The Library · Perspective
Enhancement is the larger market
Replacement is the story being told and augmentation is where the value is. The argument is economic rather than moral, which is why it is worth making carefully.
The replacement framing dominates the conversation, and it is both the less interesting outcome and, we think, the smaller market. That claim is worth making on economic grounds rather than reassuring ones, because reassurance has a poor record of surviving contact with a budget.
The economics of the two positions
Replacing a role means capturing some fraction of one salary, and requires taking on the whole of that role — including the accountability, the rare cases, and the judgement calls that were never written down. The product must be right essentially all of the time, because there is no longer anyone positioned to catch it.
Enhancing a role means making a person more effective. It requires capturing only the portion of the work that is well-specified, it can be wrong sometimes because there is someone to catch it, and it addresses everyone in the role rather than the subset whose work is fully automatable.
The second is a better business. It is also considerably easier to build, which is why the products that have found durable commercial footing are overwhelmingly of that kind, whatever their marketing says.
What the pattern has been
The historical analogy usually offered is the spreadsheet, and it is a good one. Bookkeeping did not disappear; the ledger-ruling and the arithmetic did. What remained was the part that required judgement — and the profession grew, because the cost of asking a financial question fell far enough that people asked many more of them.
The pattern is not that the job survives untouched. It is that the job decomposes into tasks, some tasks automate, and the role recomposes around what is left. What is left is reliably the part involving accountability, ambiguity, and other people.
One column captured entirely, against a modest lift applied across the whole population. The areas are the argument, and they are not close.
What humans are actually bringing
Stated precisely enough to design against, rather than as a compliment.
Knowing which question to ask. Systems answer. Deciding what is worth answering requires a model of what matters, and that model comes from a stake in the outcome.
Being accountable. Someone must be answerable when it goes wrong, with authority to change course. This cannot be delegated to a system; it can only be delegated to a person who deployed one.
Judging under genuine ambiguity. Not uncertainty with a distribution — ambiguity, where the criteria themselves are contested and the work is deciding which criteria apply.
Holding context nobody wrote down. The politics, the history, the reason the obvious solution was tried in 2019 and abandoned.
Bearing responsibility for other people. Teaching, caring, managing, negotiating. The value depends on a person being answerable to another person.
Why the framing matters commercially
A team that believes it is building a replacement designs for full autonomy, hides its uncertainty, and treats human review as a temporary embarrassment to be engineered away. It ends up with a system that is difficult to supervise precisely where supervision matters most.
A team that believes it is building enhancement designs for the handover: it exposes uncertainty, makes its reasoning inspectable, and treats the human's time as the scarce resource to be spent carefully. That system is more useful on the day it ships and considerably more useful two years later.