AI changed the work, not the standards
The useful shift is not lowering the bar. It is spending less time on friction and more time on judgement.
AI has changed the work. It has not changed what good work is.
That distinction gets lost because the tools are loud. They can write, draw, code, summarise, translate, refactor, and produce a passable first version of almost anything with a text box attached to it. That is a big deal. It is also not the same as taste, strategy, craft, responsibility, or care.
The standards did not go away. The route to meeting them changed.
The friction moved
A lot of old friction has become optional.
Starting from a blank page is easier. Turning notes into a structure is easier. Finding a first version of a component is easier. Asking the codebase a question is easier. Producing three rough directions before a meeting is easier.
That is useful because friction is expensive. It eats energy before the real problem has even appeared.
But removing friction does not remove the need to decide. In some ways it increases it. When the tool can produce more options, someone has to know which options are worth keeping. When the tool can generate code quickly, someone has to know whether the code belongs in the project. When the tool can produce confident prose, someone has to know whether it is true.
The bottleneck moves from production to judgement.
Average got cheaper
AI is very good at the middle of the bell curve. It can produce work that looks familiar, reads smoothly, and follows common patterns. For some jobs, that is enough. A lot of internal admin does not need a bespoke essay wearing a linen shirt.
For public work, product work, brand work, and tools people actually depend on, average is still average.
The web is already full of pages that are technically fine and emotionally absent. AI can make more of them. It can also help make better things if the person using it has a clear standard for what better means.
That standard has to come from the work itself: the audience, the constraints, the tone, the data, the edge cases, the business model, the thing the user is actually trying to get done.
The useful work became more visible
The part of the job that used to hide inside production is now easier to see.
A client might have thought they were paying for a website because websites require labour. Now the labour is faster, so the real value becomes clearer: understanding the business, shaping the offer, deciding what to leave out, making the interface trustworthy, designing the fallback states, writing the sentence that does not sound like every other sentence on the internet.
That is good news for work with a point of view.
It is less comfortable for work that relied on mystique. If the value was mostly that the tools were hard to use, AI is a problem. If the value is judgement, taste, systems thinking, and accountability, AI is leverage.
Responsibility did not transfer
This is the bit I keep coming back to: the person shipping the work still owns the work.
If an AI tool invents a claim, publishes inaccessible markup, breaks a flow, misreads a brief, or produces copy that sounds confident but hollow, the tool is not going to sit in the client meeting and explain itself. The responsibility lands with the person who chose to use it.
That is not a reason to avoid AI. It is a reason to use it with adult supervision.
Ask it to draft. Ask it to compare. Ask it to test. Ask it to find the awkward edge case you missed. Then check the output against the actual standard of the project.
The bar can go up
The optimistic version is not that AI lets everyone care less.
The optimistic version is that it gives careful people more room to care about the parts that matter. Less time moving boilerplate around. More time improving the interaction. Less time formatting notes. More time thinking about whether the concept is true. Less time generating the obvious option. More time finding the useful one.
That is the version I am interested in.
The tools will keep changing. The standard is simpler and more stubborn: does the work help, does it hold up, and would I still put my name on it after the shortcut has worn off?
The usefulness test is still the test I trust.