Loyalty programmes are a technology for exploiting three properties of human buyers: that comparing options is tiresome, that habits are cheaper than decisions, and that a small accumulated benefit feels disproportionately like a reason to return.
They work. They have worked for a century. And every one of the three properties belongs to the buyer, not to the transaction — which is why the arrangement becomes unstable the moment the buyer delegates.
What an agent does not have
Software purchasing on a person's behalf does not find comparison tiresome; comparison is the cheapest thing it does. It has no habits, because it has no cost of deciding afresh each time. It does not experience a points balance as a sunk relationship. It has no residual warmth towards the supplier that handled the last order well, unless someone has explicitly encoded that warmth as a parameter.
Strip those three properties out and a loyalty programme reduces to what it always was underneath: a discount, deferred and made conditional. An agent will evaluate it as a discount — net present value, probability of redemption, breakage risk — and will be entirely unmoved by the parts of the programme that were doing the emotional work.
Tier status, in particular, becomes difficult to defend. A gold tier that confers no operational advantage is a signal aimed at a self-image, and an agent has no self-image to flatter.
What survives
Some of the loyalty apparatus does survive delegation, and it is worth being precise about which parts, because they are the parts firms have tended to treat as hygiene rather than as strategy.
Reliability survives, and strengthens. An agent optimising across repeated purchases will weight a supplier's delivery variance, defect rate and dispute-resolution record — and unlike a human buyer, it will not forget a bad experience after six months, nor forgive it because the packaging was nice.
Switching costs that are real survive: integration, accumulated configuration, data that would have to be migrated. These were always the durable component of retention, and the loyalty literature has long distinguished them from the affective kind. Delegation removes the affective kind and leaves the structural kind untouched.
Terms survive, and become more important than their presentation. Better price, better guarantee, better availability. Nothing rhetorical.
The uncomfortable diagnosis
Read the list of what survives and a conclusion is hard to avoid: agents will not destroy loyalty so much as reveal how much of it was never loyalty.
A large share of measured retention has always been a mixture of inertia, comparison cost and mild confusion — which is why the same brands that report high loyalty scores also lose customers rapidly when a competitor makes switching trivial. Removing the friction does not change customer preferences. It exposes what they were.
The firms that will be hurt are the ones whose retention rested on the friction. The firms that will benefit are the ones already competing on the merits, who have been paying for the friction of the others without receiving credit for being better.
What to do about it
Two things, and they point in opposite directions from most current practice.
Make yourself legible to machines. If an agent cannot readily determine your availability, terms, specifications and service record, it will not select you — not because it judged you inferior, but because it could not evaluate you at all. Structured, accurate, machine-readable information is becoming a distribution channel, and the firms treating it as a technical chore are treating a shelf position as a chore.
Then compete on what an agent measures. Reliability, transparency, the real terms. It is a narrower and less forgiving game than the one loyalty programmes were designed for, and it rewards firms that were already good at the boring parts. That is not the worst outcome an industry has faced.