We taught the machines to paint before we taught them to carry the boxes

There is something quietly absurd about the current moment in AI. The technology's most visible achievements are paintings, songs and poems - the activities humans have done for pleasure since we first put ochre on cave walls. Meanwhile, somewhere tonight, a qualified accountant is manually ticking through a creditors ledger, a pensions administrator is re-keying member data between systems, and a remediation analyst is reading their four-hundredth identical case file of the month.

We built machines that make art and left the humans carrying the boxes.

Creativity was never the problem to solve

Nobody became a chartered accountant dreaming of duplicate invoice checks. Nobody entered pensions because they loved reconciling GMP records. People enter professions for the judgement, the client relationships, the moments where expertise actually changes an outcome - and then spend remarkable proportions of their working lives on mechanical verification that a machine should have absorbed years ago.

The promise of automation was never that machines would write our symphonies. It was that they would do the heavy lifting so we had time to.

The cultural argument matters as much as the commercial one. Art, music and writing are not inefficiencies to be optimised away - they are the point of having time at all. A technology agenda that automates human joy while leaving human drudgery intact has simply misread its brief.

The heavy lifting is the harder engineering

Here is the uncomfortable truth the generative wave glosses over: making a plausible image is, in engineering terms, more forgiving than finding the one misposted journal in ten thousand. Art tolerates infinite wrong answers. A regulated review tolerates none. Doing the heavy lifting properly demands statistical rigour, explainability, audit trails and benchmarked accuracy against senior professional judgement - which is precisely why so much of the industry chose the easier, flashier path.

We made the opposite choice. Oversight exists to take the mechanical weight out of professional work: the scanning, the cross-checking, the anomaly-hunting across thousands of field relationships that humans were never built for. The professional keeps what humans are actually for - judgement, scepticism, the signature on the file - and gets their evenings back.

Let the machines carry the boxes. The symphonies were always ours.