A 2023 Goldman Sachs report estimated that 44% of legal-sector tasks could be automated by generative AI. After that figure went public, some large law firms began scaling back their hiring plans for paralegals and first-year associates. Within a year or two, an unexpected symptom emerged: attorneys three to five years into their careers — accustomed to reviewing and revising AI-drafted work — reportedly started struggling to construct legal arguments from scratch.
Similar patterns are showing up across consulting, accounting, software development, and content strategy. The faster AI produces intermediate output, the more it erodes the training process people used to go through before that output existed. Organizations have gained efficiency, but the pipeline that carries expertise to the next generation has thinned.
Solo PMs and one-person businesses are the first to run into this problem.
What AI Eliminated Wasn't Grunt Work — It Was the Path to Mastery
Professional training used to have an invisible design built into it. Juniors spent long stretches doing tedious, repetitive tasks: research, number-crunching, drafting reports, compiling meeting notes. Most senior professionals remember that time as work they simply had to grind through — but two things were accumulating underneath it.
One was domain pattern recognition. Read the same type of contract two hundred times, and odd clauses start jumping out at you. Reconcile financial statements every week, and you begin to read the flow of numbers before you consciously analyze them. It's an instinct that can't be taught through explanation — only built through repeated exposure.
The other was friction with seniors. After being asked "why did you write it this way?" dozens of times, a junior learns, viscerally, what their senior actually cares about. The standard gets absorbed through experience before it's ever put into words.
Generative AI blocks both at once. When AI does the research, exposure to patterns shrinks. When the first draft already arrives polished, seniors lose the occasion to explain what's wrong with it. The feedback loop thins, and the apprenticeship relationship dissolves.
To Be Fair, Junior Grunt Work Wasn't Necessarily Good Training
This is where the counterargument needs to be stated fairly. Some researchers and organizational designers ask: was repetitive junior work ever really professional training — or are we romanticizing cheap labor?
Before photocopiers existed, juniors learned how workflows functioned by copying documents by hand. The photocopier eliminated that process, but it didn't collapse how expertise was developed. Before spreadsheets, financial models were calculated by hand. Spreadsheets took over that labor, and finance professionals kept getting trained anyway. From this view, AI is just the next technology in the same lineage.
There's real substance to this argument. Equating hardship with expertise is close to survivorship bias. If the same competence can be built faster, there's no reason to insist on the slower path.
But there's a difference. Photocopiers and spreadsheets replaced simple mechanical tasks. Generative AI produces output shaped like an intermediate judgment call — a draft, an analytical summary, a strategic recommendation. These arrive already looking like conclusions. If you only ever work by receiving and correcting that output, your editing skill grows, but your ability to construct reasoning from a blank page doesn't. The two capabilities aren't the same thing.
And the symptom now showing up across industries is exactly this. Since adopting AI, work has gotten faster — but the people who should be growing into senior roles have instead become people who only know how to edit.
Solo PMs and One-Person Businesses Feel This First
McKinsey and other researchers frame this as a corporate HR challenge. But for solo PMs, one-person businesses, and independent consultants, it's already a lived reality — well ahead of the corporate conversation.
Organizations have HR people whose job is to design training for the next generation. In a one-person operation, you have to sense for yourself whether your own capability is growing or stalling. And the more convenient AI gets, the harder that self-monitoring becomes.
I've found this question useful: "When was the last time I finished something from scratch, without AI?" If nothing comes to mind from the past month, your thinking may have stalled in that particular area.
It also helps to write a short answer of your own before handing a question to AI, then compare it against what AI produces and track exactly where the two diverge. That gap — accumulated over time — is where judgment actually grows.
Keeping a decision log serves the same purpose. Recording why you chose a particular direction, and what signals led you there, lets you reproduce that same quality of decision six months later, in a similar situation. An experience you don't record doesn't stay an experience — it just becomes time that passed.
How you use AI is also worth examining. Treating AI as an "output generator" and treating it as "a tool that refines my own draft" can look identical day to day, but they produce very different competence curves a year out. The first approach converges toward a state where your speed collapses without AI. The second compounds — the more you use AI, the more your own speed and depth improve together. What determines the direction of your competence isn't how much you depend on the tool, but how you use it.
People who've built enduring brands tend to share something in common: they started out clumsy, and it was inside that clumsiness that they found their own standards. They'll tell you it wasn't the polished output but the repetition of an imperfect process that built a distinctive eye. In an environment where AI hands you a polished result from the start, that process gets skipped — and without it, there's no path to discovering your own standard. Brands, expertise, a distinctive point of view: all of it grows out of that imperfect repetition.
McKinsey asks organizations whether they've woven knowledge management, role design, learning systems, and coaching into a single system. Turn that same question toward a one-person business, and it becomes: where have you designed the path for your own growth? What you choose to fill the space AI leaves behind, once it's taken the grunt work, is what will shape everything that comes after.
