The AI Leverage Ladder: Where You Sit in the AI Value Chain Determines Your Career Trajectory

The AI Leverage Ladder: Four Rungs That Decide Your Next Career Move

TLDR: Goldman Sachs now uses AI to produce 95 percent of an IPO prospectus in minutes, and the remaining 5 percent, judgment, strategy, regulatory nuance, is where all the value lives. A four-rung ladder framework helps you assess whether your current role is being commoditized or compounding in value, but the real danger is that using AI itself erodes the deep thinking you need to climb.

Summary:

Here is a piece that lands squarely at the intersection of career strategy and AI adoption, and it does so with a genuinely useful mental model. The author, Kamil, introduces what he calls the AI Leverage Ladder, a four-tier framework for understanding where your professional work sits relative to AI capabilities. At the bottom, Rung 1, you have Execution: producing outputs that AI is rapidly learning to generate. First drafts, data entry, report generation. The numbers here are sobering. Entry-level hiring dropped 73 percent between 2023 and 2025, and US programmer employment fell 27.5 percent in the same window. If you spend your day generating things AI generates faster, you are sitting on a shrinking platform.

Rung 2 is Validation, the human-in-the-loop role where you check, edit, and approve AI outputs. About 76 percent of enterprises have explicitly built these roles to catch hallucinations. It feels safe, but it is fundamentally reactive. As models improve and hallucination rates drop, the volume of work that justifies a dedicated validator shrinks. Rung 3 is Direction, where you decide what AI works on and how it should work. You scope problems, define success criteria, and choose which tasks deserve AI involvement at all. UC Berkeley research found that the most successful AI-adopting firms were the ones where people understood the problem deeply, not the technology. Domain expertise becomes the moat. Rung 4 is Architecture, designing the systems AI operates within: data pipelines, governance standards, AI product strategy, ethical guardrails. The Harvey AI example is telling. A first-year litigation associate repositioned from drafting contracts to building the system that drafts contracts. That company is now valued at 8 billion dollars.

But here is where the article gets genuinely interesting and where most career advice pieces stop short. The author confronts a paradox that almost nobody in the "learn AI to stay relevant" crowd wants to talk about. MIT Media Lab researchers found that participants using ChatGPT showed a 47 percent drop in neural connectivity compared to unaided writers, and 83 percent could not recall key points from their own AI-assisted essays minutes after writing them. Microsoft Research confirmed the pattern: higher confidence in AI leads to less critical thinking, which erodes self-confidence, which drives even greater AI reliance. The author calls it a loop that pulls you down the ladder while you think you are climbing it. A BCG and Harvard study of 758 consultants quantified the cost. For tasks outside AI's capability range, consultants who trusted AI performed 19 percentage points worse than those working without it.

Now, let me push back on a few things the article avoids thinking about. First, the ladder metaphor implies a clean hierarchy, but in reality, many roles blend multiple rungs simultaneously. A senior engineer might architect systems in the morning and validate AI output in the afternoon. The framework is useful for self-assessment, but it risks oversimplifying the messiness of actual work. Second, the article focuses entirely on individual repositioning and says nothing about organizational dynamics. What happens when entire teams try to climb to Rung 3 and 4 simultaneously? Not everyone can be the architect. There is an implicit zero-sum dimension here that goes unaddressed. Third, the cognitive debt research is compelling, but the article presents it as a universal risk without distinguishing between types of AI usage. Using AI for boilerplate versus using it for novel reasoning are very different activities with likely different cognitive impacts. The blanket warning, while directionally correct, lacks nuance.

For architects and team leads, the practical takeaway is clear. When you evaluate your team's workflow, map each recurring task to a rung. If most of the team's time sits at Rung 1 and 2, you have a structural vulnerability. The move is not to automate those tasks and call it a day. The move is to redesign roles so that people spend proportionally more time on direction and system design. And critically, build in deliberate practice time where team members work without AI assistance, specifically to maintain the domain expertise that makes their AI-assisted work valuable.

Key takeaways:

• The 95/5 split at Goldman Sachs illustrates a broader pattern: AI commoditizes output generation while judgment and strategic framing become more valuable
• Workers with AI skills command a 56 percent wage premium, double from the previous year, but the premium comes from proximity to AI's inputs, not its outputs
• Cognitive debt from AI overuse is a real and measured phenomenon, with a 47 percent drop in neural connectivity and a self-reinforcing cycle of declining critical thinking
• The most valuable career move is not learning to prompt better, it is repositioning your role one rung up from execution toward direction or architecture
• Entry-level roles are being compressed fastest, with a 73 percent drop in hiring at the execution level

Tradeoffs:

• Gain speed and volume through AI delegation but sacrifice the deep thinking practice that builds domain expertise
• Gain safety in the validation role but sacrifice long-term leverage as model accuracy improves
• Gain strategic positioning at Rung 3 and 4 but sacrifice the comfort of well-defined, repeatable tasks

The AI Leverage Ladder: Four Rungs That Decide Your Next Career Move