Why AI Came for the Professions First

Every previous wave of automation moved from the bottom of the wage scale upward. Looms before clerks. Assembly lines before accountants. The advice that followed each wave was the same and it was reasonable: get educated, get credentialed, get into work that requires judgment rather than hands.

This one arrived from the other direction. AI came for the professions first, and the people most exposed are the ones who did exactly what they were told.

Why did AI reach the professions first?

Because the work was already made of text.

A model learns from written material and produces written material. That is the whole shape of it. So the occupations most exposed are the ones whose output is documents, analysis, summaries, drafts, code and correspondence. Law. Accounting.

Marketing. Consulting. Software. Financial advice, including mine.

Meanwhile the electrician crawling through a roof cavity in a Melbourne summer is doing something no model can currently touch. Not because the task is intellectually harder, but because it happens in a body, in a physical space, with judgment applied to a situation that has never occurred in quite that form before. The robotics required to replace him are a genuinely unsolved problem and have been for forty years. The software required to draft his lawyer’s letter of advice arrived in about eighteen months.

Research from OpenAI and the University of Pennsylvania in 2023, published as “GPTs are GPTs”, mapped occupational exposure to large language models and found that higher-wage, higher-education occupations showed greater exposure than lower-wage ones. That is a reversal of every technological shift in the last two centuries, and it has not fully landed yet in how people plan their working lives.

One detail in that work matters more than the headline. The analysis was done at the level of tasks, not jobs. Very few occupations are entirely exposed. Most are made of a mix, some parts of which are highly automatable and some of which are not touched at all. That distinction is the whole story, because it tells you where inside your own work the value is moving.

What was knowledge wealth actually worth?

For about a century, knowledge wealth was the safest asset a person could hold. You spent years acquiring it, it was expensive to replicate, and it produced income for four decades. The degree, the CPA, the practising certificate. These were not simply qualifications. They were positions in a market where scarce information could be sold.

That scarcity is what has changed. The information itself is now close to free and instantly available at a quality that was, until very recently, professional grade. What you know is worth less than it was. Not nothing, but less.

It helps to be precise about what has moved, because the loose version of this argument is what makes sensible people dismiss it.

Recall of information has been commoditised almost entirely.

Structured application of known rules to a described situation has been heavily commoditised. Drafting has been commoditised. What has not moved is the ability to work out which question is actually being asked, to sit with a person while they decide something that will shape the next decade of their life, and to be the one they trusted enough to ask.

This is a repricing, not a disappearance. And it matters because knowledge wealth was the form that most professionals concentrated almost everything into. Twenty years of building one asset, on the understanding that it would hold.

Why concentration is the real exposure

Here is where the connection between forms becomes practical rather than theoretical.

Knowledge wealth on its own is now a fragile position. Knowledge wealth attached to relational wealth is not, because someone still has to be trusted enough to be asked. Knowledge attached to adaptive wealth is not, because the person who can reconfigure what they do every eighteen months keeps a floor under themselves.

Knowledge attached to reputational wealth is not, because reputation decides who gets the call in the first place.

The knowledge was never the whole product. It was the part that was easiest to charge for, so it became the part everyone measured, and eventually the part everyone confused for the thing itself.

There is a second shift underneath the first that is worth watching.

The place where people start looking is moving. Someone who needed an accountant in 2015 asked a friend or searched. Increasingly they ask a model, and the model gives them a shortlist.

Whatever you think of that, it changes what makes a professional findable. A body of published thinking, a visible position, a reputation that exists in text somewhere a system can read. That is reputational and technological wealth doing work that used to be done by a referral over coffee. The coffee still matters. It is no longer the only door.

What holds value when knowledge is cheap

Sit in a room where someone is about to sell the business they built over twenty-two years. There is a spreadsheet on the table and the spreadsheet is not the problem. A model can build that spreadsheet, and build it well. What the model cannot do is sit with the man while he works out who he is on the Monday after settlement.

That is not a sentimental point about the human touch. It is a specific claim about where value sits. When the analysis is commoditised, what remains scarce is judgment about what matters, trust built over years, the ability to hold a hard conversation, and presence in a moment a person cannot get through alone.

Those are not soft skills. In the framework I work with they are separate forms of wealth with their own logic. Character. Relational. Communication. Adaptive. They were always there. They were subsidised by the scarcity of information, which meant nobody had to be good at them to make a living.

That subsidy is ending.

The honest counterweight

It would be dishonest to write this without the other side of it, particularly for an audience that has heard a lot of confident predictions.

Capability is not adoption. The gap between what these systems can do in a demonstration and what actually changes inside a regulated professional firm is years wide, and it is made of compliance, liability, procurement, professional standards and the ordinary slowness of institutions. History also suggests the simple replacement story is usually wrong. James Bessen’s work on bank tellers is the well-known example: automated teller machines spread widely and teller employment did not collapse in the way expected, because the cost per branch fell and the number of branches rose.

Read as a caution, it says that automating tasks inside a job does not reliably remove the job.

So the reasonable position is not panic. It is that the composition of professional work is changing faster than the volume of it, and that the parts holding their value are the parts nobody was trained to build.

What this means if you are in one of these professions

The instinct is to get faster. Use the tools, produce more, protect the margin. That is a reasonable short-term move and it is not a strategy, because everyone in your field has access to the same tools and the gains get competed away inside two years.

The harder question is the one worth sitting with. If the friction and complexity you currently get paid to manage both collapse, why would anyone still come to you?

The honest answers are all found outside knowledge wealth. Which means the work is not to defend the credential. It is to build the other forms while the credential still buys you the time to do it.

Most people will not. They will get faster, then get worried, then get faster again. The professionals who come through this well will be the ones who noticed early that they had built a life on a single form of wealth, and quietly started building others.

Book Your Clarity Call