The Stump Grinder and the Singularity: Why AI's Biggest Risk Isn't What You Think

School drop-off. Another dad. The AGI timeline conversation I've now had seventeen times this year.

His forecast: AGI by 2031. Superintelligence by 2033. Every transaction transparent, audits obsolete, knowledge workers redundant. He'd clearly done the math. Had the Kurzweil curves memorized. Could articulate the end state in extraordinary detail.

So I asked: "What are you going to do about it?"

His answer: Buy a stump grinder. A chainsaw, a truck, an entirely physical business machines can't touch.

That's when I realized we're having the wrong conversation about AI disruption.

The Two Camps

I keep seeing this split—not just at school drop-off, but in every client conversation, every conference panel, every strategic planning session I'm part of.

Camp 1 is planning for the singularity. They've read the DeepMind papers. They can walk you through transformer architectures and emergent capabilities. They're intellectually fascinating. They describe 2033 in vivid, specific detail. But ask them what to do Thursday and they either shrug or reach for a chainsaw.

Camp 2 is planning for Thursday. They know AI is transformative—they're not in denial. But they're asking a fundamentally different question: What do we build in the 5-7 years where AI is extraordinary but not omniscient? Where human judgment still matters in human situations? Where transparency doesn't eliminate accountability and someone still has to stand behind the decision?

Here's what I've noticed after 18 months of having this conversation: Camp 1 is mostly people who work on AI. Camp 2 is mostly people who are responsible for humans.

The People With Thursday Problems

The managing partner who has to explain the talent strategy to 200 partners next quarter. The CFO who has to sign off on next year's hiring plan knowing the skills map is shifting monthly. The L&D leader who has to build a training program that might be obsolete—or might be the only thing keeping the firm competitive.

These people can't plan for the singularity. They have a board meeting next month.

I'm in Camp 2. Not because I think Camp 1 is wrong about the trajectory. I've watched enough disruption cycles to know the technologists often underestimate the speed of change, then overestimate the completeness of transformation.

But I can't build strategy for a future nobody can predict. I can only build for the transition that's already happening.

What the Last Disruption Cycle Taught Us

I was doing cybersecurity consulting when the internet went from "that thing researchers use" to "that thing every business depends on" between 1995 and 2001. Six years.

The companies that survived weren't the ones who correctly predicted 2001 while standing in 1995. They were the ones who solved 1996, then 1997, then 1998. They asked Thursday questions: How do we secure customer data when we're moving it online? What skills do our people need next quarter? Where do judgment and automation intersect today, not theoretically?

The companies that failed? They either denied the internet mattered (Camp Zero, I suppose) or they jumped straight to "everything will be different" without building the bridge.

The middle years are where businesses are won and lost. The end state takes care of itself. The transition requires deliberate design.

The Problem Nobody's Naming

Here's the uncomfortable part: AI is eliminating the work that accidentally taught people how to be leaders.

The partner-track path in most professional services firms has looked the same for decades. You start doing the grunt work—the detailed analysis, the research memos, the draft documents that senior people red-line into competence. You learn how deals actually work. How clients actually think. Where theory meets reality and compromises get made.

Somewhere in year 3-7, you stop being supervised on every decision. You start supervising others. You've built judgment through repetition—thousands of small decisions that taught you the patterns.

AI is compressing that timeline. The grunt work is increasingly automated or augmented to the point where one person does what three people used to do. Which sounds like pure efficiency until you ask the second-order question: Where does the next generation develop judgment if the judgment-building work disappears?

I asked a managing partner at a Big Four firm this question last month. Long pause. Then: "We're assuming AI will handle that somehow."

That's not a plan. That's a hope.

The 5-7 Year Window

If Camp 1 is right—if we get AGI by 2031 and superintelligence by 2033—this problem solves itself or becomes irrelevant. Maybe AI handles all the judgment. Maybe we're all buying stump grinders.

But if Camp 1 is wrong, or if the timeline stretches, or if AI becomes extraordinary but not omniscient, we'll have a leadership vacuum in professional services that no amount of technical capability fixes.

We'll have firms full of people who know how to use AI but don't know how to lead humans. Who can generate the analysis but can't read the room. Who optimize for the algorithm but miss the political dynamic that kills the deal.

This isn't hypothetical. I'm watching it start. The 2-3 year associate who's never built a financial model from scratch because AI does it faster—but also doesn't understand why the model breaks under certain assumptions. The consultant who can generate the deck but can't facilitate the hard conversation with the C-suite. The auditor who can run the exception report but can't explain why this particular exception matters and that one doesn't.

These aren't skill gaps. They're judgment gaps. And judgment comes from repetition we're systematically eliminating.

What Thursday Looks Like

So what's the Thursday solution?

I'm working with three firms right now on versions of the same answer: Deliberate apprenticeship models that replace what we're losing accidentally.

One firm is running "decision clinics" where senior partners walk through messy client situations—not the success stories, the ambiguous middle—and junior people learn the judgment process out loud. The stuff that used to happen implicitly during document review.

Another is creating "AI co-pilot debriefs." After using AI to generate analysis, the team reviews it together: What did AI get right? Where did it hallucinate? What context did we have to add? What would a client misunderstand? It's repetition with reflection instead of just repetition.

A third is redesigning the partner track entirely. Less time doing repetitive analysis (AI handles that). More time in client meetings earlier (where judgment actually develops). More explicit mentorship on the soft skills that AI can't replicate and firms can't afford to lose.

Are these permanent solutions? No idea. They're Thursday solutions. They work for the transition we're in, not the end state we're theorizing.

The Uncomfortable Question

Here's what keeps me up: What if we're wrong about what AI can't do, but we're also wrong about the timeline?

What if AI gets good enough to eliminate most of the grunt work by 2026, but doesn't achieve full AGI until 2040? We'll have spent 14 years not building judgment in humans because we assumed machines would handle it. Then we'll have a generation of professionals who can't do what AI automates and never learned what AI can't replace.

That's the gap between the singularity and Thursday. That's the risk nobody's pricing in.

Camp 1 isn't wrong to think about the end state. But Camp 2 can't wait for certainty. The firms that figure out how to build judgment deliberately—instead of assuming it happens accidentally—will have a massive advantage in the transition years.

And if the singularity arrives early and makes all this obsolete? Great. We'll have built leadership capability in our people that turned out to be unnecessary.

I can live with that downside.

What to Do Monday Morning

If you're responsible for developing talent in a professional services environment, here are the Thursday questions:

What work are we eliminating or automating that used to teach judgment? Be specific. Not "analysis work" but "the monthly variance report that taught associates how operational decisions show up in financial results."

Where are we assuming AI will teach people things it actually can't? AI can show patterns. It can't explain why this client operates differently from the pattern. That requires human transfer of institutional knowledge.

What does deliberate apprenticeship look like in our firm? If we can't rely on judgment developing accidentally through repetition, what's the intentional process?

Who owns this problem? L&D can't solve it alone. Partners have to be part of the solution, which means time and incentives need to align.

These aren't comfortable questions. They don't have clean answers. They require investment in something that might become obsolete.

But they're Thursday questions. And Thursday is when the board meeting happens, when the hiring plan gets approved, when the talent strategy gets locked in for next year.

The singularity can wait.

Although I did check—you can rent a stump grinder for $200 a day. You know, just in case.

Jay Schulman helps professional services firms navigate technology disruption without losing what makes them valuable. If your firm is wrestling with how AI changes talent development, reach out—I've probably seen your version of this problem twice already.