the narrow mechanical sense. But you may also lose the human tissue that keeps the organization coherent. [skeptical] And that's the bit that gets sold as a feature, right? "Look, we've reduced friction." Yeah... but sometimes friction is just another name for a person stopping you from doing something daft. [chuckles] A speed bump feels annoying until you remember the school's next to it. [reflective] I keep coming back to one question. If an organization gets faster, but loses memory, trust, and judgment... is it actually improving? Or is it just moving quicker toward mistakes it can no longer recognize? [softly] That is the question. Because innovation is not automatically progress. A company can optimize itself into amnesia. It can become very efficient at forgetting why certain practices existed, why certain people mattered, why mentorship prevented failure, why context mattered more than output volume. In risk work, we see this often: systems appear stronger on paper right before they become more fragile in reality. Wait -- "optimize itself into amnesia." That's gonna stick with me. Because you can almost see the slide deck, can't you? Headcount down, cycle time down, approvals down. Lovely clean arrows. But nowhere on the slide is "we've just lost the only manager who knew why that client escalates every March" or "we've cut the person who trained half the team." [laughs lightly] Yes. The dashboard says green; the hallway says something else. And maybe the hallway disappears too, because now everybody's remote and the software is monitoring throughput instead of noticing who seems frightened. And this is where AI changes the texture of the problem. We are not only talking about traditional automation -- moving forms faster, routing tickets, reducing repetitive tasks. We are talking about systems that now score, recommend, monitor, summarize, escalate, and increasingly coordinate. So corporations begin to believe the machine is not merely assisting management. It is becoming management. [questioning tone] Becoming management -- not just helping with admin. That's a bigger claim. You're saying the software isn't just doing the paperwork; it's quietly taking authority? Yes. Authority, perception, and in some cases legitimacy. If an employee's performance score is generated by a model, if workflow priority is assigned by an algorithm, if staffing recommendations come from predictive analytics, then power has shifted. Perhaps not formally on the org chart. But functionally, very much so. [serious] And if nobody can really explain the chain of reasoning beneath that score or ranking, then the human being on the receiving end is no longer being managed by a person they can appeal to. They're being sorted. Precisely. And sorting is not leadership. [curious] Alright, let's really sit with that missing layer. Because I think a lot of listeners have been told, for years, that middle management was dead weight. So Jacques, if we strip away the jokes and the stereotypes, what was that layer actually doing when it was healthy? [reflective] When it was healthy, it performed at least four essential functions. First, translation: leadership says, "Here is the strategy," and somebody has to turn that abstraction into weekly decisions, tradeoffs, roles, deadlines. Second, buffering: executives create pressure, markets create pressure, clients create pressure -- someone has to metabolize that pressure so frontline workers are challenged, yes, but not psychologically shattered. Third, mentorship: people do not become capable simply because a knowledge base exists. They become capable because someone more experienced corrects them, encourages them, teaches them judgment. And fourth, escalation management: the middle layer catches conflict, confusion, and operational drift before they become enterprise-level failures. [questioning tone] I want to pull on the first one -- translation. Because strategy always sounds elegant at the top. "Increase customer responsiveness." "Adopt AI across workflows." "Do more with less." But somebody in the middle has to answer the ugly practical questions like, okay, which team loses budget, who approves exceptions, what happens when the tool is wrong? Yes, and also: what does success look like for this particular team, with these particular people, under these particular constraints? Strategy without translation is slogan. The middle layer turns slogan into operating reality. [matter-of-fact] And operating reality is always grubby. It's never clean. Someone's off sick, one team's carrying an old system, another team's got a manager who's brilliant technically and hopeless with people, the budget got cut, procurement's late... even a kangaroo could trip over it. That's why I get twitchy when people say, "The AI can coordinate it." Coordinate WHAT, exactly? [dryly] Usually a simplified version of reality, which is not the same thing as reality. Algorithms are very good at processing defined signals. They are not inherently good at interpreting the emotional, political, historical, and contextual complexity inside organizations. They do not smell fear in the room. They do not notice the silence after a meeting and understand that silence means resistance. They do not recognize that a high performer is deteriorating because of burnout rather than low commitment. That silence point -- I felt that in my chest. There was a team I worked with years ago, and on paper everything looked fine. Delivery was happening, tickets were moving, updates were polished. But there