When AI Efficiency Becomes an Automation Arms Race

Lachlan Reed

[warmly] Welcome to the show! I’m Lachlan Reed with Simon Carver. Simon, I can’t get past this one ugly little image: a company saves a million bucks by replacing people with AI... and then wonders why fewer customers can afford to buy the thing. That’s like flogging your trail bike for petrol money and then acting shocked you’ve got no bike.

Simon Carver

[reflective] The “million bucks” part is exactly what gets me. Because on one spreadsheet, that looks like discipline, progress, leadership. But outside the spreadsheet, those wages were grocery money, rent money, school-shoes money. They were also demand. They were the thing keeping other businesses alive.

Lachlan Reed

Right. And that’s the trap. One of the papers we looked at lays it out pretty brutally: if AI displaces workers faster than the economy can reabsorb them, firms can automate their way to more productivity and LESS demand. The authors call it an automation arms race. Each firm grabs the savings, but the pain gets sprayed across everyone else.

Simon Carver

[questioning tone] The phrase “arms race” matters there. Because it means even smart executives can see the cliff and still keep driving. Not because they’re cartoon villains, but because if their rival cuts first, they feel they have to cut too. It’s not one bad decision. It’s a system teaching people to panic elegantly.

Lachlan Reed

[laughs softly] Panic elegantly is very corporate, mate. And there are real examples now. The paper points to more than 100,000 tech workers laid off in 2025, with AI named as a main driver in over half those cases. Customer support, ops, middle management—the bits everyone said were “safe enough” got whacked.

Simon Carver

That “100,000” sticks. I’m not going to forget that number. Because those aren’t abstract units. That’s 100,000 households adjusting dinner plans, cancelling subscriptions, delaying dentist appointments, maybe moving in with family. We talk about labor displacement like it ends at the office door. It doesn’t. It walks home.

Lachlan Reed

And there’s a bigger stat that should make leaders sit bolt upright: roughly 80% of U.S. workers hold jobs with tasks that large language models could touch in some way. Not necessarily erase the whole job, but reshape chunks of it. So if every firm goes, “Beauty, we’ll cut first and sort it later,” they’re chewing through their own customer base. Barking up the wrong tree, then chopping it down.

Simon Carver

[skeptical] Though I want to be careful here. This isn’t “technology bad.” History does show work gets reshaped, not just destroyed. New tasks appear. New occupations appear. The trouble is the timing. If the new work arrives too slowly, families can’t live on a theory.

Lachlan Reed

Exactly. That’s the difference. The hopeful version is transition. The dangerous version is a gap. And in that gap, people aren’t just workers—they’re consumers, parents, carers, neighbours. One of the books we read keeps hammering this home: people are not employee IDs on a spreadsheet. They’re the whole bloody system.

Simon Carver

I loved that reframing. Also the way it described “management by spreadsheet.” Because that’s the emotional trap, isn’t it? A layoff can look rational from 30,000 feet. But spreadsheets don’t see the kid whose dad now says no to soccer fees. They don’t see the nurse delaying car repairs. They don’t see that human beings are demand with faces.

Lachlan Reed

[calm] And weirdly—this is the bit that should worry capital just as much as labor—the paper says over-automation can hurt firm owners too. Not just workers. It’s deadweight loss. Everyone cops it. If demand drops enough, even the so-called winners end up standing in a shinier, emptier shop.

Simon Carver

That’s the whole moral and economic knot of it. Human dignity and consumer demand are not separate issues. They are the same issue viewed from two windows.

Simon Carver

[gently] So here’s the hopeful turn. The answer isn’t to freeze technology in place. It’s to design work so AI removes drudgery, not dignity. The most useful examples aren’t about replacing the person. They’re about clearing the desk so the person can finally do the part of the job that matters.

Lachlan Reed

Yep. In healthcare, for instance, the good use case is not “cheers, nurse, robot’s got it.” It’s AI doing the admin sludge—notes, scheduling, triage support—so the nurse can actually look at the patient. Be present. Same in education: automate the grading grunt work, free the teacher to mentor, explain, notice when a kid’s gone quiet.

Simon Carver

The “gone quiet” part is the point. A machine might flag performance. A teacher sees shame, boredom, fear, talent, spark. Those are different universes. And in finance, the hopeful example wasn’t some chest-thumping efficiency story. It was AI handling reconciliation and repetitive checking so humans could move toward advising, interpreting, building trust.

Lachlan Reed

[excited] City services too. That one really landed for me. Use AI to route complaints, model traffic, predict water pipe failures—great. But the bloke or woman doing the job becomes more human, not less. More time for community meetings, edge cases, helping the person who can’t navigate the app. The machine sorts; the human understands.

Simon Carver

And there’s a dignity piece inside that. When people say AI should “replace jobs,” they often mean the parts of jobs they personally find dull. But a lot of workers hear, “You are the dull part.” That’s why this has to be framed differently. Remove the busywork. Keep the judgment, empathy, creativity, care.

Lachlan Reed

Beautifully put. One of the books says this future is not the end of work, but the end of meaningless work. I reckon that’s the right target. Not zero humans. Better humans doing better bits of the job. If AI can take the copy-paste rubbish, let the person do the mentoring, the fixing, the inventive thinking.

Simon Carver

[reflective] Which puts a challenge in front of leaders and workers alike. Leaders have to stop using AI as a layoff shortcut and start using it as redesign. Workers have to ask, “What in my role is procedural, and what is deeply human?” That second question is the future.

Lachlan Reed

And policy matters too. The paper is pretty blunt that lots of popular fixes—UBI on its own, vague upskilling promises, hoping wages sort themselves out—don’t change the incentive to over-automate. The deeper fix is steering firms away from the race itself, while investing in real reabsorption and retraining.

Simon Carver

Which means the future of work isn’t just about invention. It’s about governance, patience, and values. Do we build systems that scale trust, learning, and meaning? Or systems that scale panic with prettier dashboards?

Lachlan Reed

[softly] That’s the question, isn’t it? If AI gives us back time, focus, and capability, do we spend that dividend on more human work—or just fewer humans? That’s the fork in the road. Anyway, Simon, that’s one to chew on.

Simon Carver

[warmly] It really is. Thanks for listening, everyone.