AI Came for the Rate, Not the Job

AI Came for the Rate, Not the Job

A few years ago, a client hired a developer, a designer, or a writer and assumed they were paying for expertise. The rate was the rate. Nobody opened the conversation by asking whether the work needed to be done by a person at all.

Now a good number of conversations start somewhere near "can't AI do this?"

I want to be careful here, because I am the last person who gets to act offended by that question. I have AI writing my first drafts. I have it running my deploys. When I catch myself doing the same task by hand twice in a week, I stop and turn it into a skill so I never do it manually again. I have spent the last stretch of my career deliberately handing the repetitive parts of my job to a machine.

So when a client wonders out loud whether a machine could do the thing they used to pay me for, they are asking a fair question. Pretending it is an insult is how you lose the room.

The honest answer is more interesting than either "yes, obviously" or "never, I am irreplaceable." AI did not come for the job. It came for the rate. And it is worth understanding exactly which part of the rate it took.

What AI actually repriced

Every piece of work I have ever been paid for was really two things stacked on top of each other. There was the mechanical part: the typing, the boilerplate, the wrapper code, the first pass, the fifth near-identical config file. And there was the part that decided whether any of that mechanical work was the right thing to do at all.

For twenty years, clients paid me for both in a single number, because they came bundled. You could not get the judgment without also buying the hours of typing that carried it.

AI unbundled them. The mechanical layer, the part that used to eat most of the clock, now costs almost nothing. A first draft that took an afternoon takes a few minutes. Scaffolding that took a morning takes one prompt. This is not a threat I am bravely facing down. It is a fact I use every single day, on purpose, because that work was always the least valuable thing I did. It was rent I paid to get to the interesting problem.

So when a client senses that the work got faster and expects the price to follow, they are half right. The mechanical half genuinely did collapse in value. If your rate was mostly rent on being able to type competently, that rent is coming down, and no amount of indignation stops it.

The mistake is assuming the other half moved with it.

The part the invoice never itemised

Here is the trap. For all those years, the valuable half of the work was invisible. It never appeared on an invoice. Clients did not pay a line item for "knew this API would bite us in six months" or "talked you out of the feature that would have doubled your support load." They paid for hours, and the judgment rode along inside the hours, unnamed and unpriced.

That worked fine right up until the hours got cheap. The moment AI stripped away the visible, mechanical layer, it exposed something a lot of us never had to say out loud: what were you actually paying for? If the honest answer is "the typing," then yes, that person is in trouble. If the honest answer is "someone who owns whether this is right," then nothing about that changed, except that now you have to name it.

That is the real work of this moment, and it is not technical. It is learning to charge for the thing that was always the point but never had a price tag on it. The judgment. The taste. The willingness to say no. The person on the other end of the problem who has seen this shape before and knows where it goes.

Knowing when the confident answer is wrong

Let me put a sharper point on it, because there is one skill that AI made more valuable, not less, and I do not think it gets said enough.

AI is confidently, fluently, articulately wrong on a regular basis. Not obviously wrong, which would be easy. Plausibly wrong. The kind of wrong that reads perfectly, compiles, demos clean, and quietly carries a decision that will cost you three weeks in two months. The output looks exactly as polished when it is right as when it is subtly broken. That is the whole problem.

Catching that does not come from a prompt. It comes from the twenty years. From having shipped the thing that looked fine and watching it fall over in production. From the certificate error four layers downstream of the real bug. The reason I can hand so much of my work to a model is precisely that I can tell when the model is lying to me, and I can tell because I have been burned enough times to recognise the smell.

That is the strange inversion of this whole era. The more you automate, the more valuable the judgment that supervises the automation becomes, because it is now the only thing standing between a confident wrong answer and a shipped mistake. The person who can operate the machine and knows when to overrule it is worth more than they were, not less. AI did not replace that expertise. It made it the entire job.

Stop billing for hours, start billing for the 2am phone call

Which is why the freelancers who are struggling right now are, almost without exception, the ones still selling time.

If your offer is "I will spend hours on this," you are now competing directly with a machine that has infinite hours and charges by the token. You will lose that race, and you should. It is not a race worth being in.

The offer that survives is a different shape entirely. It is not "I will do the work." It is "this will be right, and when it breaks at 2am, I am the one who picks up the phone." That is accountability, and it is the one thing you genuinely cannot buy from a model.

AI does not carry the risk. It does not stake its name on the outcome. It cannot be the person a client trusts to own the result, because owning a result means being reachable and answerable when the result goes wrong. A model is neither.

So I have stopped selling the hours and started selling the outcome, and the ownership of it. The AI is just how the outcome gets built faster. It is a power tool in the workshop, not the carpenter, and nobody hires a carpenter because they own a nice saw.

The economics changed, the job didn't

AI has not replaced my job. It has changed the economics of it, and I think that is the more useful way to see the whole thing.

If you were renting out a commodity skill, the rate is coming down, and the sooner you stop fighting that the sooner you can move. But if you were selling judgment, accountability, and the ability to know when the confident answer is wrong, this is not a downgrade. It is a raise. The machine took over the part you never enjoyed and left you with the part you were actually good at, and it made that part scarcer and more valuable than it has ever been.

The client asking "can't AI do this?" is not the enemy. They are handing you the exact opening to explain what they were really paying for all along. Most of us just never had to say it out loud before.

Have you noticed the conversation about rates changing since AI showed up?