was this strange dead air in meetings. No disagreement. No pushback. Just instant compliance. And one manager -- one person -- saw it and said, "This isn't alignment, this is fear." Two months later, half the issues spilled out. People were confused, covering for each other, terrified to admit they were lost. If that manager hadn't named it, we'd have called it efficiency. [pauses] It wasn't efficiency. It was emotional gridlock. [softly] "This isn't alignment, this is fear." That's the sentence. Because if all you're tracking is output, those two can look identical for a while. And that is why invisible labor matters. Good managers often appear inefficient to executives who only value visible metrics. They spend time clarifying. They listen. They mediate. They repeat themselves. They coach someone through uncertainty for the fourth time. None of that looks elegant on a dashboard. But it prevents error, resignation, resentment, and what I would call hidden operational decay. [curious] So let me try to explain it back, and you tell me where I get this wrong. The corporate fantasy is: if management includes status updates, check-ins, reminders, approvals, and performance reviews, then software can absorb those pieces and the layer disappears. But the missing part is that those visible tasks were only the shell. The actual job was interpretation and relationship. Almost. [short pause] I would sharpen one word. Not only interpretation and relationship -- accountability. A healthy middle layer is where accountability becomes human. A senior leader can declare a target. A system can track whether the target was met. But a manager asks, "Why did this happen? What blocked you? What did we learn? What do we change now?" That is not clerical work. That is the construction of judgment. And judgment takes time. There's no shortcut there. It's like teaching someone to ride in sand versus bitumen. You can hand them a manual, sure, but until they wobble, overcorrect, nearly bin it, and someone talks them through the feel of it... they don't KNOW it. Beautifully said. Apprenticeship is not data transfer. It is guided exposure to complexity. Which means when companies compress management, they're not just cutting salary bands. They're cutting apprenticeship pathways. Yes. And later they will say, "Why do we have fewer leaders capable of sound judgment?" They will ask this with complete surprise, after dismantling the conditions under which judgment was developed. [serious] This is where it gets eerie for me. Because now we're not just talking about fewer managers. We're talking about algorithmic management -- AI-generated performance scores, behavioral productivity analytics, automated workflow surveillance, maybe even predictive workforce reduction. That's not hypothetical language anymore. That's how many firms describe the tooling. [skeptical] "Predictive workforce reduction" is one of those phrases that sounds clean until you actually hear it. What it means in plain English is: software helps decide who becomes disposable. [calm] Yes. And the danger begins with a seductive assumption: if something can be measured, it can be managed. But measurable is not the same as meaningful. A system can quantify response times, keyboard activity, ticket closure rates, schedule adherence, sentiment proxies. What it cannot easily quantify is whether a person is building trust, mentoring quietly, preserving team stability, or preventing mistakes before they appear. The phrase I keep hearing from companies is "objective visibility." They love that phrase. The idea that the dashboard sees more fairly than a biased human being. And to be fair, sometimes it can reveal patterns humans miss. I am not anti-measurement. I am anti-idolatry. If leaders begin worshipping what the dashboard can display, they will neglect what the dashboard cannot perceive. Then culture starts to contort itself around legibility. People perform for the metric. They avoid the risky, unmeasured work. They stop raising ambiguity because ambiguity slows scores. [responds quickly] "Culture contorts itself around legibility" -- that's nasty, in the accurate sense. Because then everyone learns the trick. Don't do the important weird stuff. Do the countable stuff. Close the ticket, don't solve the root problem. Look productive, don't BE useful. [laughs softly] The spreadsheet version of being a good person. [little chuckle] Exactly. And over time this damages four things very quickly. Psychological safety, because people feel watched rather than supported. Institutional memory, because experienced people leave and their tacit knowledge was never captured. Apprenticeship culture, because nobody has time to teach what is not immediately billable or measurable. And trust in leadership, because workers sense that decisions affecting their lives are being made by systems they cannot question. I want to grab the "four things" because they're not fluffy. Psychological safety, institutional memory, apprenticeship, trust. If you lose those four, even if your quarterly numbers look shiny, you've basically taken termite damage to the frame of the house. Yes, and this is what I mean by organizational debt. Not technical debt alone -- organizational debt. You gain efficiency now, but defer the cost. The invoice arrives later as attrition, disengagement, brittle teams, shallow succession pipelines, and failures that nobody can fully explain because the context is gone. [thoughtful] Debt is the right metaphor. Because the gain feels real. Headcount