Each successive automation model brings new areas of functionality. As AI's capabilities increase, so does what we perceive as competence.
That's only half the picture. It's not just about the degree of replicability of a task or job. It's also about the perception of excellence by the consumer.
Think about the work being automated the fastest: customer service, coding, content production, all forms of clerical work. People in each of those professions would no doubt tell you they can discern which people are merely adequate at their job and which are truly excellent. This knowledge is possessed only by those who've actually done the work. They're not lying, but excellence usually isn't perceptible to those outside the field at all. Customer problems get solved or they don't. Code works or it doesn't. Performance marketing copy either converts or it doesn't. Virtuosity is blurred by a binary outcome.
By contrast, the perception of excellence is tantamount to the product's success in other professions. A film director's excellence (or lack thereof) is quite perceptible; what makes a movie worth seeing or not isn't a binary question. So is the difference between a great criminal attorney and an average one. That last one is especially illustrative because, even within the same professional umbrella (attorneys), there is wide variation in non-lawyers' ability to perceive excellence between, say, a securities lawyer and a trial attorney.
The true constraint we're headed toward is therefore not the amount of excellence in the workforce, but the number of customers who can perceive excellence. The more perceptible excellence is, the less likely that work is to be automated, no matter how advanced the models become.