lower, reporting lines simpler, decisions faster. It's not fake. It's just incomplete. The system borrows against the future. And sometimes borrows against the people who are still there. I know folks who've worked under heavy monitoring tools, and the weird part wasn't even the privacy side first. It was the feeling that every minute had to prove itself. That's exhausting. You stop experimenting. You stop helping a teammate unless it's visible. You stop breathing naturally. The whole thing goes stiff. [softly] Yes. Surveillance changes behavior, and not always toward excellence. Often toward anxiety. Human beings under constant measurement do not become wiser. They become more cautious, more performative, and sometimes more deceptive. Wait -- more deceptive? Of course. If the system rewards what is visible, people optimize visibility. If it punishes delay, they hide uncertainty. If the model ranks them continuously, they learn to manage the signal instead of the work. This is not because workers are immoral. It is because systems teach behavior. A badly designed measurement system is a curriculum in disguise. [laughs once] A curriculum in disguise -- that's going in permanent marker on the shed wall. Because every metric is basically saying, "Children, today we learn what this company actually cares about." And if what the company actually cares about is legible output over durable learning, then you've built a culture that can look strong right up until the moment something weird happens. A client crisis, a system outage, a reputational hit -- whatever it is -- and suddenly nobody knows how to think without the script. That is the brittleness. Resilience is not efficiency under normal conditions. Resilience is adaptive capacity under abnormal conditions. Organizations forget this because normal conditions flatter automation. Abnormal conditions expose what only human judgment can do. [curious] So if we're not meant to just hand the keys to the dashboard, what does leadership actually look like from here? Because the hype machine says future leaders need to be more tech-savvy, more data-driven, more automated. That's all true to a point. But what sits underneath that? [reflective] Future leadership will absolutely require fluency in AI and systems. But not as worshippers -- as governors. Leaders will need AI governance, yes. They will need to understand how automated decisions are made, where the data comes from, where bias or blind spots may live. They will need resilience thinking: how does this system fail, how do we recover, who gets harmed first? They will need systems ethics: not merely can we automate this, but should we, and with what safeguards? And they will need operational psychology -- understanding how people behave under pressure, ambiguity, and surveillance. [questioning tone] "Governors, not worshippers" -- that's such a useful distinction. Because I think some leaders hear "AI literacy" and imagine prompt tricks, tooling stacks, productivity gains. You're talking about stewardship. Oversight with teeth. Yes. Human oversight may become one of the most valuable roles in the next economy. Not passive approval -- active oversight. The person who can question a system recommendation, understand second-order effects, protect context, and intervene before efficiency becomes damage. And that's not anti-technology. That's the thing people muddle. You can love the tools and still refuse to let them flatten every human judgment call. Same as loving a GPS but still noticing the bridge is flooded. [warmly] That's a beautiful analogy. The GPS can be brilliant. It can also route you into a lake if nobody's awake at the wheel. [laughs softly] Exactly. Human-centered automation is not slower by definition. Often it is wiser. You automate repetitive burden so people can think more clearly, not so they can be removed more quickly. You use systems to support judgment, not erase it. I think that's the competitive edge people keep missing. They assume advantage means replacing people faster than your competitors. But maybe the real edge is building a place where people can still learn, still challenge, still notice problems early, still make sense of the weird stuff. [firmly] I believe that is exactly the edge. In every era, organizations are tempted by seductive simplifications. In this era, the simplification is that intelligence can be operationalized without relationship, oversight, mentorship, or moral responsibility. That is a dangerous fantasy. The greatest organizational risk in the AI era may not be technical failure. It may be the slow erosion of human judgment disguised as progress. [softly] Disguised as progress... yeah. That's the haunting part. Not collapse with alarms. Erosion with applause. [reflective] And maybe that's the question to leave hanging in the air. When the next wave of transformation rolls through -- and it will -- are companies going to choose resilient humanity, which is messy and slower and annoyingly hard to score? Or seductive optimization, where the chart looks gorgeous right up until the culture snaps? Jacques, thank you. Really. This felt like the right conversation at the right time. [warmly] Thank you both. It was an honor. If this one resonated, share it with someone wrestling with leadership, AI, or organizational change. And if you liked the episode, subscribe -- it helps people find the show. [warmly] And please join us in welcoming Jacques San Dimas to The Human Workforce Podcast. We're glad you're here with us. Thanks for listening.